diff --git "a/stage4_ckpt/trainer_state.json" "b/stage4_ckpt/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/stage4_ckpt/trainer_state.json"
@@ -0,0 +1,28096 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.99656041275047,
+  "global_step": 46800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.4126,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 1.0084,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-05,
+      "loss": 0.9879,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 0.823,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-05,
+      "loss": 0.7661,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-05,
+      "loss": 0.8244,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7e-05,
+      "loss": 0.7085,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 0.8008,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-05,
+      "loss": 0.7042,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001,
+      "loss": 0.698,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.997862334330911e-05,
+      "loss": 0.7638,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995724668661823e-05,
+      "loss": 0.6782,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.993587002992733e-05,
+      "loss": 0.7541,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991449337323642e-05,
+      "loss": 0.6904,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989311671654553e-05,
+      "loss": 0.6709,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987174005985465e-05,
+      "loss": 0.7488,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985036340316375e-05,
+      "loss": 0.6562,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982898674647285e-05,
+      "loss": 0.7306,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.980761008978197e-05,
+      "loss": 0.6881,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.978623343309106e-05,
+      "loss": 0.6585,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.976485677640017e-05,
+      "loss": 0.7287,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.974348011970929e-05,
+      "loss": 0.6576,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.972210346301839e-05,
+      "loss": 0.7271,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.97007268063275e-05,
+      "loss": 0.6686,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.96793501496366e-05,
+      "loss": 0.662,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965797349294572e-05,
+      "loss": 0.7368,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.963659683625481e-05,
+      "loss": 0.6588,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.961522017956391e-05,
+      "loss": 0.7176,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.959384352287303e-05,
+      "loss": 0.6552,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.957246686618214e-05,
+      "loss": 0.6416,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.955109020949124e-05,
+      "loss": 0.7067,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.952971355280036e-05,
+      "loss": 0.653,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.950833689610945e-05,
+      "loss": 0.7157,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.948696023941855e-05,
+      "loss": 0.6628,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.946558358272766e-05,
+      "loss": 0.6371,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.944420692603678e-05,
+      "loss": 0.7153,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.942283026934588e-05,
+      "loss": 0.6461,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.940145361265499e-05,
+      "loss": 0.6987,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.938007695596409e-05,
+      "loss": 0.6507,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93587002992732e-05,
+      "loss": 0.6305,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93373236425823e-05,
+      "loss": 0.7125,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.93159469858914e-05,
+      "loss": 0.6369,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.929457032920052e-05,
+      "loss": 0.7018,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.927319367250963e-05,
+      "loss": 0.6541,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.925181701581873e-05,
+      "loss": 0.6331,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.923044035912783e-05,
+      "loss": 0.7067,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.920906370243694e-05,
+      "loss": 0.6293,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.918768704574604e-05,
+      "loss": 0.6981,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.916631038905516e-05,
+      "loss": 0.6445,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.914493373236427e-05,
+      "loss": 0.6381,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.912355707567337e-05,
+      "loss": 0.7076,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.910218041898248e-05,
+      "loss": 0.6407,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.908080376229158e-05,
+      "loss": 0.6953,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.905942710560068e-05,
+      "loss": 0.638,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.903805044890979e-05,
+      "loss": 0.6264,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.901667379221891e-05,
+      "loss": 0.7074,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.899529713552801e-05,
+      "loss": 0.6239,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.897392047883712e-05,
+      "loss": 0.6856,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.895254382214622e-05,
+      "loss": 0.6448,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.893116716545533e-05,
+      "loss": 0.6204,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.890979050876443e-05,
+      "loss": 0.6987,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.888841385207353e-05,
+      "loss": 0.6235,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.886703719538265e-05,
+      "loss": 0.6914,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.884566053869176e-05,
+      "loss": 0.6492,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.882428388200086e-05,
+      "loss": 0.6222,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.880290722530997e-05,
+      "loss": 0.7033,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.878153056861907e-05,
+      "loss": 0.618,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.876015391192817e-05,
+      "loss": 0.6892,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.873877725523728e-05,
+      "loss": 0.6397,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.87174005985464e-05,
+      "loss": 0.63,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.86960239418555e-05,
+      "loss": 0.6927,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.86746472851646e-05,
+      "loss": 0.626,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.865327062847371e-05,
+      "loss": 0.6833,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.863189397178282e-05,
+      "loss": 0.6318,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.861051731509192e-05,
+      "loss": 0.6204,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.858914065840104e-05,
+      "loss": 0.6885,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.856776400171014e-05,
+      "loss": 0.6276,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.854638734501925e-05,
+      "loss": 0.6788,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.852501068832835e-05,
+      "loss": 0.643,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.850363403163746e-05,
+      "loss": 0.6186,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.848225737494656e-05,
+      "loss": 0.6965,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.846088071825567e-05,
+      "loss": 0.6322,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.843950406156478e-05,
+      "loss": 0.6845,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.841812740487389e-05,
+      "loss": 0.6313,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.839675074818299e-05,
+      "loss": 0.6157,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.83753740914921e-05,
+      "loss": 0.691,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.83539974348012e-05,
+      "loss": 0.6233,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.83326207781103e-05,
+      "loss": 0.6836,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.831124412141941e-05,
+      "loss": 0.6226,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.828986746472853e-05,
+      "loss": 0.6184,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.826849080803763e-05,
+      "loss": 0.688,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.824711415134674e-05,
+      "loss": 0.6189,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.822573749465584e-05,
+      "loss": 0.6648,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.820436083796495e-05,
+      "loss": 0.6359,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.818298418127405e-05,
+      "loss": 0.6218,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.816160752458316e-05,
+      "loss": 0.6822,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.814023086789227e-05,
+      "loss": 0.6069,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.811885421120138e-05,
+      "loss": 0.6762,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.809747755451047e-05,
+      "loss": 0.6256,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.807610089781959e-05,
+      "loss": 0.6233,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.805472424112869e-05,
+      "loss": 0.686,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.80333475844378e-05,
+      "loss": 0.6155,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.801197092774691e-05,
+      "loss": 0.6799,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.799059427105602e-05,
+      "loss": 0.6234,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.796921761436512e-05,
+      "loss": 0.6027,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.794784095767421e-05,
+      "loss": 0.6891,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.792646430098333e-05,
+      "loss": 0.6229,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.790508764429244e-05,
+      "loss": 0.6803,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.788371098760154e-05,
+      "loss": 0.6208,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.786233433091066e-05,
+      "loss": 0.616,
+      "step": 1100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.784095767421976e-05,
+      "loss": 0.689,
+      "step": 1110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.781958101752885e-05,
+      "loss": 0.6064,
+      "step": 1120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.779820436083797e-05,
+      "loss": 0.6776,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.777682770414708e-05,
+      "loss": 0.6283,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.775545104745618e-05,
+      "loss": 0.6218,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.773407439076529e-05,
+      "loss": 0.677,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.77126977340744e-05,
+      "loss": 0.6142,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.76913210773835e-05,
+      "loss": 0.6822,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.76699444206926e-05,
+      "loss": 0.6278,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.764856776400172e-05,
+      "loss": 0.6112,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.762719110731082e-05,
+      "loss": 0.6853,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.760581445061993e-05,
+      "loss": 0.6106,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.758443779392903e-05,
+      "loss": 0.6801,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.756306113723815e-05,
+      "loss": 0.6291,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.754168448054724e-05,
+      "loss": 0.6107,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.752030782385634e-05,
+      "loss": 0.6775,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.749893116716546e-05,
+      "loss": 0.6075,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.747755451047457e-05,
+      "loss": 0.6648,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.745617785378367e-05,
+      "loss": 0.6302,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.743480119709279e-05,
+      "loss": 0.6226,
+      "step": 1300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.741342454040188e-05,
+      "loss": 0.6778,
+      "step": 1310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.739204788371099e-05,
+      "loss": 0.6204,
+      "step": 1320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.737067122702009e-05,
+      "loss": 0.673,
+      "step": 1330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.734929457032921e-05,
+      "loss": 0.6294,
+      "step": 1340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.732791791363831e-05,
+      "loss": 0.5981,
+      "step": 1350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.730654125694742e-05,
+      "loss": 0.69,
+      "step": 1360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.728516460025654e-05,
+      "loss": 0.6094,
+      "step": 1370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.726378794356563e-05,
+      "loss": 0.6766,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.724241128687473e-05,
+      "loss": 0.6344,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.722103463018385e-05,
+      "loss": 0.5995,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.719965797349295e-05,
+      "loss": 0.6793,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.717828131680206e-05,
+      "loss": 0.6133,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.715690466011116e-05,
+      "loss": 0.6735,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.713552800342027e-05,
+      "loss": 0.6255,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.711415134672937e-05,
+      "loss": 0.6047,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.709277469003848e-05,
+      "loss": 0.6711,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.70713980333476e-05,
+      "loss": 0.6094,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.70500213766567e-05,
+      "loss": 0.6647,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.70286447199658e-05,
+      "loss": 0.6187,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.700726806327491e-05,
+      "loss": 0.6017,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.698589140658401e-05,
+      "loss": 0.6775,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.696451474989312e-05,
+      "loss": 0.6178,
+      "step": 1520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.694313809320222e-05,
+      "loss": 0.6752,
+      "step": 1530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.692176143651134e-05,
+      "loss": 0.6096,
+      "step": 1540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.690038477982044e-05,
+      "loss": 0.6106,
+      "step": 1550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.687900812312955e-05,
+      "loss": 0.6788,
+      "step": 1560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.685763146643865e-05,
+      "loss": 0.5967,
+      "step": 1570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.683625480974776e-05,
+      "loss": 0.6676,
+      "step": 1580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.681487815305686e-05,
+      "loss": 0.6145,
+      "step": 1590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.679350149636597e-05,
+      "loss": 0.5979,
+      "step": 1600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.677212483967508e-05,
+      "loss": 0.6844,
+      "step": 1610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.675074818298419e-05,
+      "loss": 0.6105,
+      "step": 1620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.672937152629329e-05,
+      "loss": 0.6555,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.67079948696024e-05,
+      "loss": 0.6192,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.66866182129115e-05,
+      "loss": 0.6031,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.66652415562206e-05,
+      "loss": 0.672,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.664386489952972e-05,
+      "loss": 0.6023,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.662248824283883e-05,
+      "loss": 0.6692,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.660111158614793e-05,
+      "loss": 0.6254,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.657973492945704e-05,
+      "loss": 0.606,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.655835827276614e-05,
+      "loss": 0.6759,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.653698161607525e-05,
+      "loss": 0.6125,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.651560495938435e-05,
+      "loss": 0.6554,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.649422830269347e-05,
+      "loss": 0.6149,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.647285164600257e-05,
+      "loss": 0.603,
+      "step": 1750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.645147498931168e-05,
+      "loss": 0.6846,
+      "step": 1760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.643009833262078e-05,
+      "loss": 0.6086,
+      "step": 1770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.640872167592989e-05,
+      "loss": 0.6677,
+      "step": 1780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.638734501923899e-05,
+      "loss": 0.6114,
+      "step": 1790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.63659683625481e-05,
+      "loss": 0.6115,
+      "step": 1800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.634459170585721e-05,
+      "loss": 0.6847,
+      "step": 1810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.632321504916632e-05,
+      "loss": 0.6067,
+      "step": 1820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.630183839247542e-05,
+      "loss": 0.6537,
+      "step": 1830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.628046173578453e-05,
+      "loss": 0.6172,
+      "step": 1840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.625908507909363e-05,
+      "loss": 0.6026,
+      "step": 1850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.623770842240274e-05,
+      "loss": 0.6696,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.621633176571184e-05,
+      "loss": 0.6133,
+      "step": 1870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.619495510902096e-05,
+      "loss": 0.6602,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.617357845233006e-05,
+      "loss": 0.61,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.615220179563917e-05,
+      "loss": 0.6035,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.613082513894827e-05,
+      "loss": 0.6798,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.610944848225738e-05,
+      "loss": 0.6032,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.608807182556648e-05,
+      "loss": 0.6576,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.60666951688756e-05,
+      "loss": 0.6136,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.60453185121847e-05,
+      "loss": 0.5976,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.602394185549381e-05,
+      "loss": 0.6673,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.60025651988029e-05,
+      "loss": 0.6047,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.598118854211202e-05,
+      "loss": 0.6624,
+      "step": 1980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.595981188542112e-05,
+      "loss": 0.6161,
+      "step": 1990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.593843522873023e-05,
+      "loss": 0.5997,
+      "step": 2000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.591705857203935e-05,
+      "loss": 0.6689,
+      "step": 2010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.589568191534845e-05,
+      "loss": 0.6012,
+      "step": 2020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.587430525865755e-05,
+      "loss": 0.6641,
+      "step": 2030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.585292860196665e-05,
+      "loss": 0.6058,
+      "step": 2040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.583155194527576e-05,
+      "loss": 0.6037,
+      "step": 2050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.581017528858487e-05,
+      "loss": 0.6592,
+      "step": 2060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.578879863189397e-05,
+      "loss": 0.6061,
+      "step": 2070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.576742197520309e-05,
+      "loss": 0.6532,
+      "step": 2080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.57460453185122e-05,
+      "loss": 0.6204,
+      "step": 2090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.572466866182129e-05,
+      "loss": 0.5996,
+      "step": 2100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.57032920051304e-05,
+      "loss": 0.6717,
+      "step": 2110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.568191534843951e-05,
+      "loss": 0.606,
+      "step": 2120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.566053869174861e-05,
+      "loss": 0.6578,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.563916203505772e-05,
+      "loss": 0.6095,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.561778537836684e-05,
+      "loss": 0.6019,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.559640872167594e-05,
+      "loss": 0.6782,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.557503206498503e-05,
+      "loss": 0.5977,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.555365540829415e-05,
+      "loss": 0.6533,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.553227875160325e-05,
+      "loss": 0.6116,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.551090209491236e-05,
+      "loss": 0.5907,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.548952543822148e-05,
+      "loss": 0.6669,
+      "step": 2210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.546814878153058e-05,
+      "loss": 0.5951,
+      "step": 2220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.544677212483967e-05,
+      "loss": 0.6599,
+      "step": 2230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.542539546814878e-05,
+      "loss": 0.6153,
+      "step": 2240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.54040188114579e-05,
+      "loss": 0.6057,
+      "step": 2250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.5382642154767e-05,
+      "loss": 0.6661,
+      "step": 2260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.53612654980761e-05,
+      "loss": 0.5933,
+      "step": 2270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.533988884138522e-05,
+      "loss": 0.6529,
+      "step": 2280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.531851218469433e-05,
+      "loss": 0.6165,
+      "step": 2290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.529713552800342e-05,
+      "loss": 0.5967,
+      "step": 2300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.527575887131252e-05,
+      "loss": 0.6707,
+      "step": 2310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.525438221462164e-05,
+      "loss": 0.6014,
+      "step": 2320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.523300555793074e-05,
+      "loss": 0.6573,
+      "step": 2330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.521162890123985e-05,
+      "loss": 0.6105,
+      "step": 2340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.519025224454897e-05,
+      "loss": 0.5912,
+      "step": 2350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.516887558785806e-05,
+      "loss": 0.6606,
+      "step": 2360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.514749893116716e-05,
+      "loss": 0.5957,
+      "step": 2370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.512612227447628e-05,
+      "loss": 0.6523,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.510474561778538e-05,
+      "loss": 0.6125,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.508336896109449e-05,
+      "loss": 0.5934,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.50619923044036e-05,
+      "loss": 0.6668,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.50406156477127e-05,
+      "loss": 0.6003,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.50192389910218e-05,
+      "loss": 0.6554,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.499786233433091e-05,
+      "loss": 0.6118,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.497648567764003e-05,
+      "loss": 0.5913,
+      "step": 2450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.495510902094913e-05,
+      "loss": 0.6718,
+      "step": 2460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.493373236425823e-05,
+      "loss": 0.601,
+      "step": 2470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.491235570756735e-05,
+      "loss": 0.658,
+      "step": 2480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.489097905087644e-05,
+      "loss": 0.6048,
+      "step": 2490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.486960239418555e-05,
+      "loss": 0.5941,
+      "step": 2500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.484822573749465e-05,
+      "loss": 0.661,
+      "step": 2510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.482684908080377e-05,
+      "loss": 0.5939,
+      "step": 2520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.480547242411288e-05,
+      "loss": 0.663,
+      "step": 2530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.478409576742198e-05,
+      "loss": 0.6024,
+      "step": 2540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.476271911073108e-05,
+      "loss": 0.5997,
+      "step": 2550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.474134245404019e-05,
+      "loss": 0.6667,
+      "step": 2560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.47199657973493e-05,
+      "loss": 0.5929,
+      "step": 2570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.46985891406584e-05,
+      "loss": 0.656,
+      "step": 2580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.467721248396752e-05,
+      "loss": 0.603,
+      "step": 2590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.465583582727662e-05,
+      "loss": 0.587,
+      "step": 2600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.463445917058572e-05,
+      "loss": 0.6677,
+      "step": 2610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.461308251389483e-05,
+      "loss": 0.5948,
+      "step": 2620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.459170585720393e-05,
+      "loss": 0.6529,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.457032920051304e-05,
+      "loss": 0.614,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.454895254382216e-05,
+      "loss": 0.5877,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.452757588713126e-05,
+      "loss": 0.6607,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.450619923044037e-05,
+      "loss": 0.5967,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.448482257374947e-05,
+      "loss": 0.6532,
+      "step": 2680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.446344591705857e-05,
+      "loss": 0.6089,
+      "step": 2690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.444206926036768e-05,
+      "loss": 0.5906,
+      "step": 2700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.442069260367678e-05,
+      "loss": 0.6591,
+      "step": 2710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.43993159469859e-05,
+      "loss": 0.5896,
+      "step": 2720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.4377939290295e-05,
+      "loss": 0.6447,
+      "step": 2730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.435656263360411e-05,
+      "loss": 0.6124,
+      "step": 2740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.433518597691322e-05,
+      "loss": 0.5872,
+      "step": 2750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.431380932022232e-05,
+      "loss": 0.6651,
+      "step": 2760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.429243266353142e-05,
+      "loss": 0.6045,
+      "step": 2770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.427105600684053e-05,
+      "loss": 0.6543,
+      "step": 2780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.424967935014965e-05,
+      "loss": 0.6058,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.422830269345875e-05,
+      "loss": 0.5967,
+      "step": 2800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.420692603676786e-05,
+      "loss": 0.6558,
+      "step": 2810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.418554938007696e-05,
+      "loss": 0.5979,
+      "step": 2820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.416417272338606e-05,
+      "loss": 0.6561,
+      "step": 2830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.414279606669517e-05,
+      "loss": 0.6006,
+      "step": 2840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.412141941000427e-05,
+      "loss": 0.5965,
+      "step": 2850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.410004275331339e-05,
+      "loss": 0.6693,
+      "step": 2860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.40786660966225e-05,
+      "loss": 0.5925,
+      "step": 2870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.40572894399316e-05,
+      "loss": 0.6511,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.40359127832407e-05,
+      "loss": 0.6053,
+      "step": 2890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.401453612654981e-05,
+      "loss": 0.5906,
+      "step": 2900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.399315946985891e-05,
+      "loss": 0.6595,
+      "step": 2910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.397178281316803e-05,
+      "loss": 0.5942,
+      "step": 2920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.395040615647714e-05,
+      "loss": 0.6487,
+      "step": 2930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.392902949978624e-05,
+      "loss": 0.6089,
+      "step": 2940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.390765284309535e-05,
+      "loss": 0.593,
+      "step": 2950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.388627618640445e-05,
+      "loss": 0.6626,
+      "step": 2960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.386489952971356e-05,
+      "loss": 0.5995,
+      "step": 2970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.384352287302266e-05,
+      "loss": 0.6542,
+      "step": 2980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.382214621633178e-05,
+      "loss": 0.6091,
+      "step": 2990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.380076955964088e-05,
+      "loss": 0.584,
+      "step": 3000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.377939290294999e-05,
+      "loss": 0.6548,
+      "step": 3010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.375801624625909e-05,
+      "loss": 0.5847,
+      "step": 3020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.37366395895682e-05,
+      "loss": 0.6549,
+      "step": 3030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.37152629328773e-05,
+      "loss": 0.6026,
+      "step": 3040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.36938862761864e-05,
+      "loss": 0.5814,
+      "step": 3050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.367250961949552e-05,
+      "loss": 0.6578,
+      "step": 3060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.365113296280463e-05,
+      "loss": 0.59,
+      "step": 3070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.362975630611373e-05,
+      "loss": 0.6517,
+      "step": 3080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.360837964942284e-05,
+      "loss": 0.6062,
+      "step": 3090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.358700299273194e-05,
+      "loss": 0.5843,
+      "step": 3100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.356562633604105e-05,
+      "loss": 0.657,
+      "step": 3110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.354424967935015e-05,
+      "loss": 0.5895,
+      "step": 3120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.352287302265927e-05,
+      "loss": 0.6563,
+      "step": 3130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.350149636596837e-05,
+      "loss": 0.5949,
+      "step": 3140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.348011970927746e-05,
+      "loss": 0.5849,
+      "step": 3150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.345874305258658e-05,
+      "loss": 0.6568,
+      "step": 3160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.343736639589569e-05,
+      "loss": 0.5969,
+      "step": 3170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.341598973920479e-05,
+      "loss": 0.6429,
+      "step": 3180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.339461308251391e-05,
+      "loss": 0.6022,
+      "step": 3190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.337323642582301e-05,
+      "loss": 0.5984,
+      "step": 3200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.33518597691321e-05,
+      "loss": 0.6602,
+      "step": 3210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.333048311244121e-05,
+      "loss": 0.5919,
+      "step": 3220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.330910645575033e-05,
+      "loss": 0.6588,
+      "step": 3230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.328772979905943e-05,
+      "loss": 0.5954,
+      "step": 3240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.326635314236854e-05,
+      "loss": 0.5893,
+      "step": 3250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.324497648567765e-05,
+      "loss": 0.657,
+      "step": 3260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.322359982898676e-05,
+      "loss": 0.5842,
+      "step": 3270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.320222317229585e-05,
+      "loss": 0.6536,
+      "step": 3280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.318084651560497e-05,
+      "loss": 0.6089,
+      "step": 3290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.315946985891407e-05,
+      "loss": 0.5822,
+      "step": 3300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.313809320222318e-05,
+      "loss": 0.6519,
+      "step": 3310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.311671654553228e-05,
+      "loss": 0.5863,
+      "step": 3320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.30953398888414e-05,
+      "loss": 0.6494,
+      "step": 3330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.307396323215049e-05,
+      "loss": 0.6011,
+      "step": 3340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.30525865754596e-05,
+      "loss": 0.5921,
+      "step": 3350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.303120991876871e-05,
+      "loss": 0.664,
+      "step": 3360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.300983326207782e-05,
+      "loss": 0.5861,
+      "step": 3370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.298845660538692e-05,
+      "loss": 0.6516,
+      "step": 3380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.296707994869603e-05,
+      "loss": 0.6073,
+      "step": 3390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.294570329200514e-05,
+      "loss": 0.5918,
+      "step": 3400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.292432663531423e-05,
+      "loss": 0.6567,
+      "step": 3410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.290294997862334e-05,
+      "loss": 0.5793,
+      "step": 3420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.288157332193246e-05,
+      "loss": 0.6401,
+      "step": 3430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.286019666524156e-05,
+      "loss": 0.5984,
+      "step": 3440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.283882000855067e-05,
+      "loss": 0.5841,
+      "step": 3450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.281744335185978e-05,
+      "loss": 0.6481,
+      "step": 3460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.279606669516888e-05,
+      "loss": 0.5909,
+      "step": 3470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.277469003847798e-05,
+      "loss": 0.6462,
+      "step": 3480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.275331338178708e-05,
+      "loss": 0.5925,
+      "step": 3490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.27319367250962e-05,
+      "loss": 0.5825,
+      "step": 3500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.271056006840531e-05,
+      "loss": 0.6616,
+      "step": 3510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.268918341171441e-05,
+      "loss": 0.5844,
+      "step": 3520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.266780675502353e-05,
+      "loss": 0.6455,
+      "step": 3530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.264643009833262e-05,
+      "loss": 0.5955,
+      "step": 3540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.262505344164173e-05,
+      "loss": 0.5768,
+      "step": 3550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.260367678495084e-05,
+      "loss": 0.6591,
+      "step": 3560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.258230012825995e-05,
+      "loss": 0.5865,
+      "step": 3570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.256092347156905e-05,
+      "loss": 0.6469,
+      "step": 3580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.253954681487816e-05,
+      "loss": 0.605,
+      "step": 3590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.251817015818726e-05,
+      "loss": 0.5949,
+      "step": 3600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.249679350149637e-05,
+      "loss": 0.6615,
+      "step": 3610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.247541684480547e-05,
+      "loss": 0.5788,
+      "step": 3620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.245404018811459e-05,
+      "loss": 0.6447,
+      "step": 3630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.243266353142369e-05,
+      "loss": 0.5975,
+      "step": 3640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.24112868747328e-05,
+      "loss": 0.5839,
+      "step": 3650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.23899102180419e-05,
+      "loss": 0.6566,
+      "step": 3660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.2368533561351e-05,
+      "loss": 0.5788,
+      "step": 3670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.234715690466011e-05,
+      "loss": 0.6503,
+      "step": 3680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.232578024796922e-05,
+      "loss": 0.6127,
+      "step": 3690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.230440359127833e-05,
+      "loss": 0.5776,
+      "step": 3700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.228302693458744e-05,
+      "loss": 0.6515,
+      "step": 3710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.226165027789654e-05,
+      "loss": 0.576,
+      "step": 3720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.224027362120565e-05,
+      "loss": 0.6421,
+      "step": 3730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.221889696451475e-05,
+      "loss": 0.5975,
+      "step": 3740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.219752030782386e-05,
+      "loss": 0.5839,
+      "step": 3750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.217614365113296e-05,
+      "loss": 0.6532,
+      "step": 3760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.215476699444208e-05,
+      "loss": 0.5793,
+      "step": 3770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.213339033775118e-05,
+      "loss": 0.6487,
+      "step": 3780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.211201368106029e-05,
+      "loss": 0.5978,
+      "step": 3790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.209063702436939e-05,
+      "loss": 0.5986,
+      "step": 3800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.20692603676785e-05,
+      "loss": 0.6474,
+      "step": 3810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.20478837109876e-05,
+      "loss": 0.5892,
+      "step": 3820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.202650705429672e-05,
+      "loss": 0.6548,
+      "step": 3830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.200513039760582e-05,
+      "loss": 0.589,
+      "step": 3840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.198375374091493e-05,
+      "loss": 0.5966,
+      "step": 3850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.196237708422403e-05,
+      "loss": 0.6635,
+      "step": 3860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.194100042753314e-05,
+      "loss": 0.5753,
+      "step": 3870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.191962377084224e-05,
+      "loss": 0.6437,
+      "step": 3880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.189824711415135e-05,
+      "loss": 0.6072,
+      "step": 3890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.187687045746046e-05,
+      "loss": 0.5745,
+      "step": 3900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.185549380076957e-05,
+      "loss": 0.6521,
+      "step": 3910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.183411714407867e-05,
+      "loss": 0.5728,
+      "step": 3920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.181274048738778e-05,
+      "loss": 0.6489,
+      "step": 3930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.179136383069688e-05,
+      "loss": 0.6041,
+      "step": 3940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.176998717400599e-05,
+      "loss": 0.5781,
+      "step": 3950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.174861051731509e-05,
+      "loss": 0.6585,
+      "step": 3960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.172723386062421e-05,
+      "loss": 0.5896,
+      "step": 3970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.170585720393331e-05,
+      "loss": 0.6493,
+      "step": 3980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.168448054724242e-05,
+      "loss": 0.5997,
+      "step": 3990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.166310389055152e-05,
+      "loss": 0.5892,
+      "step": 4000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.164172723386063e-05,
+      "loss": 0.6706,
+      "step": 4010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.162035057716973e-05,
+      "loss": 0.5901,
+      "step": 4020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.159897392047884e-05,
+      "loss": 0.647,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.157759726378795e-05,
+      "loss": 0.5852,
+      "step": 4040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.155622060709706e-05,
+      "loss": 0.5794,
+      "step": 4050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.153484395040616e-05,
+      "loss": 0.6538,
+      "step": 4060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.151346729371527e-05,
+      "loss": 0.5815,
+      "step": 4070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.149209063702437e-05,
+      "loss": 0.6542,
+      "step": 4080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.147071398033348e-05,
+      "loss": 0.6025,
+      "step": 4090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.14493373236426e-05,
+      "loss": 0.5771,
+      "step": 4100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.14279606669517e-05,
+      "loss": 0.6488,
+      "step": 4110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.14065840102608e-05,
+      "loss": 0.573,
+      "step": 4120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.13852073535699e-05,
+      "loss": 0.6592,
+      "step": 4130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.136383069687901e-05,
+      "loss": 0.5861,
+      "step": 4140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.134245404018812e-05,
+      "loss": 0.5789,
+      "step": 4150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.132107738349722e-05,
+      "loss": 0.6573,
+      "step": 4160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.129970072680634e-05,
+      "loss": 0.5851,
+      "step": 4170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.127832407011544e-05,
+      "loss": 0.649,
+      "step": 4180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.125694741342455e-05,
+      "loss": 0.6037,
+      "step": 4190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.123557075673364e-05,
+      "loss": 0.5902,
+      "step": 4200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.121419410004276e-05,
+      "loss": 0.6592,
+      "step": 4210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.119281744335186e-05,
+      "loss": 0.582,
+      "step": 4220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.117144078666097e-05,
+      "loss": 0.6449,
+      "step": 4230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.115006412997009e-05,
+      "loss": 0.6057,
+      "step": 4240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.112868747327919e-05,
+      "loss": 0.5951,
+      "step": 4250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.110731081658828e-05,
+      "loss": 0.662,
+      "step": 4260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.10859341598974e-05,
+      "loss": 0.5941,
+      "step": 4270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.10645575032065e-05,
+      "loss": 0.6517,
+      "step": 4280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.104318084651561e-05,
+      "loss": 0.5868,
+      "step": 4290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.102180418982471e-05,
+      "loss": 0.5859,
+      "step": 4300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.100042753313383e-05,
+      "loss": 0.6494,
+      "step": 4310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.097905087644294e-05,
+      "loss": 0.5869,
+      "step": 4320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.095767421975203e-05,
+      "loss": 0.6384,
+      "step": 4330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.093629756306114e-05,
+      "loss": 0.5882,
+      "step": 4340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.091492090637025e-05,
+      "loss": 0.5833,
+      "step": 4350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.089354424967935e-05,
+      "loss": 0.6447,
+      "step": 4360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.087216759298847e-05,
+      "loss": 0.5842,
+      "step": 4370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.085079093629758e-05,
+      "loss": 0.6514,
+      "step": 4380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.082941427960667e-05,
+      "loss": 0.5904,
+      "step": 4390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.080803762291577e-05,
+      "loss": 0.5752,
+      "step": 4400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.078666096622489e-05,
+      "loss": 0.6477,
+      "step": 4410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.0765284309534e-05,
+      "loss": 0.5775,
+      "step": 4420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.07439076528431e-05,
+      "loss": 0.6417,
+      "step": 4430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.072253099615222e-05,
+      "loss": 0.5974,
+      "step": 4440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.070115433946131e-05,
+      "loss": 0.5723,
+      "step": 4450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.067977768277041e-05,
+      "loss": 0.6443,
+      "step": 4460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.065840102607952e-05,
+      "loss": 0.5811,
+      "step": 4470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.063702436938863e-05,
+      "loss": 0.6378,
+      "step": 4480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.061564771269774e-05,
+      "loss": 0.5967,
+      "step": 4490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.059427105600684e-05,
+      "loss": 0.5773,
+      "step": 4500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.057289439931596e-05,
+      "loss": 0.6508,
+      "step": 4510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.055151774262505e-05,
+      "loss": 0.582,
+      "step": 4520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.053014108593416e-05,
+      "loss": 0.6422,
+      "step": 4530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.050876442924327e-05,
+      "loss": 0.5941,
+      "step": 4540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.048738777255238e-05,
+      "loss": 0.5738,
+      "step": 4550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.046601111586148e-05,
+      "loss": 0.6496,
+      "step": 4560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.044463445917059e-05,
+      "loss": 0.5751,
+      "step": 4570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.042325780247969e-05,
+      "loss": 0.6402,
+      "step": 4580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.04018811457888e-05,
+      "loss": 0.5927,
+      "step": 4590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.03805044890979e-05,
+      "loss": 0.5769,
+      "step": 4600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.035912783240702e-05,
+      "loss": 0.6474,
+      "step": 4610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.033775117571612e-05,
+      "loss": 0.5871,
+      "step": 4620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.031637451902523e-05,
+      "loss": 0.6459,
+      "step": 4630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.029499786233435e-05,
+      "loss": 0.5918,
+      "step": 4640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.027362120564344e-05,
+      "loss": 0.5836,
+      "step": 4650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.025224454895254e-05,
+      "loss": 0.6451,
+      "step": 4660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.023086789226165e-05,
+      "loss": 0.5762,
+      "step": 4670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.020949123557077e-05,
+      "loss": 0.6443,
+      "step": 4680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.018811457887987e-05,
+      "loss": 0.5904,
+      "step": 4690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.016673792218897e-05,
+      "loss": 0.5725,
+      "step": 4700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.014536126549808e-05,
+      "loss": 0.6333,
+      "step": 4710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.012398460880718e-05,
+      "loss": 0.5778,
+      "step": 4720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.010260795211629e-05,
+      "loss": 0.6443,
+      "step": 4730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.008123129542539e-05,
+      "loss": 0.5872,
+      "step": 4740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.005985463873451e-05,
+      "loss": 0.5788,
+      "step": 4750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.003847798204361e-05,
+      "loss": 0.6497,
+      "step": 4760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.001710132535272e-05,
+      "loss": 0.5853,
+      "step": 4770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.999572466866182e-05,
+      "loss": 0.6395,
+      "step": 4780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.997434801197093e-05,
+      "loss": 0.6032,
+      "step": 4790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.995297135528003e-05,
+      "loss": 0.5695,
+      "step": 4800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.993159469858915e-05,
+      "loss": 0.6521,
+      "step": 4810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.991021804189826e-05,
+      "loss": 0.5717,
+      "step": 4820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.988884138520736e-05,
+      "loss": 0.6481,
+      "step": 4830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.986746472851646e-05,
+      "loss": 0.5899,
+      "step": 4840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.984608807182557e-05,
+      "loss": 0.5897,
+      "step": 4850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.982471141513467e-05,
+      "loss": 0.641,
+      "step": 4860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.980333475844378e-05,
+      "loss": 0.5832,
+      "step": 4870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.97819581017529e-05,
+      "loss": 0.6438,
+      "step": 4880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.9760581445062e-05,
+      "loss": 0.5963,
+      "step": 4890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.97392047883711e-05,
+      "loss": 0.5721,
+      "step": 4900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.971782813168021e-05,
+      "loss": 0.6455,
+      "step": 4910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.969645147498931e-05,
+      "loss": 0.562,
+      "step": 4920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.967507481829842e-05,
+      "loss": 0.644,
+      "step": 4930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.965369816160752e-05,
+      "loss": 0.5944,
+      "step": 4940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.963232150491664e-05,
+      "loss": 0.5806,
+      "step": 4950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.961094484822575e-05,
+      "loss": 0.6485,
+      "step": 4960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.958956819153485e-05,
+      "loss": 0.585,
+      "step": 4970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.956819153484395e-05,
+      "loss": 0.6438,
+      "step": 4980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.954681487815306e-05,
+      "loss": 0.5973,
+      "step": 4990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.952543822146216e-05,
+      "loss": 0.5681,
+      "step": 5000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.950406156477127e-05,
+      "loss": 0.6313,
+      "step": 5010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.948268490808039e-05,
+      "loss": 0.5741,
+      "step": 5020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.946130825138949e-05,
+      "loss": 0.6429,
+      "step": 5030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.94399315946986e-05,
+      "loss": 0.5905,
+      "step": 5040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.94185549380077e-05,
+      "loss": 0.58,
+      "step": 5050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.93971782813168e-05,
+      "loss": 0.6431,
+      "step": 5060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.937580162462591e-05,
+      "loss": 0.5747,
+      "step": 5070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.935442496793503e-05,
+      "loss": 0.6399,
+      "step": 5080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.933304831124413e-05,
+      "loss": 0.5846,
+      "step": 5090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.931167165455324e-05,
+      "loss": 0.5798,
+      "step": 5100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.929029499786234e-05,
+      "loss": 0.6438,
+      "step": 5110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.926891834117144e-05,
+      "loss": 0.5852,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.924754168448055e-05,
+      "loss": 0.6354,
+      "step": 5130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.922616502778965e-05,
+      "loss": 0.5967,
+      "step": 5140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.920478837109877e-05,
+      "loss": 0.5783,
+      "step": 5150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.918341171440788e-05,
+      "loss": 0.6457,
+      "step": 5160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.916203505771698e-05,
+      "loss": 0.5745,
+      "step": 5170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.914065840102609e-05,
+      "loss": 0.633,
+      "step": 5180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.911928174433519e-05,
+      "loss": 0.5989,
+      "step": 5190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.90979050876443e-05,
+      "loss": 0.5863,
+      "step": 5200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.90765284309534e-05,
+      "loss": 0.6467,
+      "step": 5210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.905515177426252e-05,
+      "loss": 0.5777,
+      "step": 5220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.903377511757162e-05,
+      "loss": 0.6356,
+      "step": 5230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.901239846088071e-05,
+      "loss": 0.5919,
+      "step": 5240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.899102180418983e-05,
+      "loss": 0.5756,
+      "step": 5250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.896964514749894e-05,
+      "loss": 0.6475,
+      "step": 5260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.894826849080804e-05,
+      "loss": 0.5882,
+      "step": 5270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.892689183411714e-05,
+      "loss": 0.6429,
+      "step": 5280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.890551517742626e-05,
+      "loss": 0.5866,
+      "step": 5290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.888413852073537e-05,
+      "loss": 0.5816,
+      "step": 5300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.886276186404446e-05,
+      "loss": 0.6504,
+      "step": 5310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.884138520735358e-05,
+      "loss": 0.5742,
+      "step": 5320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.882000855066268e-05,
+      "loss": 0.6353,
+      "step": 5330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.879863189397178e-05,
+      "loss": 0.585,
+      "step": 5340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.87772552372809e-05,
+      "loss": 0.5748,
+      "step": 5350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.875587858059001e-05,
+      "loss": 0.6476,
+      "step": 5360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.87345019238991e-05,
+      "loss": 0.579,
+      "step": 5370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.87131252672082e-05,
+      "loss": 0.6323,
+      "step": 5380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.869174861051732e-05,
+      "loss": 0.5885,
+      "step": 5390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.867037195382643e-05,
+      "loss": 0.563,
+      "step": 5400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.864899529713553e-05,
+      "loss": 0.6455,
+      "step": 5410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.862761864044465e-05,
+      "loss": 0.5843,
+      "step": 5420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.860624198375375e-05,
+      "loss": 0.6458,
+      "step": 5430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.858486532706284e-05,
+      "loss": 0.5807,
+      "step": 5440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.856348867037196e-05,
+      "loss": 0.5734,
+      "step": 5450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.854211201368107e-05,
+      "loss": 0.6472,
+      "step": 5460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.852073535699017e-05,
+      "loss": 0.5796,
+      "step": 5470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.849935870029928e-05,
+      "loss": 0.6374,
+      "step": 5480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.847798204360839e-05,
+      "loss": 0.5877,
+      "step": 5490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.845660538691748e-05,
+      "loss": 0.5767,
+      "step": 5500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.843522873022659e-05,
+      "loss": 0.6475,
+      "step": 5510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.84138520735357e-05,
+      "loss": 0.5748,
+      "step": 5520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.839247541684481e-05,
+      "loss": 0.628,
+      "step": 5530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.837109876015392e-05,
+      "loss": 0.5823,
+      "step": 5540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.834972210346302e-05,
+      "loss": 0.5658,
+      "step": 5550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.832834544677214e-05,
+      "loss": 0.6425,
+      "step": 5560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.830696879008123e-05,
+      "loss": 0.5654,
+      "step": 5570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.828559213339033e-05,
+      "loss": 0.6343,
+      "step": 5580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.826421547669945e-05,
+      "loss": 0.5908,
+      "step": 5590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.824283882000856e-05,
+      "loss": 0.5664,
+      "step": 5600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.822146216331766e-05,
+      "loss": 0.6408,
+      "step": 5610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.820008550662678e-05,
+      "loss": 0.5767,
+      "step": 5620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.817870884993587e-05,
+      "loss": 0.6289,
+      "step": 5630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.815733219324497e-05,
+      "loss": 0.5879,
+      "step": 5640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.813595553655408e-05,
+      "loss": 0.5727,
+      "step": 5650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.81145788798632e-05,
+      "loss": 0.6488,
+      "step": 5660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.80932022231723e-05,
+      "loss": 0.5742,
+      "step": 5670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.80718255664814e-05,
+      "loss": 0.6282,
+      "step": 5680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.805044890979051e-05,
+      "loss": 0.5916,
+      "step": 5690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.802907225309962e-05,
+      "loss": 0.5768,
+      "step": 5700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.800769559640872e-05,
+      "loss": 0.6477,
+      "step": 5710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.798631893971784e-05,
+      "loss": 0.5747,
+      "step": 5720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.796494228302694e-05,
+      "loss": 0.6366,
+      "step": 5730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.794356562633605e-05,
+      "loss": 0.5956,
+      "step": 5740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.792218896964515e-05,
+      "loss": 0.5823,
+      "step": 5750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.790081231295426e-05,
+      "loss": 0.6509,
+      "step": 5760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.787943565626336e-05,
+      "loss": 0.577,
+      "step": 5770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.785805899957246e-05,
+      "loss": 0.6375,
+      "step": 5780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.783668234288158e-05,
+      "loss": 0.5858,
+      "step": 5790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.781530568619069e-05,
+      "loss": 0.5803,
+      "step": 5800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.779392902949979e-05,
+      "loss": 0.642,
+      "step": 5810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77725523728089e-05,
+      "loss": 0.5635,
+      "step": 5820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.7751175716118e-05,
+      "loss": 0.6303,
+      "step": 5830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.77297990594271e-05,
+      "loss": 0.5854,
+      "step": 5840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.770842240273621e-05,
+      "loss": 0.5705,
+      "step": 5850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.768704574604533e-05,
+      "loss": 0.6394,
+      "step": 5860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.766566908935443e-05,
+      "loss": 0.5694,
+      "step": 5870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.764429243266354e-05,
+      "loss": 0.6324,
+      "step": 5880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.762291577597264e-05,
+      "loss": 0.5907,
+      "step": 5890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.760153911928175e-05,
+      "loss": 0.582,
+      "step": 5900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.758016246259085e-05,
+      "loss": 0.6426,
+      "step": 5910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.755878580589995e-05,
+      "loss": 0.5812,
+      "step": 5920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.753740914920907e-05,
+      "loss": 0.6322,
+      "step": 5930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.751603249251818e-05,
+      "loss": 0.5888,
+      "step": 5940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.749465583582728e-05,
+      "loss": 0.5756,
+      "step": 5950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.747327917913639e-05,
+      "loss": 0.6421,
+      "step": 5960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.745190252244549e-05,
+      "loss": 0.5776,
+      "step": 5970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.74305258657546e-05,
+      "loss": 0.6281,
+      "step": 5980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.740914920906371e-05,
+      "loss": 0.5991,
+      "step": 5990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.738777255237282e-05,
+      "loss": 0.5764,
+      "step": 6000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.736639589568192e-05,
+      "loss": 0.6396,
+      "step": 6010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.734501923899103e-05,
+      "loss": 0.5759,
+      "step": 6020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.732364258230013e-05,
+      "loss": 0.6337,
+      "step": 6030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.730226592560924e-05,
+      "loss": 0.5965,
+      "step": 6040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.728088926891834e-05,
+      "loss": 0.5752,
+      "step": 6050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.725951261222746e-05,
+      "loss": 0.6431,
+      "step": 6060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.723813595553656e-05,
+      "loss": 0.564,
+      "step": 6070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.721675929884567e-05,
+      "loss": 0.6388,
+      "step": 6080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.719538264215477e-05,
+      "loss": 0.592,
+      "step": 6090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.717400598546388e-05,
+      "loss": 0.5683,
+      "step": 6100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.715262932877298e-05,
+      "loss": 0.6348,
+      "step": 6110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.713125267208209e-05,
+      "loss": 0.5744,
+      "step": 6120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.71098760153912e-05,
+      "loss": 0.6318,
+      "step": 6130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.708849935870031e-05,
+      "loss": 0.5927,
+      "step": 6140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.706712270200941e-05,
+      "loss": 0.5735,
+      "step": 6150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.704574604531852e-05,
+      "loss": 0.643,
+      "step": 6160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.702436938862762e-05,
+      "loss": 0.5788,
+      "step": 6170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.700299273193673e-05,
+      "loss": 0.6335,
+      "step": 6180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.698161607524583e-05,
+      "loss": 0.5813,
+      "step": 6190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.696023941855495e-05,
+      "loss": 0.5684,
+      "step": 6200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.693886276186405e-05,
+      "loss": 0.631,
+      "step": 6210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.691748610517316e-05,
+      "loss": 0.5674,
+      "step": 6220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.689610944848226e-05,
+      "loss": 0.6316,
+      "step": 6230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.687473279179137e-05,
+      "loss": 0.5891,
+      "step": 6240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.685335613510047e-05,
+      "loss": 0.5761,
+      "step": 6250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.683197947840959e-05,
+      "loss": 0.6419,
+      "step": 6260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.68106028217187e-05,
+      "loss": 0.5742,
+      "step": 6270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.67892261650278e-05,
+      "loss": 0.6342,
+      "step": 6280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.676784950833689e-05,
+      "loss": 0.5912,
+      "step": 6290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.674647285164601e-05,
+      "loss": 0.5759,
+      "step": 6300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.672509619495511e-05,
+      "loss": 0.6286,
+      "step": 6310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.670371953826422e-05,
+      "loss": 0.5775,
+      "step": 6320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.668234288157333e-05,
+      "loss": 0.6346,
+      "step": 6330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.666096622488244e-05,
+      "loss": 0.5801,
+      "step": 6340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.663958956819154e-05,
+      "loss": 0.5638,
+      "step": 6350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.661821291150063e-05,
+      "loss": 0.642,
+      "step": 6360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.659683625480975e-05,
+      "loss": 0.5698,
+      "step": 6370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.657545959811886e-05,
+      "loss": 0.6293,
+      "step": 6380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.655408294142796e-05,
+      "loss": 0.5844,
+      "step": 6390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.653270628473708e-05,
+      "loss": 0.5729,
+      "step": 6400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.651132962804618e-05,
+      "loss": 0.6443,
+      "step": 6410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.648995297135528e-05,
+      "loss": 0.5785,
+      "step": 6420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.64685763146644e-05,
+      "loss": 0.6354,
+      "step": 6430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.64471996579735e-05,
+      "loss": 0.5819,
+      "step": 6440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.64258230012826e-05,
+      "loss": 0.5606,
+      "step": 6450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.640444634459171e-05,
+      "loss": 0.6505,
+      "step": 6460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.638306968790082e-05,
+      "loss": 0.5764,
+      "step": 6470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.636169303120992e-05,
+      "loss": 0.6339,
+      "step": 6480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.634031637451902e-05,
+      "loss": 0.5911,
+      "step": 6490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.631893971782814e-05,
+      "loss": 0.5716,
+      "step": 6500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.629756306113724e-05,
+      "loss": 0.6489,
+      "step": 6510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.627618640444635e-05,
+      "loss": 0.5706,
+      "step": 6520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.625480974775547e-05,
+      "loss": 0.6311,
+      "step": 6530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.623343309106457e-05,
+      "loss": 0.59,
+      "step": 6540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.621205643437366e-05,
+      "loss": 0.5642,
+      "step": 6550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.619067977768277e-05,
+      "loss": 0.6452,
+      "step": 6560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.616930312099188e-05,
+      "loss": 0.5715,
+      "step": 6570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.614792646430099e-05,
+      "loss": 0.6174,
+      "step": 6580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.612654980761009e-05,
+      "loss": 0.5702,
+      "step": 6590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.610517315091921e-05,
+      "loss": 0.5677,
+      "step": 6600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.60837964942283e-05,
+      "loss": 0.6382,
+      "step": 6610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.60624198375374e-05,
+      "loss": 0.5763,
+      "step": 6620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.604104318084651e-05,
+      "loss": 0.6281,
+      "step": 6630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.601966652415563e-05,
+      "loss": 0.5864,
+      "step": 6640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.599828986746473e-05,
+      "loss": 0.5705,
+      "step": 6650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.597691321077384e-05,
+      "loss": 0.6454,
+      "step": 6660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.595553655408296e-05,
+      "loss": 0.5756,
+      "step": 6670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.593415989739205e-05,
+      "loss": 0.6367,
+      "step": 6680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.591278324070115e-05,
+      "loss": 0.5747,
+      "step": 6690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.589140658401027e-05,
+      "loss": 0.5756,
+      "step": 6700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.587002992731937e-05,
+      "loss": 0.6373,
+      "step": 6710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.584865327062848e-05,
+      "loss": 0.5703,
+      "step": 6720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.582727661393758e-05,
+      "loss": 0.6384,
+      "step": 6730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.580589995724669e-05,
+      "loss": 0.5776,
+      "step": 6740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.578452330055579e-05,
+      "loss": 0.5724,
+      "step": 6750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.57631466438649e-05,
+      "loss": 0.6356,
+      "step": 6760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.574176998717401e-05,
+      "loss": 0.5704,
+      "step": 6770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.572039333048312e-05,
+      "loss": 0.6317,
+      "step": 6780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.569901667379222e-05,
+      "loss": 0.5927,
+      "step": 6790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.567764001710134e-05,
+      "loss": 0.5644,
+      "step": 6800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.565626336041043e-05,
+      "loss": 0.647,
+      "step": 6810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.563488670371954e-05,
+      "loss": 0.5739,
+      "step": 6820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.561351004702864e-05,
+      "loss": 0.6314,
+      "step": 6830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.559213339033776e-05,
+      "loss": 0.5791,
+      "step": 6840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.557075673364686e-05,
+      "loss": 0.5664,
+      "step": 6850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.554938007695597e-05,
+      "loss": 0.6443,
+      "step": 6860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.552800342026507e-05,
+      "loss": 0.5794,
+      "step": 6870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.550662676357418e-05,
+      "loss": 0.6261,
+      "step": 6880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.548525010688328e-05,
+      "loss": 0.5711,
+      "step": 6890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.546387345019239e-05,
+      "loss": 0.5646,
+      "step": 6900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.54424967935015e-05,
+      "loss": 0.6257,
+      "step": 6910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.542112013681061e-05,
+      "loss": 0.5686,
+      "step": 6920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.539974348011971e-05,
+      "loss": 0.6298,
+      "step": 6930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.537836682342882e-05,
+      "loss": 0.5772,
+      "step": 6940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.535699016673792e-05,
+      "loss": 0.5551,
+      "step": 6950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.533561351004703e-05,
+      "loss": 0.6515,
+      "step": 6960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.531423685335615e-05,
+      "loss": 0.5604,
+      "step": 6970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.529286019666525e-05,
+      "loss": 0.6233,
+      "step": 6980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.527148353997435e-05,
+      "loss": 0.5859,
+      "step": 6990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.525010688328346e-05,
+      "loss": 0.5653,
+      "step": 7000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.522873022659256e-05,
+      "loss": 0.6506,
+      "step": 7010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.520735356990167e-05,
+      "loss": 0.5721,
+      "step": 7020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.518597691321077e-05,
+      "loss": 0.6293,
+      "step": 7030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.516460025651989e-05,
+      "loss": 0.5837,
+      "step": 7040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.5143223599829e-05,
+      "loss": 0.5629,
+      "step": 7050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.51218469431381e-05,
+      "loss": 0.6381,
+      "step": 7060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.51004702864472e-05,
+      "loss": 0.5668,
+      "step": 7070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.507909362975631e-05,
+      "loss": 0.6315,
+      "step": 7080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.505771697306541e-05,
+      "loss": 0.5883,
+      "step": 7090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.503634031637452e-05,
+      "loss": 0.5641,
+      "step": 7100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.501496365968364e-05,
+      "loss": 0.6337,
+      "step": 7110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.499358700299274e-05,
+      "loss": 0.5743,
+      "step": 7120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.497221034630184e-05,
+      "loss": 0.6268,
+      "step": 7130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.495083368961095e-05,
+      "loss": 0.5807,
+      "step": 7140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.492945703292005e-05,
+      "loss": 0.566,
+      "step": 7150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.490808037622916e-05,
+      "loss": 0.6325,
+      "step": 7160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.488670371953826e-05,
+      "loss": 0.5665,
+      "step": 7170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.486532706284738e-05,
+      "loss": 0.6362,
+      "step": 7180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.484395040615649e-05,
+      "loss": 0.5832,
+      "step": 7190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.482257374946559e-05,
+      "loss": 0.5689,
+      "step": 7200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.48011970927747e-05,
+      "loss": 0.6282,
+      "step": 7210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.47798204360838e-05,
+      "loss": 0.5712,
+      "step": 7220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.47584437793929e-05,
+      "loss": 0.626,
+      "step": 7230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.473706712270202e-05,
+      "loss": 0.5838,
+      "step": 7240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.471569046601113e-05,
+      "loss": 0.5713,
+      "step": 7250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.469431380932023e-05,
+      "loss": 0.6296,
+      "step": 7260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.467293715262932e-05,
+      "loss": 0.5758,
+      "step": 7270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.465156049593844e-05,
+      "loss": 0.6305,
+      "step": 7280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.463018383924754e-05,
+      "loss": 0.5794,
+      "step": 7290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.460880718255665e-05,
+      "loss": 0.5734,
+      "step": 7300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.458743052586577e-05,
+      "loss": 0.6443,
+      "step": 7310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.456605386917487e-05,
+      "loss": 0.5678,
+      "step": 7320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.454467721248398e-05,
+      "loss": 0.6281,
+      "step": 7330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.452330055579308e-05,
+      "loss": 0.5862,
+      "step": 7340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.450192389910218e-05,
+      "loss": 0.566,
+      "step": 7350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.448054724241129e-05,
+      "loss": 0.6405,
+      "step": 7360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.44591705857204e-05,
+      "loss": 0.5694,
+      "step": 7370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.443779392902951e-05,
+      "loss": 0.6264,
+      "step": 7380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.441641727233862e-05,
+      "loss": 0.5824,
+      "step": 7390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.439504061564771e-05,
+      "loss": 0.5669,
+      "step": 7400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.437366395895683e-05,
+      "loss": 0.6339,
+      "step": 7410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.435228730226593e-05,
+      "loss": 0.5642,
+      "step": 7420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.433091064557503e-05,
+      "loss": 0.6182,
+      "step": 7430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.430953398888414e-05,
+      "loss": 0.5776,
+      "step": 7440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.428815733219326e-05,
+      "loss": 0.5681,
+      "step": 7450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.426678067550236e-05,
+      "loss": 0.6213,
+      "step": 7460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.424540401881145e-05,
+      "loss": 0.5612,
+      "step": 7470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.422402736212057e-05,
+      "loss": 0.6248,
+      "step": 7480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.420265070542967e-05,
+      "loss": 0.5795,
+      "step": 7490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.418127404873878e-05,
+      "loss": 0.56,
+      "step": 7500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.41598973920479e-05,
+      "loss": 0.6377,
+      "step": 7510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.4138520735357e-05,
+      "loss": 0.5756,
+      "step": 7520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.411714407866609e-05,
+      "loss": 0.6235,
+      "step": 7530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.40957674219752e-05,
+      "loss": 0.5808,
+      "step": 7540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.407439076528432e-05,
+      "loss": 0.571,
+      "step": 7550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.405301410859342e-05,
+      "loss": 0.6342,
+      "step": 7560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.403163745190252e-05,
+      "loss": 0.5678,
+      "step": 7570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.401026079521164e-05,
+      "loss": 0.6195,
+      "step": 7580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.398888413852075e-05,
+      "loss": 0.5756,
+      "step": 7590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.396750748182984e-05,
+      "loss": 0.5772,
+      "step": 7600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.394613082513896e-05,
+      "loss": 0.6286,
+      "step": 7610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.392475416844806e-05,
+      "loss": 0.5622,
+      "step": 7620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.390337751175717e-05,
+      "loss": 0.627,
+      "step": 7630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.388200085506627e-05,
+      "loss": 0.5709,
+      "step": 7640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.386062419837539e-05,
+      "loss": 0.5667,
+      "step": 7650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.383924754168448e-05,
+      "loss": 0.6382,
+      "step": 7660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.381787088499358e-05,
+      "loss": 0.572,
+      "step": 7670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.37964942283027e-05,
+      "loss": 0.6292,
+      "step": 7680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.37751175716118e-05,
+      "loss": 0.5808,
+      "step": 7690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.375374091492091e-05,
+      "loss": 0.5674,
+      "step": 7700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.373236425823001e-05,
+      "loss": 0.6286,
+      "step": 7710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.371098760153912e-05,
+      "loss": 0.5692,
+      "step": 7720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.368961094484822e-05,
+      "loss": 0.6254,
+      "step": 7730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.366823428815733e-05,
+      "loss": 0.5832,
+      "step": 7740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.364685763146645e-05,
+      "loss": 0.565,
+      "step": 7750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.362548097477555e-05,
+      "loss": 0.6298,
+      "step": 7760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.360410431808466e-05,
+      "loss": 0.567,
+      "step": 7770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.358272766139377e-05,
+      "loss": 0.6213,
+      "step": 7780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.356135100470286e-05,
+      "loss": 0.574,
+      "step": 7790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.353997434801197e-05,
+      "loss": 0.5699,
+      "step": 7800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.351859769132107e-05,
+      "loss": 0.6268,
+      "step": 7810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.349722103463019e-05,
+      "loss": 0.5742,
+      "step": 7820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.34758443779393e-05,
+      "loss": 0.6264,
+      "step": 7830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.34544677212484e-05,
+      "loss": 0.5777,
+      "step": 7840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 8.34330910645575e-05,
+      "loss": 0.5623,
+      "step": 7850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.341171440786661e-05,
+      "loss": 0.6235,
+      "step": 7860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.339033775117571e-05,
+      "loss": 0.5674,
+      "step": 7870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.336896109448483e-05,
+      "loss": 0.6271,
+      "step": 7880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.334758443779394e-05,
+      "loss": 0.5822,
+      "step": 7890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.332620778110304e-05,
+      "loss": 0.5667,
+      "step": 7900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.330483112441215e-05,
+      "loss": 0.6348,
+      "step": 7910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.328345446772125e-05,
+      "loss": 0.5647,
+      "step": 7920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.326207781103035e-05,
+      "loss": 0.6239,
+      "step": 7930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.324070115433946e-05,
+      "loss": 0.5812,
+      "step": 7940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.321932449764858e-05,
+      "loss": 0.5693,
+      "step": 7950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.319794784095768e-05,
+      "loss": 0.6279,
+      "step": 7960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.317657118426679e-05,
+      "loss": 0.5656,
+      "step": 7970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.315519452757589e-05,
+      "loss": 0.6269,
+      "step": 7980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.3133817870885e-05,
+      "loss": 0.5905,
+      "step": 7990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.31124412141941e-05,
+      "loss": 0.5605,
+      "step": 8000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.30910645575032e-05,
+      "loss": 0.636,
+      "step": 8010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.306968790081232e-05,
+      "loss": 0.5722,
+      "step": 8020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.304831124412143e-05,
+      "loss": 0.6249,
+      "step": 8030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.302693458743053e-05,
+      "loss": 0.5768,
+      "step": 8040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.300555793073964e-05,
+      "loss": 0.5677,
+      "step": 8050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.298418127404874e-05,
+      "loss": 0.6426,
+      "step": 8060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.296280461735784e-05,
+      "loss": 0.5684,
+      "step": 8070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 8.294142796066695e-05,
+      "loss": 0.6204,
+      "step": 8080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.292005130397607e-05,
+      "loss": 0.5883,
+      "step": 8090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.289867464728517e-05,
+      "loss": 0.5655,
+      "step": 8100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.287729799059428e-05,
+      "loss": 0.6434,
+      "step": 8110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.285592133390338e-05,
+      "loss": 0.5635,
+      "step": 8120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.283454467721249e-05,
+      "loss": 0.6223,
+      "step": 8130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.281316802052159e-05,
+      "loss": 0.5878,
+      "step": 8140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.279179136383071e-05,
+      "loss": 0.5602,
+      "step": 8150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.277041470713981e-05,
+      "loss": 0.6272,
+      "step": 8160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.274903805044892e-05,
+      "loss": 0.5667,
+      "step": 8170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.272766139375802e-05,
+      "loss": 0.6361,
+      "step": 8180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.270628473706713e-05,
+      "loss": 0.571,
+      "step": 8190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.268490808037623e-05,
+      "loss": 0.5656,
+      "step": 8200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.266353142368534e-05,
+      "loss": 0.6388,
+      "step": 8210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.264215476699445e-05,
+      "loss": 0.5653,
+      "step": 8220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.262077811030356e-05,
+      "loss": 0.6226,
+      "step": 8230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.259940145361266e-05,
+      "loss": 0.5663,
+      "step": 8240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.257802479692177e-05,
+      "loss": 0.5657,
+      "step": 8250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.255664814023087e-05,
+      "loss": 0.6299,
+      "step": 8260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.253527148353998e-05,
+      "loss": 0.5626,
+      "step": 8270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.251389482684908e-05,
+      "loss": 0.628,
+      "step": 8280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.24925181701582e-05,
+      "loss": 0.5787,
+      "step": 8290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.24711415134673e-05,
+      "loss": 0.5607,
+      "step": 8300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.244976485677641e-05,
+      "loss": 0.6295,
+      "step": 8310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 8.242838820008551e-05,
+      "loss": 0.5609,
+      "step": 8320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.240701154339462e-05,
+      "loss": 0.6311,
+      "step": 8330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.238563488670372e-05,
+      "loss": 0.5823,
+      "step": 8340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.236425823001283e-05,
+      "loss": 0.5694,
+      "step": 8350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.234288157332194e-05,
+      "loss": 0.6357,
+      "step": 8360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.232150491663105e-05,
+      "loss": 0.5648,
+      "step": 8370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.230012825994015e-05,
+      "loss": 0.628,
+      "step": 8380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.227875160324926e-05,
+      "loss": 0.5765,
+      "step": 8390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.225737494655836e-05,
+      "loss": 0.5685,
+      "step": 8400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.223599828986747e-05,
+      "loss": 0.6356,
+      "step": 8410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.221462163317658e-05,
+      "loss": 0.5612,
+      "step": 8420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.219324497648569e-05,
+      "loss": 0.6249,
+      "step": 8430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.217186831979479e-05,
+      "loss": 0.5797,
+      "step": 8440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.215049166310388e-05,
+      "loss": 0.5656,
+      "step": 8450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.2129115006413e-05,
+      "loss": 0.6283,
+      "step": 8460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.21077383497221e-05,
+      "loss": 0.5533,
+      "step": 8470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.208636169303121e-05,
+      "loss": 0.632,
+      "step": 8480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.206498503634033e-05,
+      "loss": 0.5861,
+      "step": 8490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.204360837964943e-05,
+      "loss": 0.5651,
+      "step": 8500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.202223172295852e-05,
+      "loss": 0.639,
+      "step": 8510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.200085506626763e-05,
+      "loss": 0.5659,
+      "step": 8520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.197947840957675e-05,
+      "loss": 0.621,
+      "step": 8530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.195810175288585e-05,
+      "loss": 0.5816,
+      "step": 8540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 8.193672509619496e-05,
+      "loss": 0.5661,
+      "step": 8550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.191534843950407e-05,
+      "loss": 0.6287,
+      "step": 8560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.189397178281318e-05,
+      "loss": 0.5672,
+      "step": 8570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.187259512612227e-05,
+      "loss": 0.6281,
+      "step": 8580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.185121846943139e-05,
+      "loss": 0.588,
+      "step": 8590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.182984181274049e-05,
+      "loss": 0.5646,
+      "step": 8600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.18084651560496e-05,
+      "loss": 0.631,
+      "step": 8610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.17870884993587e-05,
+      "loss": 0.5629,
+      "step": 8620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.176571184266782e-05,
+      "loss": 0.6205,
+      "step": 8630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.174433518597691e-05,
+      "loss": 0.5734,
+      "step": 8640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.172295852928601e-05,
+      "loss": 0.5663,
+      "step": 8650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.170158187259513e-05,
+      "loss": 0.6379,
+      "step": 8660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.168020521590424e-05,
+      "loss": 0.5636,
+      "step": 8670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.165882855921334e-05,
+      "loss": 0.6246,
+      "step": 8680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.163745190252246e-05,
+      "loss": 0.5738,
+      "step": 8690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.161607524583156e-05,
+      "loss": 0.5652,
+      "step": 8700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.159469858914066e-05,
+      "loss": 0.6369,
+      "step": 8710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.157332193244976e-05,
+      "loss": 0.5641,
+      "step": 8720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.155194527575888e-05,
+      "loss": 0.6235,
+      "step": 8730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.153056861906798e-05,
+      "loss": 0.575,
+      "step": 8740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.150919196237709e-05,
+      "loss": 0.5654,
+      "step": 8750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.14878153056862e-05,
+      "loss": 0.6288,
+      "step": 8760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.14664386489953e-05,
+      "loss": 0.5654,
+      "step": 8770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.14450619923044e-05,
+      "loss": 0.6312,
+      "step": 8780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 8.14236853356135e-05,
+      "loss": 0.5766,
+      "step": 8790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.140230867892262e-05,
+      "loss": 0.5603,
+      "step": 8800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.138093202223173e-05,
+      "loss": 0.6398,
+      "step": 8810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.135955536554083e-05,
+      "loss": 0.5615,
+      "step": 8820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.133817870884995e-05,
+      "loss": 0.6243,
+      "step": 8830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.131680205215904e-05,
+      "loss": 0.5705,
+      "step": 8840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.129542539546815e-05,
+      "loss": 0.5594,
+      "step": 8850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.127404873877726e-05,
+      "loss": 0.6236,
+      "step": 8860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.125267208208637e-05,
+      "loss": 0.5609,
+      "step": 8870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.123129542539547e-05,
+      "loss": 0.6247,
+      "step": 8880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.120991876870458e-05,
+      "loss": 0.5712,
+      "step": 8890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.118854211201368e-05,
+      "loss": 0.563,
+      "step": 8900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.116716545532279e-05,
+      "loss": 0.632,
+      "step": 8910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.114578879863189e-05,
+      "loss": 0.566,
+      "step": 8920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.112441214194101e-05,
+      "loss": 0.6196,
+      "step": 8930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.110303548525011e-05,
+      "loss": 0.5776,
+      "step": 8940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.108165882855922e-05,
+      "loss": 0.5607,
+      "step": 8950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.106028217186832e-05,
+      "loss": 0.6289,
+      "step": 8960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.103890551517743e-05,
+      "loss": 0.5669,
+      "step": 8970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.101752885848653e-05,
+      "loss": 0.6181,
+      "step": 8980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.099615220179564e-05,
+      "loss": 0.5894,
+      "step": 8990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.097477554510475e-05,
+      "loss": 0.5576,
+      "step": 9000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.095339888841386e-05,
+      "loss": 0.6289,
+      "step": 9010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.093202223172296e-05,
+      "loss": 0.5601,
+      "step": 9020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.091064557503207e-05,
+      "loss": 0.6134,
+      "step": 9030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.088926891834117e-05,
+      "loss": 0.5778,
+      "step": 9040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.086789226165028e-05,
+      "loss": 0.5657,
+      "step": 9050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.084651560495938e-05,
+      "loss": 0.6255,
+      "step": 9060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.08251389482685e-05,
+      "loss": 0.5682,
+      "step": 9070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.08037622915776e-05,
+      "loss": 0.6174,
+      "step": 9080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.078238563488671e-05,
+      "loss": 0.5725,
+      "step": 9090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.076100897819581e-05,
+      "loss": 0.5647,
+      "step": 9100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.073963232150492e-05,
+      "loss": 0.6373,
+      "step": 9110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.071825566481402e-05,
+      "loss": 0.5684,
+      "step": 9120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.069687900812314e-05,
+      "loss": 0.624,
+      "step": 9130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.067550235143224e-05,
+      "loss": 0.5755,
+      "step": 9140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.065412569474135e-05,
+      "loss": 0.5609,
+      "step": 9150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.063274903805045e-05,
+      "loss": 0.6277,
+      "step": 9160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.061137238135956e-05,
+      "loss": 0.5597,
+      "step": 9170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.058999572466866e-05,
+      "loss": 0.6225,
+      "step": 9180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.056861906797777e-05,
+      "loss": 0.5845,
+      "step": 9190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.054724241128688e-05,
+      "loss": 0.5635,
+      "step": 9200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.052586575459599e-05,
+      "loss": 0.6314,
+      "step": 9210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.05044890979051e-05,
+      "loss": 0.5622,
+      "step": 9220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.04831124412142e-05,
+      "loss": 0.616,
+      "step": 9230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.04617357845233e-05,
+      "loss": 0.5695,
+      "step": 9240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 8.044035912783241e-05,
+      "loss": 0.5652,
+      "step": 9250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.041898247114151e-05,
+      "loss": 0.6347,
+      "step": 9260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.039760581445063e-05,
+      "loss": 0.5591,
+      "step": 9270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.037622915775973e-05,
+      "loss": 0.6265,
+      "step": 9280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.035485250106884e-05,
+      "loss": 0.5802,
+      "step": 9290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.033347584437794e-05,
+      "loss": 0.5697,
+      "step": 9300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.031209918768705e-05,
+      "loss": 0.6226,
+      "step": 9310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.029072253099615e-05,
+      "loss": 0.5622,
+      "step": 9320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.026934587430526e-05,
+      "loss": 0.6127,
+      "step": 9330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.024796921761438e-05,
+      "loss": 0.5773,
+      "step": 9340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.022659256092348e-05,
+      "loss": 0.5543,
+      "step": 9350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.020521590423258e-05,
+      "loss": 0.6252,
+      "step": 9360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.018383924754169e-05,
+      "loss": 0.5658,
+      "step": 9370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.01624625908508e-05,
+      "loss": 0.6248,
+      "step": 9380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.01410859341599e-05,
+      "loss": 0.5814,
+      "step": 9390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.011970927746902e-05,
+      "loss": 0.5603,
+      "step": 9400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.009833262077812e-05,
+      "loss": 0.6265,
+      "step": 9410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.007695596408722e-05,
+      "loss": 0.5623,
+      "step": 9420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.005557930739632e-05,
+      "loss": 0.6081,
+      "step": 9430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.003420265070543e-05,
+      "loss": 0.5702,
+      "step": 9440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 8.001282599401454e-05,
+      "loss": 0.5622,
+      "step": 9450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.999144933732364e-05,
+      "loss": 0.625,
+      "step": 9460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.997007268063276e-05,
+      "loss": 0.5675,
+      "step": 9470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.994869602394187e-05,
+      "loss": 0.6168,
+      "step": 9480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 7.992731936725097e-05,
+      "loss": 0.5726,
+      "step": 9490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.990594271056007e-05,
+      "loss": 0.5665,
+      "step": 9500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.988456605386918e-05,
+      "loss": 0.6256,
+      "step": 9510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.986318939717828e-05,
+      "loss": 0.567,
+      "step": 9520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.984181274048739e-05,
+      "loss": 0.6187,
+      "step": 9530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.98204360837965e-05,
+      "loss": 0.5773,
+      "step": 9540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.979905942710561e-05,
+      "loss": 0.5619,
+      "step": 9550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.97776827704147e-05,
+      "loss": 0.6334,
+      "step": 9560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.975630611372382e-05,
+      "loss": 0.564,
+      "step": 9570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.973492945703292e-05,
+      "loss": 0.6292,
+      "step": 9580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.971355280034203e-05,
+      "loss": 0.5727,
+      "step": 9590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.969217614365113e-05,
+      "loss": 0.5622,
+      "step": 9600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.967079948696025e-05,
+      "loss": 0.625,
+      "step": 9610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.964942283026936e-05,
+      "loss": 0.5764,
+      "step": 9620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.962804617357845e-05,
+      "loss": 0.6248,
+      "step": 9630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.960666951688756e-05,
+      "loss": 0.5801,
+      "step": 9640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.958529286019667e-05,
+      "loss": 0.5671,
+      "step": 9650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.956391620350577e-05,
+      "loss": 0.6266,
+      "step": 9660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.954253954681489e-05,
+      "loss": 0.5657,
+      "step": 9670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.9521162890124e-05,
+      "loss": 0.6161,
+      "step": 9680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.949978623343309e-05,
+      "loss": 0.5695,
+      "step": 9690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.947840957674219e-05,
+      "loss": 0.5676,
+      "step": 9700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.945703292005131e-05,
+      "loss": 0.6205,
+      "step": 9710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 7.943565626336041e-05,
+      "loss": 0.5673,
+      "step": 9720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.941427960666952e-05,
+      "loss": 0.6226,
+      "step": 9730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.939290294997864e-05,
+      "loss": 0.56,
+      "step": 9740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.937152629328773e-05,
+      "loss": 0.5605,
+      "step": 9750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.935014963659683e-05,
+      "loss": 0.6242,
+      "step": 9760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.932877297990595e-05,
+      "loss": 0.5564,
+      "step": 9770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.930739632321506e-05,
+      "loss": 0.6175,
+      "step": 9780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.928601966652416e-05,
+      "loss": 0.5786,
+      "step": 9790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.926464300983326e-05,
+      "loss": 0.5682,
+      "step": 9800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.924326635314238e-05,
+      "loss": 0.6296,
+      "step": 9810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.922188969645147e-05,
+      "loss": 0.5602,
+      "step": 9820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.920051303976058e-05,
+      "loss": 0.6192,
+      "step": 9830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.91791363830697e-05,
+      "loss": 0.5676,
+      "step": 9840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.91577597263788e-05,
+      "loss": 0.5669,
+      "step": 9850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.91363830696879e-05,
+      "loss": 0.619,
+      "step": 9860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.911500641299701e-05,
+      "loss": 0.5639,
+      "step": 9870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.909362975630611e-05,
+      "loss": 0.6206,
+      "step": 9880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.907225309961522e-05,
+      "loss": 0.5823,
+      "step": 9890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.905087644292432e-05,
+      "loss": 0.5588,
+      "step": 9900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.902949978623344e-05,
+      "loss": 0.6308,
+      "step": 9910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.900812312954255e-05,
+      "loss": 0.5541,
+      "step": 9920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.898674647285165e-05,
+      "loss": 0.6127,
+      "step": 9930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.896536981616077e-05,
+      "loss": 0.5731,
+      "step": 9940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.894399315946986e-05,
+      "loss": 0.5519,
+      "step": 9950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 7.892261650277896e-05,
+      "loss": 0.629,
+      "step": 9960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.890123984608807e-05,
+      "loss": 0.5542,
+      "step": 9970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.887986318939719e-05,
+      "loss": 0.6201,
+      "step": 9980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.885848653270629e-05,
+      "loss": 0.5656,
+      "step": 9990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.88371098760154e-05,
+      "loss": 0.556,
+      "step": 10000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.88157332193245e-05,
+      "loss": 0.6346,
+      "step": 10010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.87943565626336e-05,
+      "loss": 0.5688,
+      "step": 10020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.877297990594271e-05,
+      "loss": 0.6181,
+      "step": 10030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.875160324925183e-05,
+      "loss": 0.5622,
+      "step": 10040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.873022659256093e-05,
+      "loss": 0.5596,
+      "step": 10050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.870884993587004e-05,
+      "loss": 0.6293,
+      "step": 10060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.868747327917914e-05,
+      "loss": 0.5648,
+      "step": 10070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.866609662248824e-05,
+      "loss": 0.6267,
+      "step": 10080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.864471996579735e-05,
+      "loss": 0.5801,
+      "step": 10090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.862334330910645e-05,
+      "loss": 0.5693,
+      "step": 10100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.860196665241557e-05,
+      "loss": 0.6275,
+      "step": 10110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.858058999572468e-05,
+      "loss": 0.5735,
+      "step": 10120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.855921333903378e-05,
+      "loss": 0.6187,
+      "step": 10130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.853783668234289e-05,
+      "loss": 0.569,
+      "step": 10140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.851646002565199e-05,
+      "loss": 0.5442,
+      "step": 10150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.84950833689611e-05,
+      "loss": 0.6257,
+      "step": 10160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.84737067122702e-05,
+      "loss": 0.5603,
+      "step": 10170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.845233005557932e-05,
+      "loss": 0.6104,
+      "step": 10180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 7.843095339888842e-05,
+      "loss": 0.5717,
+      "step": 10190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.840957674219753e-05,
+      "loss": 0.5603,
+      "step": 10200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.838820008550663e-05,
+      "loss": 0.613,
+      "step": 10210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.836682342881573e-05,
+      "loss": 0.5574,
+      "step": 10220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.834544677212484e-05,
+      "loss": 0.6197,
+      "step": 10230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.832407011543394e-05,
+      "loss": 0.5706,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.830269345874306e-05,
+      "loss": 0.5641,
+      "step": 10250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.828131680205217e-05,
+      "loss": 0.6277,
+      "step": 10260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.825994014536127e-05,
+      "loss": 0.5599,
+      "step": 10270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.823856348867038e-05,
+      "loss": 0.6138,
+      "step": 10280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.821718683197948e-05,
+      "loss": 0.5742,
+      "step": 10290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.819581017528858e-05,
+      "loss": 0.5614,
+      "step": 10300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.81744335185977e-05,
+      "loss": 0.6212,
+      "step": 10310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.815305686190681e-05,
+      "loss": 0.55,
+      "step": 10320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.813168020521591e-05,
+      "loss": 0.6049,
+      "step": 10330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.811030354852502e-05,
+      "loss": 0.569,
+      "step": 10340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.808892689183412e-05,
+      "loss": 0.5514,
+      "step": 10350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.806755023514323e-05,
+      "loss": 0.6248,
+      "step": 10360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.804617357845233e-05,
+      "loss": 0.563,
+      "step": 10370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.802479692176145e-05,
+      "loss": 0.6117,
+      "step": 10380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.800342026507055e-05,
+      "loss": 0.567,
+      "step": 10390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.798204360837966e-05,
+      "loss": 0.554,
+      "step": 10400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.796066695168876e-05,
+      "loss": 0.6228,
+      "step": 10410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.793929029499787e-05,
+      "loss": 0.5516,
+      "step": 10420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.791791363830697e-05,
+      "loss": 0.6168,
+      "step": 10430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.789653698161607e-05,
+      "loss": 0.5747,
+      "step": 10440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.787516032492519e-05,
+      "loss": 0.5578,
+      "step": 10450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.78537836682343e-05,
+      "loss": 0.6337,
+      "step": 10460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.78324070115434e-05,
+      "loss": 0.5529,
+      "step": 10470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.78110303548525e-05,
+      "loss": 0.6234,
+      "step": 10480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.778965369816161e-05,
+      "loss": 0.5685,
+      "step": 10490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.776827704147072e-05,
+      "loss": 0.5613,
+      "step": 10500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.774690038477982e-05,
+      "loss": 0.6168,
+      "step": 10510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.772552372808894e-05,
+      "loss": 0.5514,
+      "step": 10520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.770414707139804e-05,
+      "loss": 0.6233,
+      "step": 10530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.768277041470713e-05,
+      "loss": 0.5803,
+      "step": 10540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.766139375801625e-05,
+      "loss": 0.5612,
+      "step": 10550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.764001710132536e-05,
+      "loss": 0.6265,
+      "step": 10560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.761864044463446e-05,
+      "loss": 0.5668,
+      "step": 10570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.759726378794358e-05,
+      "loss": 0.6183,
+      "step": 10580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.757588713125268e-05,
+      "loss": 0.5676,
+      "step": 10590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.755451047456179e-05,
+      "loss": 0.5597,
+      "step": 10600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.753313381787088e-05,
+      "loss": 0.6171,
+      "step": 10610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.751175716118e-05,
+      "loss": 0.5523,
+      "step": 10620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.74903805044891e-05,
+      "loss": 0.6123,
+      "step": 10630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.74690038477982e-05,
+      "loss": 0.5646,
+      "step": 10640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.744762719110732e-05,
+      "loss": 0.5457,
+      "step": 10650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.742625053441643e-05,
+      "loss": 0.6312,
+      "step": 10660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.740487387772552e-05,
+      "loss": 0.5652,
+      "step": 10670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.738349722103462e-05,
+      "loss": 0.6169,
+      "step": 10680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.736212056434374e-05,
+      "loss": 0.5657,
+      "step": 10690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.734074390765285e-05,
+      "loss": 0.5537,
+      "step": 10700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.731936725096195e-05,
+      "loss": 0.6173,
+      "step": 10710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.729799059427107e-05,
+      "loss": 0.5588,
+      "step": 10720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.727661393758017e-05,
+      "loss": 0.6074,
+      "step": 10730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.725523728088926e-05,
+      "loss": 0.565,
+      "step": 10740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.723386062419838e-05,
+      "loss": 0.5519,
+      "step": 10750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.721248396750749e-05,
+      "loss": 0.6164,
+      "step": 10760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.719110731081659e-05,
+      "loss": 0.5565,
+      "step": 10770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.71697306541257e-05,
+      "loss": 0.6169,
+      "step": 10780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.714835399743481e-05,
+      "loss": 0.5726,
+      "step": 10790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.71269773407439e-05,
+      "loss": 0.5591,
+      "step": 10800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.710560068405301e-05,
+      "loss": 0.6299,
+      "step": 10810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.708422402736213e-05,
+      "loss": 0.5664,
+      "step": 10820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.706284737067123e-05,
+      "loss": 0.6203,
+      "step": 10830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.704147071398034e-05,
+      "loss": 0.5733,
+      "step": 10840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.702009405728945e-05,
+      "loss": 0.554,
+      "step": 10850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.699871740059856e-05,
+      "loss": 0.6272,
+      "step": 10860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.697734074390765e-05,
+      "loss": 0.5607,
+      "step": 10870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.695596408721675e-05,
+      "loss": 0.6187,
+      "step": 10880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 7.693458743052587e-05,
+      "loss": 0.5696,
+      "step": 10890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.691321077383498e-05,
+      "loss": 0.5585,
+      "step": 10900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.689183411714408e-05,
+      "loss": 0.6249,
+      "step": 10910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.68704574604532e-05,
+      "loss": 0.5538,
+      "step": 10920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.684908080376229e-05,
+      "loss": 0.6157,
+      "step": 10930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.68277041470714e-05,
+      "loss": 0.5661,
+      "step": 10940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.68063274903805e-05,
+      "loss": 0.5522,
+      "step": 10950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.678495083368962e-05,
+      "loss": 0.6285,
+      "step": 10960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.676357417699872e-05,
+      "loss": 0.5496,
+      "step": 10970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.674219752030783e-05,
+      "loss": 0.6155,
+      "step": 10980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.672082086361693e-05,
+      "loss": 0.5656,
+      "step": 10990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.669944420692604e-05,
+      "loss": 0.561,
+      "step": 11000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.667806755023514e-05,
+      "loss": 0.6313,
+      "step": 11010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.665669089354426e-05,
+      "loss": 0.5597,
+      "step": 11020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.663531423685336e-05,
+      "loss": 0.6134,
+      "step": 11030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.661393758016247e-05,
+      "loss": 0.5615,
+      "step": 11040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.659256092347157e-05,
+      "loss": 0.5518,
+      "step": 11050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.657118426678068e-05,
+      "loss": 0.6231,
+      "step": 11060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.654980761008978e-05,
+      "loss": 0.559,
+      "step": 11070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.652843095339889e-05,
+      "loss": 0.6163,
+      "step": 11080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.6507054296708e-05,
+      "loss": 0.5667,
+      "step": 11090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.648567764001711e-05,
+      "loss": 0.5631,
+      "step": 11100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.646430098332621e-05,
+      "loss": 0.6288,
+      "step": 11110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.644292432663532e-05,
+      "loss": 0.5643,
+      "step": 11120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.642154766994442e-05,
+      "loss": 0.6178,
+      "step": 11130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.640017101325353e-05,
+      "loss": 0.5725,
+      "step": 11140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.637879435656263e-05,
+      "loss": 0.5579,
+      "step": 11150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.635741769987175e-05,
+      "loss": 0.631,
+      "step": 11160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.633604104318085e-05,
+      "loss": 0.5479,
+      "step": 11170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.631466438648996e-05,
+      "loss": 0.6115,
+      "step": 11180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.629328772979906e-05,
+      "loss": 0.5702,
+      "step": 11190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.627191107310817e-05,
+      "loss": 0.5543,
+      "step": 11200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.625053441641727e-05,
+      "loss": 0.6261,
+      "step": 11210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.622915775972638e-05,
+      "loss": 0.5602,
+      "step": 11220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.62077811030355e-05,
+      "loss": 0.6155,
+      "step": 11230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.61864044463446e-05,
+      "loss": 0.5688,
+      "step": 11240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.61650277896537e-05,
+      "loss": 0.5544,
+      "step": 11250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.614365113296281e-05,
+      "loss": 0.6304,
+      "step": 11260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.612227447627191e-05,
+      "loss": 0.5553,
+      "step": 11270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.610089781958102e-05,
+      "loss": 0.6187,
+      "step": 11280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.607952116289013e-05,
+      "loss": 0.5706,
+      "step": 11290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.605814450619924e-05,
+      "loss": 0.5547,
+      "step": 11300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.603676784950834e-05,
+      "loss": 0.62,
+      "step": 11310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.601539119281745e-05,
+      "loss": 0.5592,
+      "step": 11320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.599401453612655e-05,
+      "loss": 0.6133,
+      "step": 11330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.597263787943566e-05,
+      "loss": 0.5815,
+      "step": 11340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.595126122274476e-05,
+      "loss": 0.5567,
+      "step": 11350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 7.592988456605388e-05,
+      "loss": 0.623,
+      "step": 11360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.590850790936298e-05,
+      "loss": 0.5623,
+      "step": 11370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.588713125267209e-05,
+      "loss": 0.6251,
+      "step": 11380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.586575459598119e-05,
+      "loss": 0.5729,
+      "step": 11390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.58443779392903e-05,
+      "loss": 0.5551,
+      "step": 11400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.58230012825994e-05,
+      "loss": 0.6208,
+      "step": 11410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.58016246259085e-05,
+      "loss": 0.5593,
+      "step": 11420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.578024796921762e-05,
+      "loss": 0.6249,
+      "step": 11430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.575887131252673e-05,
+      "loss": 0.5602,
+      "step": 11440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.573749465583583e-05,
+      "loss": 0.5558,
+      "step": 11450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.571611799914494e-05,
+      "loss": 0.6125,
+      "step": 11460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.569474134245404e-05,
+      "loss": 0.5575,
+      "step": 11470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.567336468576315e-05,
+      "loss": 0.6187,
+      "step": 11480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.565198802907225e-05,
+      "loss": 0.5689,
+      "step": 11490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.563061137238137e-05,
+      "loss": 0.553,
+      "step": 11500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.560923471569047e-05,
+      "loss": 0.622,
+      "step": 11510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.558785805899958e-05,
+      "loss": 0.5622,
+      "step": 11520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.556648140230868e-05,
+      "loss": 0.6229,
+      "step": 11530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.554510474561779e-05,
+      "loss": 0.5647,
+      "step": 11540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.552372808892689e-05,
+      "loss": 0.5506,
+      "step": 11550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.550235143223601e-05,
+      "loss": 0.6218,
+      "step": 11560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.548097477554511e-05,
+      "loss": 0.5543,
+      "step": 11570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.545959811885422e-05,
+      "loss": 0.6117,
+      "step": 11580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.543822146216331e-05,
+      "loss": 0.569,
+      "step": 11590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 7.541684480547243e-05,
+      "loss": 0.56,
+      "step": 11600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.539546814878153e-05,
+      "loss": 0.6231,
+      "step": 11610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.537409149209064e-05,
+      "loss": 0.565,
+      "step": 11620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.535271483539976e-05,
+      "loss": 0.6171,
+      "step": 11630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.533133817870886e-05,
+      "loss": 0.5682,
+      "step": 11640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.530996152201796e-05,
+      "loss": 0.5618,
+      "step": 11650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.528858486532707e-05,
+      "loss": 0.6262,
+      "step": 11660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.526720820863617e-05,
+      "loss": 0.5622,
+      "step": 11670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.524583155194528e-05,
+      "loss": 0.6232,
+      "step": 11680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.522445489525438e-05,
+      "loss": 0.56,
+      "step": 11690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.52030782385635e-05,
+      "loss": 0.5568,
+      "step": 11700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.51817015818726e-05,
+      "loss": 0.6131,
+      "step": 11710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.51603249251817e-05,
+      "loss": 0.557,
+      "step": 11720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.513894826849081e-05,
+      "loss": 0.6222,
+      "step": 11730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.511757161179992e-05,
+      "loss": 0.5621,
+      "step": 11740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.509619495510902e-05,
+      "loss": 0.5572,
+      "step": 11750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.507481829841813e-05,
+      "loss": 0.6251,
+      "step": 11760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.505344164172725e-05,
+      "loss": 0.5536,
+      "step": 11770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.503206498503634e-05,
+      "loss": 0.6123,
+      "step": 11780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.501068832834544e-05,
+      "loss": 0.5724,
+      "step": 11790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.498931167165456e-05,
+      "loss": 0.5607,
+      "step": 11800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.496793501496366e-05,
+      "loss": 0.6355,
+      "step": 11810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.494655835827277e-05,
+      "loss": 0.5486,
+      "step": 11820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 7.492518170158189e-05,
+      "loss": 0.6026,
+      "step": 11830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.490380504489099e-05,
+      "loss": 0.5734,
+      "step": 11840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.488242838820008e-05,
+      "loss": 0.5543,
+      "step": 11850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.486105173150919e-05,
+      "loss": 0.616,
+      "step": 11860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.48396750748183e-05,
+      "loss": 0.5626,
+      "step": 11870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.481829841812741e-05,
+      "loss": 0.614,
+      "step": 11880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.479692176143651e-05,
+      "loss": 0.5675,
+      "step": 11890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.477554510474563e-05,
+      "loss": 0.5544,
+      "step": 11900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.475416844805472e-05,
+      "loss": 0.625,
+      "step": 11910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.473279179136383e-05,
+      "loss": 0.5612,
+      "step": 11920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.471141513467293e-05,
+      "loss": 0.6091,
+      "step": 11930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.469003847798205e-05,
+      "loss": 0.5697,
+      "step": 11940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.466866182129115e-05,
+      "loss": 0.5563,
+      "step": 11950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.464728516460026e-05,
+      "loss": 0.6193,
+      "step": 11960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.462590850790938e-05,
+      "loss": 0.5459,
+      "step": 11970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.460453185121847e-05,
+      "loss": 0.6076,
+      "step": 11980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.458315519452757e-05,
+      "loss": 0.5711,
+      "step": 11990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.456177853783669e-05,
+      "loss": 0.5569,
+      "step": 12000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.45404018811458e-05,
+      "loss": 0.615,
+      "step": 12010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.45190252244549e-05,
+      "loss": 0.5621,
+      "step": 12020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.4497648567764e-05,
+      "loss": 0.6007,
+      "step": 12030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.447627191107311e-05,
+      "loss": 0.5629,
+      "step": 12040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.445489525438221e-05,
+      "loss": 0.559,
+      "step": 12050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.443351859769132e-05,
+      "loss": 0.6299,
+      "step": 12060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 7.441214194100044e-05,
+      "loss": 0.5519,
+      "step": 12070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.439076528430954e-05,
+      "loss": 0.6161,
+      "step": 12080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.436938862761864e-05,
+      "loss": 0.5714,
+      "step": 12090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.434801197092776e-05,
+      "loss": 0.5558,
+      "step": 12100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.432663531423685e-05,
+      "loss": 0.6178,
+      "step": 12110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.430525865754596e-05,
+      "loss": 0.5439,
+      "step": 12120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.428388200085506e-05,
+      "loss": 0.595,
+      "step": 12130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.426250534416418e-05,
+      "loss": 0.5671,
+      "step": 12140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.424112868747328e-05,
+      "loss": 0.5569,
+      "step": 12150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.421975203078239e-05,
+      "loss": 0.6265,
+      "step": 12160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.41983753740915e-05,
+      "loss": 0.5506,
+      "step": 12170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.41769987174006e-05,
+      "loss": 0.6197,
+      "step": 12180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.41556220607097e-05,
+      "loss": 0.5604,
+      "step": 12190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.413424540401881e-05,
+      "loss": 0.5427,
+      "step": 12200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.411286874732793e-05,
+      "loss": 0.6256,
+      "step": 12210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.409149209063703e-05,
+      "loss": 0.5521,
+      "step": 12220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.407011543394613e-05,
+      "loss": 0.6234,
+      "step": 12230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.404873877725524e-05,
+      "loss": 0.5811,
+      "step": 12240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.402736212056434e-05,
+      "loss": 0.5592,
+      "step": 12250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.400598546387345e-05,
+      "loss": 0.6214,
+      "step": 12260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.398460880718257e-05,
+      "loss": 0.5693,
+      "step": 12270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.396323215049167e-05,
+      "loss": 0.6104,
+      "step": 12280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.394185549380078e-05,
+      "loss": 0.5646,
+      "step": 12290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.392047883710988e-05,
+      "loss": 0.5454,
+      "step": 12300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.389910218041898e-05,
+      "loss": 0.6205,
+      "step": 12310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.387772552372809e-05,
+      "loss": 0.5636,
+      "step": 12320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.385634886703719e-05,
+      "loss": 0.6103,
+      "step": 12330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.383497221034631e-05,
+      "loss": 0.564,
+      "step": 12340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.381359555365542e-05,
+      "loss": 0.5465,
+      "step": 12350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.379221889696452e-05,
+      "loss": 0.6199,
+      "step": 12360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.377084224027362e-05,
+      "loss": 0.5572,
+      "step": 12370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.374946558358273e-05,
+      "loss": 0.6157,
+      "step": 12380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.372808892689183e-05,
+      "loss": 0.5651,
+      "step": 12390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.370671227020094e-05,
+      "loss": 0.5573,
+      "step": 12400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.368533561351006e-05,
+      "loss": 0.6123,
+      "step": 12410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.366395895681916e-05,
+      "loss": 0.5453,
+      "step": 12420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.364258230012827e-05,
+      "loss": 0.6098,
+      "step": 12430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.362120564343737e-05,
+      "loss": 0.5699,
+      "step": 12440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.359982898674647e-05,
+      "loss": 0.5535,
+      "step": 12450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.357845233005558e-05,
+      "loss": 0.6205,
+      "step": 12460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.355707567336468e-05,
+      "loss": 0.5481,
+      "step": 12470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.35356990166738e-05,
+      "loss": 0.6061,
+      "step": 12480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.35143223599829e-05,
+      "loss": 0.5694,
+      "step": 12490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.349294570329201e-05,
+      "loss": 0.5581,
+      "step": 12500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.347156904660112e-05,
+      "loss": 0.6265,
+      "step": 12510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.345019238991022e-05,
+      "loss": 0.5452,
+      "step": 12520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.342881573321932e-05,
+      "loss": 0.6147,
+      "step": 12530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 7.340743907652844e-05,
+      "loss": 0.5748,
+      "step": 12540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.338606241983755e-05,
+      "loss": 0.5463,
+      "step": 12550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.336468576314665e-05,
+      "loss": 0.6215,
+      "step": 12560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.334330910645574e-05,
+      "loss": 0.5459,
+      "step": 12570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.332193244976486e-05,
+      "loss": 0.6213,
+      "step": 12580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.330055579307396e-05,
+      "loss": 0.5602,
+      "step": 12590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.327917913638307e-05,
+      "loss": 0.5452,
+      "step": 12600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.325780247969219e-05,
+      "loss": 0.6142,
+      "step": 12610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.323642582300129e-05,
+      "loss": 0.5541,
+      "step": 12620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.32150491663104e-05,
+      "loss": 0.6155,
+      "step": 12630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.31936725096195e-05,
+      "loss": 0.5671,
+      "step": 12640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.31722958529286e-05,
+      "loss": 0.5519,
+      "step": 12650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.315091919623771e-05,
+      "loss": 0.6196,
+      "step": 12660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.312954253954681e-05,
+      "loss": 0.5538,
+      "step": 12670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.310816588285593e-05,
+      "loss": 0.6077,
+      "step": 12680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.308678922616504e-05,
+      "loss": 0.5656,
+      "step": 12690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.306541256947413e-05,
+      "loss": 0.5537,
+      "step": 12700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.304403591278325e-05,
+      "loss": 0.6226,
+      "step": 12710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.302265925609235e-05,
+      "loss": 0.5604,
+      "step": 12720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.300128259940145e-05,
+      "loss": 0.6109,
+      "step": 12730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.297990594271056e-05,
+      "loss": 0.5684,
+      "step": 12740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.295852928601968e-05,
+      "loss": 0.5563,
+      "step": 12750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.293715262932878e-05,
+      "loss": 0.6166,
+      "step": 12760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 7.291577597263787e-05,
+      "loss": 0.5489,
+      "step": 12770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.289439931594699e-05,
+      "loss": 0.6062,
+      "step": 12780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.28730226592561e-05,
+      "loss": 0.5662,
+      "step": 12790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.28516460025652e-05,
+      "loss": 0.5488,
+      "step": 12800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.283026934587432e-05,
+      "loss": 0.6219,
+      "step": 12810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.280889268918342e-05,
+      "loss": 0.5632,
+      "step": 12820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.278751603249251e-05,
+      "loss": 0.6184,
+      "step": 12830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.276613937580162e-05,
+      "loss": 0.5746,
+      "step": 12840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.274476271911074e-05,
+      "loss": 0.5566,
+      "step": 12850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.272338606241984e-05,
+      "loss": 0.6176,
+      "step": 12860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.270200940572895e-05,
+      "loss": 0.5477,
+      "step": 12870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.268063274903806e-05,
+      "loss": 0.6051,
+      "step": 12880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.265925609234717e-05,
+      "loss": 0.5711,
+      "step": 12890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.263787943565626e-05,
+      "loss": 0.5521,
+      "step": 12900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.261650277896538e-05,
+      "loss": 0.6286,
+      "step": 12910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.259512612227448e-05,
+      "loss": 0.5531,
+      "step": 12920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.257374946558359e-05,
+      "loss": 0.6108,
+      "step": 12930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.255237280889269e-05,
+      "loss": 0.5689,
+      "step": 12940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.253099615220181e-05,
+      "loss": 0.5504,
+      "step": 12950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.25096194955109e-05,
+      "loss": 0.6156,
+      "step": 12960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.248824283882e-05,
+      "loss": 0.5504,
+      "step": 12970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.246686618212912e-05,
+      "loss": 0.6042,
+      "step": 12980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.244548952543823e-05,
+      "loss": 0.5681,
+      "step": 12990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 7.242411286874733e-05,
+      "loss": 0.5633,
+      "step": 13000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.240273621205644e-05,
+      "loss": 0.6267,
+      "step": 13010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.238135955536554e-05,
+      "loss": 0.5553,
+      "step": 13020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.235998289867464e-05,
+      "loss": 0.6106,
+      "step": 13030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.233860624198375e-05,
+      "loss": 0.5787,
+      "step": 13040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.231722958529287e-05,
+      "loss": 0.5542,
+      "step": 13050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.229585292860197e-05,
+      "loss": 0.6155,
+      "step": 13060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.227447627191108e-05,
+      "loss": 0.5513,
+      "step": 13070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.22530996152202e-05,
+      "loss": 0.6095,
+      "step": 13080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.223172295852929e-05,
+      "loss": 0.5629,
+      "step": 13090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.221034630183839e-05,
+      "loss": 0.5403,
+      "step": 13100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.21889696451475e-05,
+      "loss": 0.6222,
+      "step": 13110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.216759298845661e-05,
+      "loss": 0.5589,
+      "step": 13120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.214621633176572e-05,
+      "loss": 0.6008,
+      "step": 13130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.212483967507482e-05,
+      "loss": 0.5636,
+      "step": 13140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.210346301838393e-05,
+      "loss": 0.5466,
+      "step": 13150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.208208636169303e-05,
+      "loss": 0.6106,
+      "step": 13160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.206070970500213e-05,
+      "loss": 0.5593,
+      "step": 13170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.203933304831125e-05,
+      "loss": 0.6045,
+      "step": 13180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.201795639162036e-05,
+      "loss": 0.5636,
+      "step": 13190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.199657973492946e-05,
+      "loss": 0.5514,
+      "step": 13200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.197520307823857e-05,
+      "loss": 0.6155,
+      "step": 13210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.195382642154767e-05,
+      "loss": 0.5519,
+      "step": 13220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.193244976485678e-05,
+      "loss": 0.615,
+      "step": 13230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.191107310816588e-05,
+      "loss": 0.562,
+      "step": 13240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.1889696451475e-05,
+      "loss": 0.5528,
+      "step": 13250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.18683197947841e-05,
+      "loss": 0.6204,
+      "step": 13260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.184694313809321e-05,
+      "loss": 0.546,
+      "step": 13270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.182556648140231e-05,
+      "loss": 0.6118,
+      "step": 13280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.180418982471142e-05,
+      "loss": 0.5642,
+      "step": 13290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.178281316802052e-05,
+      "loss": 0.5498,
+      "step": 13300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.176143651132962e-05,
+      "loss": 0.6148,
+      "step": 13310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.174005985463874e-05,
+      "loss": 0.5511,
+      "step": 13320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.171868319794785e-05,
+      "loss": 0.6013,
+      "step": 13330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.169730654125695e-05,
+      "loss": 0.5647,
+      "step": 13340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.167592988456606e-05,
+      "loss": 0.5514,
+      "step": 13350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.165455322787516e-05,
+      "loss": 0.6165,
+      "step": 13360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.163317657118427e-05,
+      "loss": 0.5494,
+      "step": 13370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.161179991449337e-05,
+      "loss": 0.6209,
+      "step": 13380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.159042325780249e-05,
+      "loss": 0.5546,
+      "step": 13390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.156904660111159e-05,
+      "loss": 0.553,
+      "step": 13400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.15476699444207e-05,
+      "loss": 0.6189,
+      "step": 13410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.15262932877298e-05,
+      "loss": 0.5616,
+      "step": 13420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.15049166310389e-05,
+      "loss": 0.6123,
+      "step": 13430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.148353997434801e-05,
+      "loss": 0.5707,
+      "step": 13440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.146216331765713e-05,
+      "loss": 0.543,
+      "step": 13450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.144078666096623e-05,
+      "loss": 0.616,
+      "step": 13460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.141941000427534e-05,
+      "loss": 0.5563,
+      "step": 13470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.139803334758444e-05,
+      "loss": 0.6117,
+      "step": 13480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.137665669089355e-05,
+      "loss": 0.5642,
+      "step": 13490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.135528003420265e-05,
+      "loss": 0.5511,
+      "step": 13500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.133390337751176e-05,
+      "loss": 0.6144,
+      "step": 13510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.131252672082087e-05,
+      "loss": 0.5564,
+      "step": 13520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.129115006412998e-05,
+      "loss": 0.6064,
+      "step": 13530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.126977340743908e-05,
+      "loss": 0.5533,
+      "step": 13540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.124839675074819e-05,
+      "loss": 0.5549,
+      "step": 13550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.122702009405729e-05,
+      "loss": 0.6233,
+      "step": 13560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.12056434373664e-05,
+      "loss": 0.5525,
+      "step": 13570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.11842667806755e-05,
+      "loss": 0.6054,
+      "step": 13580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.116289012398462e-05,
+      "loss": 0.5663,
+      "step": 13590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.114151346729372e-05,
+      "loss": 0.5531,
+      "step": 13600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.112013681060283e-05,
+      "loss": 0.6149,
+      "step": 13610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.109876015391193e-05,
+      "loss": 0.5489,
+      "step": 13620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.107738349722104e-05,
+      "loss": 0.6104,
+      "step": 13630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.105600684053014e-05,
+      "loss": 0.5654,
+      "step": 13640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.103463018383925e-05,
+      "loss": 0.5569,
+      "step": 13650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.101325352714836e-05,
+      "loss": 0.6236,
+      "step": 13660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.099187687045747e-05,
+      "loss": 0.5589,
+      "step": 13670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.097050021376657e-05,
+      "loss": 0.6043,
+      "step": 13680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.094912355707568e-05,
+      "loss": 0.5591,
+      "step": 13690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.092774690038478e-05,
+      "loss": 0.5528,
+      "step": 13700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.090637024369389e-05,
+      "loss": 0.6183,
+      "step": 13710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.0884993587003e-05,
+      "loss": 0.5464,
+      "step": 13720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.086361693031211e-05,
+      "loss": 0.6075,
+      "step": 13730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.084224027362121e-05,
+      "loss": 0.5572,
+      "step": 13740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.08208636169303e-05,
+      "loss": 0.5489,
+      "step": 13750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.079948696023942e-05,
+      "loss": 0.6223,
+      "step": 13760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.077811030354853e-05,
+      "loss": 0.548,
+      "step": 13770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.075673364685763e-05,
+      "loss": 0.6048,
+      "step": 13780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.073535699016675e-05,
+      "loss": 0.5744,
+      "step": 13790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.071398033347585e-05,
+      "loss": 0.543,
+      "step": 13800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.069260367678495e-05,
+      "loss": 0.6093,
+      "step": 13810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.067122702009405e-05,
+      "loss": 0.5512,
+      "step": 13820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.064985036340317e-05,
+      "loss": 0.6117,
+      "step": 13830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.062847370671227e-05,
+      "loss": 0.5643,
+      "step": 13840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.060709705002138e-05,
+      "loss": 0.5591,
+      "step": 13850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.05857203933305e-05,
+      "loss": 0.6178,
+      "step": 13860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.05643437366396e-05,
+      "loss": 0.5458,
+      "step": 13870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.054296707994869e-05,
+      "loss": 0.6113,
+      "step": 13880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.052159042325781e-05,
+      "loss": 0.5706,
+      "step": 13890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.050021376656691e-05,
+      "loss": 0.5493,
+      "step": 13900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.047883710987602e-05,
+      "loss": 0.6264,
+      "step": 13910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.045746045318512e-05,
+      "loss": 0.5506,
+      "step": 13920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.043608379649424e-05,
+      "loss": 0.6167,
+      "step": 13930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.041470713980333e-05,
+      "loss": 0.5699,
+      "step": 13940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.039333048311244e-05,
+      "loss": 0.5513,
+      "step": 13950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.037195382642155e-05,
+      "loss": 0.6107,
+      "step": 13960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.035057716973066e-05,
+      "loss": 0.5505,
+      "step": 13970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.032920051303976e-05,
+      "loss": 0.6095,
+      "step": 13980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.030782385634888e-05,
+      "loss": 0.5674,
+      "step": 13990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.028644719965799e-05,
+      "loss": 0.5424,
+      "step": 14000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.026507054296708e-05,
+      "loss": 0.6185,
+      "step": 14010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.024369388627618e-05,
+      "loss": 0.5587,
+      "step": 14020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.02223172295853e-05,
+      "loss": 0.6128,
+      "step": 14030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.02009405728944e-05,
+      "loss": 0.567,
+      "step": 14040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.017956391620351e-05,
+      "loss": 0.555,
+      "step": 14050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.015818725951263e-05,
+      "loss": 0.6101,
+      "step": 14060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.013681060282172e-05,
+      "loss": 0.5385,
+      "step": 14070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.011543394613082e-05,
+      "loss": 0.6093,
+      "step": 14080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.009405728943993e-05,
+      "loss": 0.569,
+      "step": 14090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.007268063274904e-05,
+      "loss": 0.5558,
+      "step": 14100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.005130397605815e-05,
+      "loss": 0.6149,
+      "step": 14110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.002992731936725e-05,
+      "loss": 0.5541,
+      "step": 14120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.000855066267637e-05,
+      "loss": 0.6064,
+      "step": 14130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.998717400598546e-05,
+      "loss": 0.5779,
+      "step": 14140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.996579734929457e-05,
+      "loss": 0.5517,
+      "step": 14150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.994442069260368e-05,
+      "loss": 0.6183,
+      "step": 14160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.992304403591279e-05,
+      "loss": 0.555,
+      "step": 14170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.99016673792219e-05,
+      "loss": 0.6076,
+      "step": 14180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9880290722531e-05,
+      "loss": 0.5624,
+      "step": 14190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.98589140658401e-05,
+      "loss": 0.5534,
+      "step": 14200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983753740914921e-05,
+      "loss": 0.6184,
+      "step": 14210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981616075245831e-05,
+      "loss": 0.5568,
+      "step": 14220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979478409576743e-05,
+      "loss": 0.6007,
+      "step": 14230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977340743907653e-05,
+      "loss": 0.5623,
+      "step": 14240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975203078238564e-05,
+      "loss": 0.5427,
+      "step": 14250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973065412569474e-05,
+      "loss": 0.6178,
+      "step": 14260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970927746900385e-05,
+      "loss": 0.5477,
+      "step": 14270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.968790081231295e-05,
+      "loss": 0.6166,
+      "step": 14280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.966652415562206e-05,
+      "loss": 0.5634,
+      "step": 14290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964514749893117e-05,
+      "loss": 0.5478,
+      "step": 14300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962377084224028e-05,
+      "loss": 0.616,
+      "step": 14310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960239418554938e-05,
+      "loss": 0.5468,
+      "step": 14320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.958101752885849e-05,
+      "loss": 0.6115,
+      "step": 14330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955964087216759e-05,
+      "loss": 0.5634,
+      "step": 14340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95382642154767e-05,
+      "loss": 0.5445,
+      "step": 14350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95168875587858e-05,
+      "loss": 0.6135,
+      "step": 14360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949551090209492e-05,
+      "loss": 0.5562,
+      "step": 14370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947413424540402e-05,
+      "loss": 0.6056,
+      "step": 14380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.945275758871313e-05,
+      "loss": 0.5616,
+      "step": 14390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943138093202223e-05,
+      "loss": 0.5483,
+      "step": 14400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941000427533134e-05,
+      "loss": 0.6095,
+      "step": 14410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.938862761864044e-05,
+      "loss": 0.5591,
+      "step": 14420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.936725096194956e-05,
+      "loss": 0.6013,
+      "step": 14430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.934587430525867e-05,
+      "loss": 0.563,
+      "step": 14440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.932449764856777e-05,
+      "loss": 0.5485,
+      "step": 14450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.930312099187687e-05,
+      "loss": 0.6175,
+      "step": 14460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.928174433518598e-05,
+      "loss": 0.5431,
+      "step": 14470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.926036767849508e-05,
+      "loss": 0.608,
+      "step": 14480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.923899102180419e-05,
+      "loss": 0.5576,
+      "step": 14490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.92176143651133e-05,
+      "loss": 0.5476,
+      "step": 14500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.919623770842241e-05,
+      "loss": 0.6157,
+      "step": 14510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.917486105173151e-05,
+      "loss": 0.5514,
+      "step": 14520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.915348439504062e-05,
+      "loss": 0.6143,
+      "step": 14530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.913210773834972e-05,
+      "loss": 0.5628,
+      "step": 14540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.911073108165883e-05,
+      "loss": 0.5491,
+      "step": 14550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.908935442496793e-05,
+      "loss": 0.6153,
+      "step": 14560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.906797776827705e-05,
+      "loss": 0.5478,
+      "step": 14570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.904660111158616e-05,
+      "loss": 0.615,
+      "step": 14580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.902522445489526e-05,
+      "loss": 0.5627,
+      "step": 14590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.900384779820436e-05,
+      "loss": 0.5443,
+      "step": 14600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.898247114151347e-05,
+      "loss": 0.6171,
+      "step": 14610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.896109448482257e-05,
+      "loss": 0.55,
+      "step": 14620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.893971782813168e-05,
+      "loss": 0.6092,
+      "step": 14630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.89183411714408e-05,
+      "loss": 0.5593,
+      "step": 14640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.88969645147499e-05,
+      "loss": 0.5364,
+      "step": 14650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.8875587858059e-05,
+      "loss": 0.6095,
+      "step": 14660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.885421120136811e-05,
+      "loss": 0.5511,
+      "step": 14670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.883283454467721e-05,
+      "loss": 0.6058,
+      "step": 14680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.881145788798632e-05,
+      "loss": 0.5652,
+      "step": 14690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.879008123129544e-05,
+      "loss": 0.5472,
+      "step": 14700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.876870457460454e-05,
+      "loss": 0.6106,
+      "step": 14710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.874732791791365e-05,
+      "loss": 0.5514,
+      "step": 14720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.872595126122274e-05,
+      "loss": 0.6062,
+      "step": 14730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.870457460453185e-05,
+      "loss": 0.5597,
+      "step": 14740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.868319794784096e-05,
+      "loss": 0.5524,
+      "step": 14750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.866182129115006e-05,
+      "loss": 0.6109,
+      "step": 14760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.864044463445918e-05,
+      "loss": 0.5465,
+      "step": 14770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.861906797776829e-05,
+      "loss": 0.6039,
+      "step": 14780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.859769132107739e-05,
+      "loss": 0.5574,
+      "step": 14790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.85763146643865e-05,
+      "loss": 0.5574,
+      "step": 14800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.85549380076956e-05,
+      "loss": 0.6085,
+      "step": 14810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.85335613510047e-05,
+      "loss": 0.5468,
+      "step": 14820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.851218469431381e-05,
+      "loss": 0.6182,
+      "step": 14830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.849080803762293e-05,
+      "loss": 0.5614,
+      "step": 14840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.846943138093203e-05,
+      "loss": 0.5475,
+      "step": 14850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.844805472424112e-05,
+      "loss": 0.6165,
+      "step": 14860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.842667806755024e-05,
+      "loss": 0.5572,
+      "step": 14870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.840530141085934e-05,
+      "loss": 0.6132,
+      "step": 14880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.838392475416845e-05,
+      "loss": 0.5622,
+      "step": 14890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.836254809747755e-05,
+      "loss": 0.5423,
+      "step": 14900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.834117144078667e-05,
+      "loss": 0.6082,
+      "step": 14910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.831979478409578e-05,
+      "loss": 0.5545,
+      "step": 14920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.829841812740487e-05,
+      "loss": 0.6018,
+      "step": 14930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.827704147071399e-05,
+      "loss": 0.5695,
+      "step": 14940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.825566481402309e-05,
+      "loss": 0.5433,
+      "step": 14950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.82342881573322e-05,
+      "loss": 0.6234,
+      "step": 14960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.821291150064131e-05,
+      "loss": 0.5469,
+      "step": 14970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.819153484395042e-05,
+      "loss": 0.599,
+      "step": 14980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.817015818725951e-05,
+      "loss": 0.556,
+      "step": 14990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.814878153056861e-05,
+      "loss": 0.5496,
+      "step": 15000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.812740487387773e-05,
+      "loss": 0.6209,
+      "step": 15010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.810602821718684e-05,
+      "loss": 0.5453,
+      "step": 15020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.808465156049594e-05,
+      "loss": 0.6127,
+      "step": 15030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.806327490380506e-05,
+      "loss": 0.5519,
+      "step": 15040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.804189824711415e-05,
+      "loss": 0.5505,
+      "step": 15050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.802052159042325e-05,
+      "loss": 0.6182,
+      "step": 15060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.799914493373237e-05,
+      "loss": 0.5555,
+      "step": 15070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.797776827704148e-05,
+      "loss": 0.6014,
+      "step": 15080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.795639162035058e-05,
+      "loss": 0.566,
+      "step": 15090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.793501496365968e-05,
+      "loss": 0.5498,
+      "step": 15100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.79136383069688e-05,
+      "loss": 0.6135,
+      "step": 15110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.78922616502779e-05,
+      "loss": 0.5434,
+      "step": 15120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.7870884993587e-05,
+      "loss": 0.5988,
+      "step": 15130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.784950833689612e-05,
+      "loss": 0.564,
+      "step": 15140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.782813168020522e-05,
+      "loss": 0.5464,
+      "step": 15150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.780675502351433e-05,
+      "loss": 0.612,
+      "step": 15160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.778537836682343e-05,
+      "loss": 0.5407,
+      "step": 15170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.776400171013253e-05,
+      "loss": 0.608,
+      "step": 15180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.774262505344164e-05,
+      "loss": 0.5674,
+      "step": 15190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.772124839675074e-05,
+      "loss": 0.5369,
+      "step": 15200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.769987174005986e-05,
+      "loss": 0.6188,
+      "step": 15210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.767849508336897e-05,
+      "loss": 0.5516,
+      "step": 15220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.765711842667807e-05,
+      "loss": 0.6131,
+      "step": 15230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.763574176998719e-05,
+      "loss": 0.5624,
+      "step": 15240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.761436511329628e-05,
+      "loss": 0.5444,
+      "step": 15250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.759298845660538e-05,
+      "loss": 0.6128,
+      "step": 15260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.757161179991449e-05,
+      "loss": 0.5486,
+      "step": 15270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.75502351432236e-05,
+      "loss": 0.5926,
+      "step": 15280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.752885848653271e-05,
+      "loss": 0.5479,
+      "step": 15290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.750748182984182e-05,
+      "loss": 0.542,
+      "step": 15300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.748610517315092e-05,
+      "loss": 0.6213,
+      "step": 15310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.746472851646002e-05,
+      "loss": 0.5526,
+      "step": 15320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.744335185976913e-05,
+      "loss": 0.607,
+      "step": 15330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.742197520307825e-05,
+      "loss": 0.5571,
+      "step": 15340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.740059854638735e-05,
+      "loss": 0.5377,
+      "step": 15350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.737922188969646e-05,
+      "loss": 0.6241,
+      "step": 15360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.735784523300556e-05,
+      "loss": 0.5449,
+      "step": 15370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.733646857631467e-05,
+      "loss": 0.6074,
+      "step": 15380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.731509191962377e-05,
+      "loss": 0.5726,
+      "step": 15390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.729371526293287e-05,
+      "loss": 0.5424,
+      "step": 15400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.727233860624199e-05,
+      "loss": 0.6141,
+      "step": 15410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.72509619495511e-05,
+      "loss": 0.5557,
+      "step": 15420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.72295852928602e-05,
+      "loss": 0.598,
+      "step": 15430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.72082086361693e-05,
+      "loss": 0.5532,
+      "step": 15440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.718683197947841e-05,
+      "loss": 0.5503,
+      "step": 15450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.716545532278751e-05,
+      "loss": 0.6274,
+      "step": 15460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.714407866609662e-05,
+      "loss": 0.5524,
+      "step": 15470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.712270200940574e-05,
+      "loss": 0.6094,
+      "step": 15480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.710132535271484e-05,
+      "loss": 0.5622,
+      "step": 15490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.707994869602395e-05,
+      "loss": 0.5431,
+      "step": 15500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.705857203933305e-05,
+      "loss": 0.6123,
+      "step": 15510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.703719538264216e-05,
+      "loss": 0.5468,
+      "step": 15520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.701581872595126e-05,
+      "loss": 0.612,
+      "step": 15530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.699444206926036e-05,
+      "loss": 0.5655,
+      "step": 15540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.697306541256948e-05,
+      "loss": 0.5501,
+      "step": 15550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.695168875587859e-05,
+      "loss": 0.6009,
+      "step": 15560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.693031209918769e-05,
+      "loss": 0.5442,
+      "step": 15570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.69089354424968e-05,
+      "loss": 0.5972,
+      "step": 15580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.68875587858059e-05,
+      "loss": 0.5633,
+      "step": 15590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.6866182129115e-05,
+      "loss": 0.5434,
+      "step": 15600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.684480547242412e-05,
+      "loss": 0.6107,
+      "step": 15610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.682342881573323e-05,
+      "loss": 0.5438,
+      "step": 15620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.680205215904233e-05,
+      "loss": 0.6036,
+      "step": 15630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.678067550235144e-05,
+      "loss": 0.5584,
+      "step": 15640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.675929884566054e-05,
+      "loss": 0.5496,
+      "step": 15650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.673792218896965e-05,
+      "loss": 0.6098,
+      "step": 15660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.671654553227875e-05,
+      "loss": 0.559,
+      "step": 15670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.669516887558787e-05,
+      "loss": 0.6044,
+      "step": 15680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.667379221889697e-05,
+      "loss": 0.5554,
+      "step": 15690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.665241556220608e-05,
+      "loss": 0.5469,
+      "step": 15700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.663103890551518e-05,
+      "loss": 0.6054,
+      "step": 15710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.660966224882429e-05,
+      "loss": 0.5529,
+      "step": 15720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.658828559213339e-05,
+      "loss": 0.6096,
+      "step": 15730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.65669089354425e-05,
+      "loss": 0.5618,
+      "step": 15740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.654553227875161e-05,
+      "loss": 0.5471,
+      "step": 15750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.652415562206072e-05,
+      "loss": 0.6238,
+      "step": 15760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.650277896536982e-05,
+      "loss": 0.5481,
+      "step": 15770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.648140230867893e-05,
+      "loss": 0.603,
+      "step": 15780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.646002565198803e-05,
+      "loss": 0.5574,
+      "step": 15790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.643864899529714e-05,
+      "loss": 0.5511,
+      "step": 15800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.641727233860624e-05,
+      "loss": 0.6084,
+      "step": 15810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.639589568191536e-05,
+      "loss": 0.5473,
+      "step": 15820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.637451902522446e-05,
+      "loss": 0.6085,
+      "step": 15830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.635314236853355e-05,
+      "loss": 0.5656,
+      "step": 15840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.633176571184267e-05,
+      "loss": 0.5479,
+      "step": 15850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.631038905515178e-05,
+      "loss": 0.5982,
+      "step": 15860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.628901239846088e-05,
+      "loss": 0.5513,
+      "step": 15870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.626763574177e-05,
+      "loss": 0.5952,
+      "step": 15880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.62462590850791e-05,
+      "loss": 0.5609,
+      "step": 15890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.622488242838821e-05,
+      "loss": 0.5543,
+      "step": 15900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.62035057716973e-05,
+      "loss": 0.611,
+      "step": 15910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.618212911500642e-05,
+      "loss": 0.556,
+      "step": 15920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.616075245831552e-05,
+      "loss": 0.6082,
+      "step": 15930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.613937580162463e-05,
+      "loss": 0.5591,
+      "step": 15940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.611799914493374e-05,
+      "loss": 0.5439,
+      "step": 15950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.609662248824285e-05,
+      "loss": 0.6095,
+      "step": 15960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.607524583155194e-05,
+      "loss": 0.5463,
+      "step": 15970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.605386917486104e-05,
+      "loss": 0.6001,
+      "step": 15980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.603249251817016e-05,
+      "loss": 0.567,
+      "step": 15990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.601111586147927e-05,
+      "loss": 0.5419,
+      "step": 16000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.598973920478837e-05,
+      "loss": 0.6156,
+      "step": 16010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.596836254809749e-05,
+      "loss": 0.5433,
+      "step": 16020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.59469858914066e-05,
+      "loss": 0.5996,
+      "step": 16030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.592560923471568e-05,
+      "loss": 0.5492,
+      "step": 16040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.59042325780248e-05,
+      "loss": 0.5573,
+      "step": 16050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.588285592133391e-05,
+      "loss": 0.6128,
+      "step": 16060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.586147926464301e-05,
+      "loss": 0.5459,
+      "step": 16070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.584010260795212e-05,
+      "loss": 0.5999,
+      "step": 16080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.581872595126123e-05,
+      "loss": 0.5648,
+      "step": 16090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.579734929457033e-05,
+      "loss": 0.5447,
+      "step": 16100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.577597263787943e-05,
+      "loss": 0.6173,
+      "step": 16110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.575459598118855e-05,
+      "loss": 0.5517,
+      "step": 16120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.573321932449765e-05,
+      "loss": 0.5995,
+      "step": 16130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.571184266780676e-05,
+      "loss": 0.5584,
+      "step": 16140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.569046601111588e-05,
+      "loss": 0.5453,
+      "step": 16150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.566908935442498e-05,
+      "loss": 0.6224,
+      "step": 16160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.564771269773407e-05,
+      "loss": 0.5403,
+      "step": 16170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.562633604104318e-05,
+      "loss": 0.6052,
+      "step": 16180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.56049593843523e-05,
+      "loss": 0.5669,
+      "step": 16190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.55835827276614e-05,
+      "loss": 0.5475,
+      "step": 16200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.55622060709705e-05,
+      "loss": 0.5999,
+      "step": 16210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.554082941427962e-05,
+      "loss": 0.5393,
+      "step": 16220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.551945275758871e-05,
+      "loss": 0.6079,
+      "step": 16230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.549807610089782e-05,
+      "loss": 0.5649,
+      "step": 16240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.547669944420692e-05,
+      "loss": 0.5443,
+      "step": 16250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.545532278751604e-05,
+      "loss": 0.6134,
+      "step": 16260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.543394613082514e-05,
+      "loss": 0.5366,
+      "step": 16270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.541256947413425e-05,
+      "loss": 0.6047,
+      "step": 16280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.539119281744335e-05,
+      "loss": 0.5592,
+      "step": 16290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.536981616075246e-05,
+      "loss": 0.5327,
+      "step": 16300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.534843950406156e-05,
+      "loss": 0.6025,
+      "step": 16310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.532706284737068e-05,
+      "loss": 0.5395,
+      "step": 16320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.530568619067978e-05,
+      "loss": 0.605,
+      "step": 16330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.528430953398889e-05,
+      "loss": 0.568,
+      "step": 16340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.526293287729799e-05,
+      "loss": 0.5483,
+      "step": 16350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.52415562206071e-05,
+      "loss": 0.6058,
+      "step": 16360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.52201795639162e-05,
+      "loss": 0.5525,
+      "step": 16370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.51988029072253e-05,
+      "loss": 0.6069,
+      "step": 16380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.517742625053442e-05,
+      "loss": 0.5755,
+      "step": 16390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.515604959384353e-05,
+      "loss": 0.5432,
+      "step": 16400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.513467293715263e-05,
+      "loss": 0.6073,
+      "step": 16410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.511329628046174e-05,
+      "loss": 0.5423,
+      "step": 16420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.509191962377084e-05,
+      "loss": 0.6055,
+      "step": 16430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.507054296707995e-05,
+      "loss": 0.5449,
+      "step": 16440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.504916631038905e-05,
+      "loss": 0.5477,
+      "step": 16450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.502778965369817e-05,
+      "loss": 0.6013,
+      "step": 16460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.500641299700727e-05,
+      "loss": 0.5393,
+      "step": 16470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.498503634031638e-05,
+      "loss": 0.6078,
+      "step": 16480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.496365968362548e-05,
+      "loss": 0.5569,
+      "step": 16490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.494228302693459e-05,
+      "loss": 0.5439,
+      "step": 16500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.492090637024369e-05,
+      "loss": 0.6167,
+      "step": 16510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.48995297135528e-05,
+      "loss": 0.546,
+      "step": 16520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.487815305686191e-05,
+      "loss": 0.6,
+      "step": 16530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.485677640017102e-05,
+      "loss": 0.5609,
+      "step": 16540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.483539974348012e-05,
+      "loss": 0.5456,
+      "step": 16550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.481402308678923e-05,
+      "loss": 0.6172,
+      "step": 16560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.479264643009833e-05,
+      "loss": 0.5521,
+      "step": 16570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.477126977340744e-05,
+      "loss": 0.6042,
+      "step": 16580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.474989311671656e-05,
+      "loss": 0.5636,
+      "step": 16590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.472851646002566e-05,
+      "loss": 0.5445,
+      "step": 16600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.470713980333476e-05,
+      "loss": 0.6134,
+      "step": 16610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.468576314664387e-05,
+      "loss": 0.5471,
+      "step": 16620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.466438648995297e-05,
+      "loss": 0.5964,
+      "step": 16630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.464300983326208e-05,
+      "loss": 0.5617,
+      "step": 16640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.462163317657118e-05,
+      "loss": 0.5519,
+      "step": 16650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.46002565198803e-05,
+      "loss": 0.6119,
+      "step": 16660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.45788798631894e-05,
+      "loss": 0.5514,
+      "step": 16670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.455750320649851e-05,
+      "loss": 0.6003,
+      "step": 16680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.453612654980761e-05,
+      "loss": 0.5506,
+      "step": 16690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.451474989311672e-05,
+      "loss": 0.5413,
+      "step": 16700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.449337323642582e-05,
+      "loss": 0.6038,
+      "step": 16710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.447199657973493e-05,
+      "loss": 0.5328,
+      "step": 16720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.445061992304405e-05,
+      "loss": 0.6035,
+      "step": 16730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.442924326635315e-05,
+      "loss": 0.5497,
+      "step": 16740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.440786660966225e-05,
+      "loss": 0.5468,
+      "step": 16750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.438648995297136e-05,
+      "loss": 0.6047,
+      "step": 16760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.436511329628046e-05,
+      "loss": 0.5367,
+      "step": 16770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.434373663958957e-05,
+      "loss": 0.6069,
+      "step": 16780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.432235998289867e-05,
+      "loss": 0.55,
+      "step": 16790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.430098332620779e-05,
+      "loss": 0.5481,
+      "step": 16800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.42796066695169e-05,
+      "loss": 0.6097,
+      "step": 16810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.4258230012826e-05,
+      "loss": 0.5391,
+      "step": 16820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.42368533561351e-05,
+      "loss": 0.5986,
+      "step": 16830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.421547669944421e-05,
+      "loss": 0.5548,
+      "step": 16840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.419410004275331e-05,
+      "loss": 0.5509,
+      "step": 16850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.417272338606243e-05,
+      "loss": 0.6078,
+      "step": 16860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.415134672937154e-05,
+      "loss": 0.5482,
+      "step": 16870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.412997007268064e-05,
+      "loss": 0.5988,
+      "step": 16880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.410859341598973e-05,
+      "loss": 0.5504,
+      "step": 16890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.408721675929885e-05,
+      "loss": 0.5439,
+      "step": 16900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.406584010260795e-05,
+      "loss": 0.6126,
+      "step": 16910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.404446344591706e-05,
+      "loss": 0.5525,
+      "step": 16920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.402308678922618e-05,
+      "loss": 0.6029,
+      "step": 16930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.400171013253528e-05,
+      "loss": 0.5571,
+      "step": 16940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.398033347584439e-05,
+      "loss": 0.5438,
+      "step": 16950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.395895681915349e-05,
+      "loss": 0.6036,
+      "step": 16960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.39375801624626e-05,
+      "loss": 0.5452,
+      "step": 16970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.39162035057717e-05,
+      "loss": 0.5876,
+      "step": 16980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.38948268490808e-05,
+      "loss": 0.5511,
+      "step": 16990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.387345019238992e-05,
+      "loss": 0.5408,
+      "step": 17000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.385207353569903e-05,
+      "loss": 0.6054,
+      "step": 17010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.383069687900812e-05,
+      "loss": 0.5533,
+      "step": 17020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.380932022231723e-05,
+      "loss": 0.605,
+      "step": 17030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.378794356562634e-05,
+      "loss": 0.5545,
+      "step": 17040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.376656690893544e-05,
+      "loss": 0.5409,
+      "step": 17050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.374519025224455e-05,
+      "loss": 0.6025,
+      "step": 17060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.372381359555367e-05,
+      "loss": 0.539,
+      "step": 17070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.370243693886276e-05,
+      "loss": 0.5994,
+      "step": 17080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.368106028217186e-05,
+      "loss": 0.5615,
+      "step": 17090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.365968362548098e-05,
+      "loss": 0.5353,
+      "step": 17100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.363830696879008e-05,
+      "loss": 0.621,
+      "step": 17110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.361693031209919e-05,
+      "loss": 0.5477,
+      "step": 17120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.359555365540831e-05,
+      "loss": 0.6055,
+      "step": 17130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.357417699871741e-05,
+      "loss": 0.5412,
+      "step": 17140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.35528003420265e-05,
+      "loss": 0.5579,
+      "step": 17150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.353142368533561e-05,
+      "loss": 0.6094,
+      "step": 17160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.351004702864473e-05,
+      "loss": 0.5529,
+      "step": 17170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.348867037195383e-05,
+      "loss": 0.5846,
+      "step": 17180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.346729371526293e-05,
+      "loss": 0.5621,
+      "step": 17190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.344591705857205e-05,
+      "loss": 0.5463,
+      "step": 17200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.342454040188114e-05,
+      "loss": 0.61,
+      "step": 17210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.340316374519025e-05,
+      "loss": 0.5293,
+      "step": 17220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.338178708849937e-05,
+      "loss": 0.5979,
+      "step": 17230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.336041043180847e-05,
+      "loss": 0.5604,
+      "step": 17240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.333903377511757e-05,
+      "loss": 0.5422,
+      "step": 17250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.331765711842668e-05,
+      "loss": 0.604,
+      "step": 17260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.32962804617358e-05,
+      "loss": 0.5511,
+      "step": 17270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.327490380504489e-05,
+      "loss": 0.598,
+      "step": 17280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.325352714835399e-05,
+      "loss": 0.5573,
+      "step": 17290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.323215049166311e-05,
+      "loss": 0.5378,
+      "step": 17300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.321077383497222e-05,
+      "loss": 0.6017,
+      "step": 17310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.318939717828132e-05,
+      "loss": 0.5463,
+      "step": 17320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.316802052159042e-05,
+      "loss": 0.5951,
+      "step": 17330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.314664386489953e-05,
+      "loss": 0.5614,
+      "step": 17340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.312526720820863e-05,
+      "loss": 0.5425,
+      "step": 17350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.310389055151774e-05,
+      "loss": 0.6129,
+      "step": 17360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.308251389482686e-05,
+      "loss": 0.5388,
+      "step": 17370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.306113723813596e-05,
+      "loss": 0.5997,
+      "step": 17380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.303976058144506e-05,
+      "loss": 0.5555,
+      "step": 17390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.301838392475418e-05,
+      "loss": 0.5479,
+      "step": 17400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.299700726806327e-05,
+      "loss": 0.6108,
+      "step": 17410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.297563061137238e-05,
+      "loss": 0.5383,
+      "step": 17420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.295425395468148e-05,
+      "loss": 0.5983,
+      "step": 17430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.29328772979906e-05,
+      "loss": 0.5641,
+      "step": 17440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.29115006412997e-05,
+      "loss": 0.5459,
+      "step": 17450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.289012398460881e-05,
+      "loss": 0.6169,
+      "step": 17460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.286874732791791e-05,
+      "loss": 0.5524,
+      "step": 17470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.284737067122702e-05,
+      "loss": 0.6054,
+      "step": 17480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.282599401453612e-05,
+      "loss": 0.5504,
+      "step": 17490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.280461735784524e-05,
+      "loss": 0.5439,
+      "step": 17500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.278324070115435e-05,
+      "loss": 0.6144,
+      "step": 17510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.276186404446345e-05,
+      "loss": 0.5403,
+      "step": 17520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.274048738777256e-05,
+      "loss": 0.5901,
+      "step": 17530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.271911073108166e-05,
+      "loss": 0.5567,
+      "step": 17540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.269773407439076e-05,
+      "loss": 0.5438,
+      "step": 17550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.267635741769987e-05,
+      "loss": 0.6143,
+      "step": 17560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.265498076100899e-05,
+      "loss": 0.5406,
+      "step": 17570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.263360410431809e-05,
+      "loss": 0.6106,
+      "step": 17580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.26122274476272e-05,
+      "loss": 0.5512,
+      "step": 17590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.25908507909363e-05,
+      "loss": 0.5435,
+      "step": 17600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.25694741342454e-05,
+      "loss": 0.6053,
+      "step": 17610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.254809747755451e-05,
+      "loss": 0.5473,
+      "step": 17620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.252672082086361e-05,
+      "loss": 0.6014,
+      "step": 17630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.250534416417273e-05,
+      "loss": 0.5564,
+      "step": 17640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.248396750748184e-05,
+      "loss": 0.5487,
+      "step": 17650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.246259085079094e-05,
+      "loss": 0.6089,
+      "step": 17660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.244121419410005e-05,
+      "loss": 0.5461,
+      "step": 17670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.241983753740915e-05,
+      "loss": 0.5984,
+      "step": 17680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.239846088071825e-05,
+      "loss": 0.5527,
+      "step": 17690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.237708422402736e-05,
+      "loss": 0.5457,
+      "step": 17700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.235570756733648e-05,
+      "loss": 0.6033,
+      "step": 17710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.233433091064558e-05,
+      "loss": 0.5414,
+      "step": 17720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.231295425395469e-05,
+      "loss": 0.5885,
+      "step": 17730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.229157759726379e-05,
+      "loss": 0.5565,
+      "step": 17740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.22702009405729e-05,
+      "loss": 0.5529,
+      "step": 17750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.2248824283882e-05,
+      "loss": 0.6132,
+      "step": 17760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.222744762719112e-05,
+      "loss": 0.5507,
+      "step": 17770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.220607097050022e-05,
+      "loss": 0.5943,
+      "step": 17780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.218469431380933e-05,
+      "loss": 0.5488,
+      "step": 17790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.216331765711843e-05,
+      "loss": 0.5418,
+      "step": 17800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.214194100042754e-05,
+      "loss": 0.6067,
+      "step": 17810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.212056434373664e-05,
+      "loss": 0.5479,
+      "step": 17820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.209918768704574e-05,
+      "loss": 0.594,
+      "step": 17830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.207781103035486e-05,
+      "loss": 0.5497,
+      "step": 17840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.205643437366397e-05,
+      "loss": 0.5378,
+      "step": 17850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.203505771697307e-05,
+      "loss": 0.6112,
+      "step": 17860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.201368106028216e-05,
+      "loss": 0.5372,
+      "step": 17870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.199230440359128e-05,
+      "loss": 0.6025,
+      "step": 17880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.197092774690039e-05,
+      "loss": 0.5464,
+      "step": 17890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.194955109020949e-05,
+      "loss": 0.5466,
+      "step": 17900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.192817443351861e-05,
+      "loss": 0.6114,
+      "step": 17910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.190679777682771e-05,
+      "loss": 0.5411,
+      "step": 17920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.188542112013682e-05,
+      "loss": 0.5994,
+      "step": 17930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.186404446344592e-05,
+      "loss": 0.5553,
+      "step": 17940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.184266780675503e-05,
+      "loss": 0.5492,
+      "step": 17950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.182129115006413e-05,
+      "loss": 0.6096,
+      "step": 17960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.179991449337324e-05,
+      "loss": 0.5391,
+      "step": 17970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.177853783668235e-05,
+      "loss": 0.6039,
+      "step": 17980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.175716117999146e-05,
+      "loss": 0.5528,
+      "step": 17990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.173578452330055e-05,
+      "loss": 0.5364,
+      "step": 18000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.171440786660967e-05,
+      "loss": 0.6045,
+      "step": 18010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.169303120991877e-05,
+      "loss": 0.5494,
+      "step": 18020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.167165455322788e-05,
+      "loss": 0.5933,
+      "step": 18030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.1650277896537e-05,
+      "loss": 0.5572,
+      "step": 18040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.16289012398461e-05,
+      "loss": 0.5318,
+      "step": 18050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.16075245831552e-05,
+      "loss": 0.6114,
+      "step": 18060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.15861479264643e-05,
+      "loss": 0.5443,
+      "step": 18070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.156477126977341e-05,
+      "loss": 0.5965,
+      "step": 18080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.154339461308252e-05,
+      "loss": 0.5559,
+      "step": 18090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.152201795639162e-05,
+      "loss": 0.5402,
+      "step": 18100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.150064129970074e-05,
+      "loss": 0.5973,
+      "step": 18110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.147926464300984e-05,
+      "loss": 0.5467,
+      "step": 18120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.145788798631893e-05,
+      "loss": 0.6031,
+      "step": 18130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.143651132962804e-05,
+      "loss": 0.5514,
+      "step": 18140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.141513467293716e-05,
+      "loss": 0.5423,
+      "step": 18150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.139375801624626e-05,
+      "loss": 0.6089,
+      "step": 18160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.137238135955537e-05,
+      "loss": 0.5418,
+      "step": 18170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.135100470286448e-05,
+      "loss": 0.5969,
+      "step": 18180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.132962804617359e-05,
+      "loss": 0.5611,
+      "step": 18190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.130825138948268e-05,
+      "loss": 0.5402,
+      "step": 18200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.12868747327918e-05,
+      "loss": 0.5977,
+      "step": 18210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.12654980761009e-05,
+      "loss": 0.5402,
+      "step": 18220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.124412141941e-05,
+      "loss": 0.6007,
+      "step": 18230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.122274476271911e-05,
+      "loss": 0.553,
+      "step": 18240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.120136810602823e-05,
+      "loss": 0.5448,
+      "step": 18250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.117999144933732e-05,
+      "loss": 0.6071,
+      "step": 18260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.115861479264642e-05,
+      "loss": 0.5442,
+      "step": 18270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.113723813595554e-05,
+      "loss": 0.601,
+      "step": 18280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.111586147926465e-05,
+      "loss": 0.5626,
+      "step": 18290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.109448482257375e-05,
+      "loss": 0.5438,
+      "step": 18300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.107310816588287e-05,
+      "loss": 0.6059,
+      "step": 18310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.105173150919196e-05,
+      "loss": 0.5412,
+      "step": 18320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.103035485250107e-05,
+      "loss": 0.598,
+      "step": 18330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.1008978195810176e-05,
+      "loss": 0.5517,
+      "step": 18340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.098760153911929e-05,
+      "loss": 0.5358,
+      "step": 18350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.096622488242839e-05,
+      "loss": 0.6126,
+      "step": 18360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.09448482257375e-05,
+      "loss": 0.5423,
+      "step": 18370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.092347156904661e-05,
+      "loss": 0.5889,
+      "step": 18380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.090209491235571e-05,
+      "loss": 0.5559,
+      "step": 18390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.088071825566482e-05,
+      "loss": 0.5475,
+      "step": 18400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0859341598973915e-05,
+      "loss": 0.5988,
+      "step": 18410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.083796494228303e-05,
+      "loss": 0.5447,
+      "step": 18420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.081658828559214e-05,
+      "loss": 0.5964,
+      "step": 18430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.079521162890124e-05,
+      "loss": 0.5417,
+      "step": 18440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.077383497221035e-05,
+      "loss": 0.5367,
+      "step": 18450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.075245831551946e-05,
+      "loss": 0.6078,
+      "step": 18460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.073108165882856e-05,
+      "loss": 0.536,
+      "step": 18470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0709705002137673e-05,
+      "loss": 0.5941,
+      "step": 18480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.068832834544678e-05,
+      "loss": 0.558,
+      "step": 18490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.066695168875588e-05,
+      "loss": 0.5506,
+      "step": 18500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.064557503206498e-05,
+      "loss": 0.612,
+      "step": 18510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.06241983753741e-05,
+      "loss": 0.5463,
+      "step": 18520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.06028217186832e-05,
+      "loss": 0.6018,
+      "step": 18530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.05814450619923e-05,
+      "loss": 0.5613,
+      "step": 18540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.056006840530142e-05,
+      "loss": 0.5428,
+      "step": 18550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.053869174861052e-05,
+      "loss": 0.6173,
+      "step": 18560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.051731509191962e-05,
+      "loss": 0.5419,
+      "step": 18570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.049593843522874e-05,
+      "loss": 0.5883,
+      "step": 18580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0474561778537843e-05,
+      "loss": 0.5496,
+      "step": 18590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.045318512184695e-05,
+      "loss": 0.5453,
+      "step": 18600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0431808465156046e-05,
+      "loss": 0.6091,
+      "step": 18610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.0410431808465164e-05,
+      "loss": 0.5405,
+      "step": 18620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.038905515177427e-05,
+      "loss": 0.6,
+      "step": 18630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0367678495083366e-05,
+      "loss": 0.5491,
+      "step": 18640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0346301838392484e-05,
+      "loss": 0.5428,
+      "step": 18650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.032492518170159e-05,
+      "loss": 0.6001,
+      "step": 18660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0303548525010686e-05,
+      "loss": 0.554,
+      "step": 18670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.028217186831979e-05,
+      "loss": 0.5888,
+      "step": 18680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.026079521162891e-05,
+      "loss": 0.551,
+      "step": 18690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0239418554938007e-05,
+      "loss": 0.5353,
+      "step": 18700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.021804189824711e-05,
+      "loss": 0.6005,
+      "step": 18710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.019666524155623e-05,
+      "loss": 0.5479,
+      "step": 18720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.017528858486533e-05,
+      "loss": 0.5869,
+      "step": 18730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.015391192817443e-05,
+      "loss": 0.5578,
+      "step": 18740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.013253527148355e-05,
+      "loss": 0.5463,
+      "step": 18750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0111158614792654e-05,
+      "loss": 0.6061,
+      "step": 18760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.008978195810175e-05,
+      "loss": 0.542,
+      "step": 18770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0068405301410856e-05,
+      "loss": 0.6112,
+      "step": 18780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0047028644719974e-05,
+      "loss": 0.5618,
+      "step": 18790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.002565198802907e-05,
+      "loss": 0.5326,
+      "step": 18800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.0004275331338177e-05,
+      "loss": 0.6123,
+      "step": 18810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.9982898674647295e-05,
+      "loss": 0.5456,
+      "step": 18820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.996152201795639e-05,
+      "loss": 0.594,
+      "step": 18830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.99401453612655e-05,
+      "loss": 0.5503,
+      "step": 18840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.9918768704574615e-05,
+      "loss": 0.5468,
+      "step": 18850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.989739204788371e-05,
+      "loss": 0.6077,
+      "step": 18860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.987601539119282e-05,
+      "loss": 0.5418,
+      "step": 18870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.985463873450192e-05,
+      "loss": 0.5957,
+      "step": 18880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.983326207781104e-05,
+      "loss": 0.5498,
+      "step": 18890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.981188542112014e-05,
+      "loss": 0.5389,
+      "step": 18900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.979050876442924e-05,
+      "loss": 0.6052,
+      "step": 18910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.976913210773836e-05,
+      "loss": 0.5395,
+      "step": 18920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.974775545104746e-05,
+      "loss": 0.5939,
+      "step": 18930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.972637879435656e-05,
+      "loss": 0.554,
+      "step": 18940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.970500213766567e-05,
+      "loss": 0.5297,
+      "step": 18950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.968362548097478e-05,
+      "loss": 0.5997,
+      "step": 18960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.966224882428388e-05,
+      "loss": 0.5491,
+      "step": 18970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.964087216759299e-05,
+      "loss": 0.6009,
+      "step": 18980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.96194955109021e-05,
+      "loss": 0.5508,
+      "step": 18990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.95981188542112e-05,
+      "loss": 0.5384,
+      "step": 19000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.957674219752031e-05,
+      "loss": 0.6029,
+      "step": 19010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.955536554082942e-05,
+      "loss": 0.5443,
+      "step": 19020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.953398888413852e-05,
+      "loss": 0.5974,
+      "step": 19030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.951261222744763e-05,
+      "loss": 0.5557,
+      "step": 19040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.949123557075673e-05,
+      "loss": 0.5353,
+      "step": 19050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.9469858914065844e-05,
+      "loss": 0.5967,
+      "step": 19060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.944848225737495e-05,
+      "loss": 0.5462,
+      "step": 19070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.942710560068405e-05,
+      "loss": 0.5933,
+      "step": 19080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.9405728943993164e-05,
+      "loss": 0.55,
+      "step": 19090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 5.938435228730227e-05,
+      "loss": 0.5356,
+      "step": 19100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.936297563061137e-05,
+      "loss": 0.6037,
+      "step": 19110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.9341598973920484e-05,
+      "loss": 0.5441,
+      "step": 19120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.932022231722959e-05,
+      "loss": 0.6127,
+      "step": 19130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.929884566053869e-05,
+      "loss": 0.5548,
+      "step": 19140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.92774690038478e-05,
+      "loss": 0.5314,
+      "step": 19150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.925609234715691e-05,
+      "loss": 0.5982,
+      "step": 19160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.9234715690466013e-05,
+      "loss": 0.5398,
+      "step": 19170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.921333903377512e-05,
+      "loss": 0.592,
+      "step": 19180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.919196237708423e-05,
+      "loss": 0.5492,
+      "step": 19190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.9170585720393334e-05,
+      "loss": 0.538,
+      "step": 19200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.914920906370244e-05,
+      "loss": 0.6043,
+      "step": 19210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.912783240701154e-05,
+      "loss": 0.5316,
+      "step": 19220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.9106455750320654e-05,
+      "loss": 0.5946,
+      "step": 19230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.908507909362976e-05,
+      "loss": 0.5544,
+      "step": 19240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.906370243693886e-05,
+      "loss": 0.546,
+      "step": 19250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.9042325780247974e-05,
+      "loss": 0.6037,
+      "step": 19260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.902094912355708e-05,
+      "loss": 0.5547,
+      "step": 19270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.8999572466866183e-05,
+      "loss": 0.6003,
+      "step": 19280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.8978195810175295e-05,
+      "loss": 0.5537,
+      "step": 19290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.89568191534844e-05,
+      "loss": 0.528,
+      "step": 19300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.8935442496793504e-05,
+      "loss": 0.6007,
+      "step": 19310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.891406584010261e-05,
+      "loss": 0.5431,
+      "step": 19320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.889268918341172e-05,
+      "loss": 0.5949,
+      "step": 19330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8871312526720824e-05,
+      "loss": 0.5474,
+      "step": 19340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.884993587002993e-05,
+      "loss": 0.5324,
+      "step": 19350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.882855921333904e-05,
+      "loss": 0.6074,
+      "step": 19360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8807182556648144e-05,
+      "loss": 0.5434,
+      "step": 19370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.878580589995725e-05,
+      "loss": 0.5858,
+      "step": 19380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.876442924326636e-05,
+      "loss": 0.5517,
+      "step": 19390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8743052586575465e-05,
+      "loss": 0.5515,
+      "step": 19400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.872167592988457e-05,
+      "loss": 0.6073,
+      "step": 19410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8700299273193674e-05,
+      "loss": 0.5404,
+      "step": 19420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8678922616502785e-05,
+      "loss": 0.5955,
+      "step": 19430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.865754595981189e-05,
+      "loss": 0.5563,
+      "step": 19440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8636169303120994e-05,
+      "loss": 0.5411,
+      "step": 19450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8614792646430105e-05,
+      "loss": 0.605,
+      "step": 19460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.859341598973921e-05,
+      "loss": 0.5447,
+      "step": 19470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8572039333048314e-05,
+      "loss": 0.5972,
+      "step": 19480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.855066267635741e-05,
+      "loss": 0.5486,
+      "step": 19490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.852928601966653e-05,
+      "loss": 0.5396,
+      "step": 19500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8507909362975635e-05,
+      "loss": 0.5956,
+      "step": 19510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.848653270628473e-05,
+      "loss": 0.5347,
+      "step": 19520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.846515604959385e-05,
+      "loss": 0.6002,
+      "step": 19530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8443779392902955e-05,
+      "loss": 0.5509,
+      "step": 19540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.842240273621206e-05,
+      "loss": 0.5313,
+      "step": 19550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.840102607952117e-05,
+      "loss": 0.6076,
+      "step": 19560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 5.8379649422830275e-05,
+      "loss": 0.5359,
+      "step": 19570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.835827276613938e-05,
+      "loss": 0.5892,
+      "step": 19580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.833689610944848e-05,
+      "loss": 0.5579,
+      "step": 19590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8315519452757596e-05,
+      "loss": 0.539,
+      "step": 19600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.82941427960667e-05,
+      "loss": 0.6003,
+      "step": 19610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.82727661393758e-05,
+      "loss": 0.5345,
+      "step": 19620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8251389482684916e-05,
+      "loss": 0.5962,
+      "step": 19630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.823001282599402e-05,
+      "loss": 0.5534,
+      "step": 19640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.820863616930312e-05,
+      "loss": 0.5389,
+      "step": 19650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8187259512612236e-05,
+      "loss": 0.6036,
+      "step": 19660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.816588285592134e-05,
+      "loss": 0.5484,
+      "step": 19670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8144506199230445e-05,
+      "loss": 0.5932,
+      "step": 19680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.812312954253954e-05,
+      "loss": 0.5513,
+      "step": 19690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.810175288584866e-05,
+      "loss": 0.5332,
+      "step": 19700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8080376229157765e-05,
+      "loss": 0.5999,
+      "step": 19710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.805899957246686e-05,
+      "loss": 0.5469,
+      "step": 19720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.803762291577598e-05,
+      "loss": 0.5805,
+      "step": 19730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.8016246259085086e-05,
+      "loss": 0.5482,
+      "step": 19740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.7994869602394184e-05,
+      "loss": 0.5346,
+      "step": 19750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.797349294570329e-05,
+      "loss": 0.6019,
+      "step": 19760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.7952116289012406e-05,
+      "loss": 0.5406,
+      "step": 19770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.7930739632321504e-05,
+      "loss": 0.5919,
+      "step": 19780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.790936297563061e-05,
+      "loss": 0.5551,
+      "step": 19790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 5.7887986318939726e-05,
+      "loss": 0.539,
+      "step": 19800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7866609662248824e-05,
+      "loss": 0.6048,
+      "step": 19810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.784523300555793e-05,
+      "loss": 0.5319,
+      "step": 19820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.782385634886705e-05,
+      "loss": 0.5966,
+      "step": 19830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.780247969217615e-05,
+      "loss": 0.5552,
+      "step": 19840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.778110303548525e-05,
+      "loss": 0.5344,
+      "step": 19850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7759726378794353e-05,
+      "loss": 0.6041,
+      "step": 19860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.773834972210347e-05,
+      "loss": 0.5436,
+      "step": 19870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.771697306541257e-05,
+      "loss": 0.591,
+      "step": 19880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7695596408721674e-05,
+      "loss": 0.5445,
+      "step": 19890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.767421975203079e-05,
+      "loss": 0.5381,
+      "step": 19900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.765284309533989e-05,
+      "loss": 0.6059,
+      "step": 19910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7631466438648994e-05,
+      "loss": 0.535,
+      "step": 19920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.761008978195811e-05,
+      "loss": 0.5897,
+      "step": 19930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.758871312526721e-05,
+      "loss": 0.5422,
+      "step": 19940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7567336468576314e-05,
+      "loss": 0.537,
+      "step": 19950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.754595981188542e-05,
+      "loss": 0.5993,
+      "step": 19960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.752458315519453e-05,
+      "loss": 0.5404,
+      "step": 19970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7503206498503635e-05,
+      "loss": 0.5844,
+      "step": 19980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.748182984181274e-05,
+      "loss": 0.5508,
+      "step": 19990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.746045318512186e-05,
+      "loss": 0.5198,
+      "step": 20000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7439076528430955e-05,
+      "loss": 0.6004,
+      "step": 20010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.741769987174006e-05,
+      "loss": 0.5405,
+      "step": 20020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7396323215049164e-05,
+      "loss": 0.5959,
+      "step": 20030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 5.7374946558358275e-05,
+      "loss": 0.5421,
+      "step": 20040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.735356990166738e-05,
+      "loss": 0.5265,
+      "step": 20050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7332193244976484e-05,
+      "loss": 0.6,
+      "step": 20060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7310816588285596e-05,
+      "loss": 0.5298,
+      "step": 20070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.72894399315947e-05,
+      "loss": 0.5941,
+      "step": 20080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7268063274903805e-05,
+      "loss": 0.5567,
+      "step": 20090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7246686618212916e-05,
+      "loss": 0.5233,
+      "step": 20100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.722530996152202e-05,
+      "loss": 0.6093,
+      "step": 20110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7203933304831125e-05,
+      "loss": 0.5326,
+      "step": 20120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.718255664814023e-05,
+      "loss": 0.6004,
+      "step": 20130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.716117999144934e-05,
+      "loss": 0.5669,
+      "step": 20140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7139803334758445e-05,
+      "loss": 0.5367,
+      "step": 20150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.711842667806755e-05,
+      "loss": 0.596,
+      "step": 20160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.709705002137666e-05,
+      "loss": 0.5525,
+      "step": 20170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7075673364685766e-05,
+      "loss": 0.5978,
+      "step": 20180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.705429670799487e-05,
+      "loss": 0.5462,
+      "step": 20190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.703292005130398e-05,
+      "loss": 0.54,
+      "step": 20200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.7011543394613086e-05,
+      "loss": 0.5983,
+      "step": 20210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.699016673792219e-05,
+      "loss": 0.5388,
+      "step": 20220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.6968790081231295e-05,
+      "loss": 0.597,
+      "step": 20230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.6947413424540406e-05,
+      "loss": 0.5617,
+      "step": 20240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.692603676784951e-05,
+      "loss": 0.5306,
+      "step": 20250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.6904660111158615e-05,
+      "loss": 0.601,
+      "step": 20260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 5.6883283454467727e-05,
+      "loss": 0.54,
+      "step": 20270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.686190679777683e-05,
+      "loss": 0.5905,
+      "step": 20280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6840530141085936e-05,
+      "loss": 0.5599,
+      "step": 20290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.681915348439504e-05,
+      "loss": 0.5438,
+      "step": 20300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.679777682770415e-05,
+      "loss": 0.6022,
+      "step": 20310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6776400171013256e-05,
+      "loss": 0.5244,
+      "step": 20320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.675502351432236e-05,
+      "loss": 0.5905,
+      "step": 20330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.673364685763147e-05,
+      "loss": 0.5503,
+      "step": 20340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6712270200940576e-05,
+      "loss": 0.5322,
+      "step": 20350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.669089354424968e-05,
+      "loss": 0.6064,
+      "step": 20360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.666951688755879e-05,
+      "loss": 0.5453,
+      "step": 20370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6648140230867896e-05,
+      "loss": 0.593,
+      "step": 20380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6626763574177e-05,
+      "loss": 0.5486,
+      "step": 20390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6605386917486105e-05,
+      "loss": 0.5426,
+      "step": 20400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.658401026079522e-05,
+      "loss": 0.6055,
+      "step": 20410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.656263360410432e-05,
+      "loss": 0.541,
+      "step": 20420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6541256947413426e-05,
+      "loss": 0.5992,
+      "step": 20430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.651988029072254e-05,
+      "loss": 0.5508,
+      "step": 20440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.649850363403164e-05,
+      "loss": 0.5352,
+      "step": 20450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6477126977340746e-05,
+      "loss": 0.5986,
+      "step": 20460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.645575032064986e-05,
+      "loss": 0.5446,
+      "step": 20470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.643437366395896e-05,
+      "loss": 0.5984,
+      "step": 20480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.6412997007268066e-05,
+      "loss": 0.5426,
+      "step": 20490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.639162035057717e-05,
+      "loss": 0.5391,
+      "step": 20500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 5.637024369388628e-05,
+      "loss": 0.5997,
+      "step": 20510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.634886703719539e-05,
+      "loss": 0.5297,
+      "step": 20520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.632749038050449e-05,
+      "loss": 0.5924,
+      "step": 20530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.63061137238136e-05,
+      "loss": 0.5502,
+      "step": 20540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.628473706712271e-05,
+      "loss": 0.5297,
+      "step": 20550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.626336041043181e-05,
+      "loss": 0.5915,
+      "step": 20560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.624198375374091e-05,
+      "loss": 0.5363,
+      "step": 20570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.622060709705003e-05,
+      "loss": 0.5982,
+      "step": 20580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.619923044035913e-05,
+      "loss": 0.5542,
+      "step": 20590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.617785378366823e-05,
+      "loss": 0.5359,
+      "step": 20600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.615647712697735e-05,
+      "loss": 0.5999,
+      "step": 20610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.613510047028645e-05,
+      "loss": 0.5455,
+      "step": 20620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.611372381359556e-05,
+      "loss": 0.6033,
+      "step": 20630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.609234715690467e-05,
+      "loss": 0.5505,
+      "step": 20640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.607097050021377e-05,
+      "loss": 0.5404,
+      "step": 20650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.604959384352288e-05,
+      "loss": 0.6059,
+      "step": 20660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.6028217186831975e-05,
+      "loss": 0.532,
+      "step": 20670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.600684053014109e-05,
+      "loss": 0.5928,
+      "step": 20680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.59854638734502e-05,
+      "loss": 0.5459,
+      "step": 20690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.5964087216759295e-05,
+      "loss": 0.5339,
+      "step": 20700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.594271056006841e-05,
+      "loss": 0.5905,
+      "step": 20710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.592133390337752e-05,
+      "loss": 0.5389,
+      "step": 20720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.5899957246686615e-05,
+      "loss": 0.5874,
+      "step": 20730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 5.587858058999573e-05,
+      "loss": 0.5474,
+      "step": 20740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.585720393330484e-05,
+      "loss": 0.5267,
+      "step": 20750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5835827276613936e-05,
+      "loss": 0.606,
+      "step": 20760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.581445061992304e-05,
+      "loss": 0.5355,
+      "step": 20770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.579307396323216e-05,
+      "loss": 0.5976,
+      "step": 20780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.577169730654126e-05,
+      "loss": 0.5564,
+      "step": 20790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.575032064985036e-05,
+      "loss": 0.5323,
+      "step": 20800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.572894399315948e-05,
+      "loss": 0.6026,
+      "step": 20810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.570756733646858e-05,
+      "loss": 0.5389,
+      "step": 20820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.568619067977768e-05,
+      "loss": 0.5955,
+      "step": 20830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5664814023086785e-05,
+      "loss": 0.546,
+      "step": 20840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.56434373663959e-05,
+      "loss": 0.5335,
+      "step": 20850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5622060709705e-05,
+      "loss": 0.6122,
+      "step": 20860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5600684053014106e-05,
+      "loss": 0.5249,
+      "step": 20870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5579307396323224e-05,
+      "loss": 0.5895,
+      "step": 20880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.555793073963232e-05,
+      "loss": 0.5518,
+      "step": 20890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5536554082941426e-05,
+      "loss": 0.5336,
+      "step": 20900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5515177426250544e-05,
+      "loss": 0.5982,
+      "step": 20910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.549380076955965e-05,
+      "loss": 0.5328,
+      "step": 20920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5472424112868746e-05,
+      "loss": 0.5856,
+      "step": 20930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.545104745617785e-05,
+      "loss": 0.5567,
+      "step": 20940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.542967079948697e-05,
+      "loss": 0.5324,
+      "step": 20950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.5408294142796066e-05,
+      "loss": 0.5969,
+      "step": 20960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.538691748610517e-05,
+      "loss": 0.5389,
+      "step": 20970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.536554082941429e-05,
+      "loss": 0.591,
+      "step": 20980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.534416417272339e-05,
+      "loss": 0.5547,
+      "step": 20990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.532278751603249e-05,
+      "loss": 0.5378,
+      "step": 21000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.530141085934161e-05,
+      "loss": 0.5931,
+      "step": 21010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.528003420265071e-05,
+      "loss": 0.546,
+      "step": 21020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.525865754595981e-05,
+      "loss": 0.5945,
+      "step": 21030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.5237280889268916e-05,
+      "loss": 0.5528,
+      "step": 21040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.521590423257803e-05,
+      "loss": 0.5408,
+      "step": 21050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.519452757588713e-05,
+      "loss": 0.605,
+      "step": 21060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.5173150919196236e-05,
+      "loss": 0.5429,
+      "step": 21070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.5151774262505355e-05,
+      "loss": 0.5868,
+      "step": 21080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.513039760581445e-05,
+      "loss": 0.5473,
+      "step": 21090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.510902094912356e-05,
+      "loss": 0.5264,
+      "step": 21100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.508764429243266e-05,
+      "loss": 0.5953,
+      "step": 21110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.506626763574177e-05,
+      "loss": 0.5401,
+      "step": 21120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.504489097905088e-05,
+      "loss": 0.6003,
+      "step": 21130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.502351432235998e-05,
+      "loss": 0.5497,
+      "step": 21140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.500213766566909e-05,
+      "loss": 0.523,
+      "step": 21150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.49807610089782e-05,
+      "loss": 0.5955,
+      "step": 21160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.49593843522873e-05,
+      "loss": 0.5367,
+      "step": 21170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.493800769559641e-05,
+      "loss": 0.5901,
+      "step": 21180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.491663103890552e-05,
+      "loss": 0.5389,
+      "step": 21190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.489525438221462e-05,
+      "loss": 0.5353,
+      "step": 21200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.487387772552373e-05,
+      "loss": 0.6045,
+      "step": 21210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.485250106883284e-05,
+      "loss": 0.5351,
+      "step": 21220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.483112441214194e-05,
+      "loss": 0.5895,
+      "step": 21230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.480974775545105e-05,
+      "loss": 0.5514,
+      "step": 21240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.478837109876016e-05,
+      "loss": 0.5362,
+      "step": 21250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.476699444206926e-05,
+      "loss": 0.5933,
+      "step": 21260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.474561778537837e-05,
+      "loss": 0.5358,
+      "step": 21270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.472424112868748e-05,
+      "loss": 0.5993,
+      "step": 21280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.470286447199658e-05,
+      "loss": 0.5455,
+      "step": 21290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.468148781530569e-05,
+      "loss": 0.5417,
+      "step": 21300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.466011115861479e-05,
+      "loss": 0.6006,
+      "step": 21310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4638734501923903e-05,
+      "loss": 0.5393,
+      "step": 21320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.461735784523301e-05,
+      "loss": 0.5911,
+      "step": 21330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.459598118854211e-05,
+      "loss": 0.5507,
+      "step": 21340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4574604531851224e-05,
+      "loss": 0.5345,
+      "step": 21350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.455322787516033e-05,
+      "loss": 0.5991,
+      "step": 21360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.453185121846943e-05,
+      "loss": 0.5205,
+      "step": 21370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.451047456177854e-05,
+      "loss": 0.6007,
+      "step": 21380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.448909790508765e-05,
+      "loss": 0.5589,
+      "step": 21390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.446772124839675e-05,
+      "loss": 0.5331,
+      "step": 21400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.444634459170586e-05,
+      "loss": 0.6008,
+      "step": 21410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.442496793501497e-05,
+      "loss": 0.5388,
+      "step": 21420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.440359127832407e-05,
+      "loss": 0.5969,
+      "step": 21430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.438221462163318e-05,
+      "loss": 0.5484,
+      "step": 21440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.436083796494229e-05,
+      "loss": 0.5328,
+      "step": 21450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4339461308251394e-05,
+      "loss": 0.5929,
+      "step": 21460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.43180846515605e-05,
+      "loss": 0.535,
+      "step": 21470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.42967079948696e-05,
+      "loss": 0.5848,
+      "step": 21480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4275331338178714e-05,
+      "loss": 0.5469,
+      "step": 21490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.425395468148782e-05,
+      "loss": 0.5395,
+      "step": 21500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.423257802479692e-05,
+      "loss": 0.5943,
+      "step": 21510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4211201368106034e-05,
+      "loss": 0.5385,
+      "step": 21520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.418982471141514e-05,
+      "loss": 0.5847,
+      "step": 21530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.416844805472424e-05,
+      "loss": 0.5442,
+      "step": 21540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4147071398033355e-05,
+      "loss": 0.5322,
+      "step": 21550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.412569474134246e-05,
+      "loss": 0.6063,
+      "step": 21560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4104318084651564e-05,
+      "loss": 0.5256,
+      "step": 21570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.408294142796067e-05,
+      "loss": 0.5824,
+      "step": 21580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.406156477126978e-05,
+      "loss": 0.549,
+      "step": 21590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.4040188114578884e-05,
+      "loss": 0.5375,
+      "step": 21600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.401881145788799e-05,
+      "loss": 0.6146,
+      "step": 21610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.39974348011971e-05,
+      "loss": 0.5359,
+      "step": 21620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3976058144506204e-05,
+      "loss": 0.5871,
+      "step": 21630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.395468148781531e-05,
+      "loss": 0.552,
+      "step": 21640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3933304831124406e-05,
+      "loss": 0.5365,
+      "step": 21650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3911928174433525e-05,
+      "loss": 0.5981,
+      "step": 21660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.389055151774263e-05,
+      "loss": 0.5403,
+      "step": 21670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.386917486105173e-05,
+      "loss": 0.5914,
+      "step": 21680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3847798204360845e-05,
+      "loss": 0.5402,
+      "step": 21690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.382642154766995e-05,
+      "loss": 0.5311,
+      "step": 21700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3805044890979054e-05,
+      "loss": 0.5993,
+      "step": 21710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3783668234288165e-05,
+      "loss": 0.5373,
+      "step": 21720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.376229157759727e-05,
+      "loss": 0.5993,
+      "step": 21730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3740914920906374e-05,
+      "loss": 0.5491,
+      "step": 21740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.371953826421547e-05,
+      "loss": 0.5345,
+      "step": 21750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.369816160752459e-05,
+      "loss": 0.597,
+      "step": 21760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3676784950833695e-05,
+      "loss": 0.5348,
+      "step": 21770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.365540829414279e-05,
+      "loss": 0.597,
+      "step": 21780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.363403163745191e-05,
+      "loss": 0.5489,
+      "step": 21790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3612654980761015e-05,
+      "loss": 0.536,
+      "step": 21800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.359127832407011e-05,
+      "loss": 0.6037,
+      "step": 21810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.356990166737923e-05,
+      "loss": 0.5439,
+      "step": 21820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3548525010688335e-05,
+      "loss": 0.5853,
+      "step": 21830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.352714835399743e-05,
+      "loss": 0.5537,
+      "step": 21840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.350577169730654e-05,
+      "loss": 0.5325,
+      "step": 21850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3484395040615655e-05,
+      "loss": 0.5928,
+      "step": 21860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.346301838392476e-05,
+      "loss": 0.5373,
+      "step": 21870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.344164172723386e-05,
+      "loss": 0.5879,
+      "step": 21880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.3420265070542976e-05,
+      "loss": 0.5412,
+      "step": 21890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.339888841385208e-05,
+      "loss": 0.5255,
+      "step": 21900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.337751175716118e-05,
+      "loss": 0.5921,
+      "step": 21910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.335613510047028e-05,
+      "loss": 0.5433,
+      "step": 21920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.33347584437794e-05,
+      "loss": 0.5957,
+      "step": 21930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.33133817870885e-05,
+      "loss": 0.5426,
+      "step": 21940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.32920051303976e-05,
+      "loss": 0.5399,
+      "step": 21950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.327062847370672e-05,
+      "loss": 0.5924,
+      "step": 21960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.324925181701582e-05,
+      "loss": 0.5414,
+      "step": 21970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.322787516032492e-05,
+      "loss": 0.5904,
+      "step": 21980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.320649850363404e-05,
+      "loss": 0.5486,
+      "step": 21990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.318512184694314e-05,
+      "loss": 0.5334,
+      "step": 22000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3163745190252243e-05,
+      "loss": 0.6065,
+      "step": 22010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.314236853356135e-05,
+      "loss": 0.5367,
+      "step": 22020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3120991876870466e-05,
+      "loss": 0.5958,
+      "step": 22030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3099615220179564e-05,
+      "loss": 0.5565,
+      "step": 22040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.307823856348867e-05,
+      "loss": 0.5287,
+      "step": 22050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3056861906797786e-05,
+      "loss": 0.6005,
+      "step": 22060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3035485250106884e-05,
+      "loss": 0.5422,
+      "step": 22070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.301410859341599e-05,
+      "loss": 0.5889,
+      "step": 22080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2992731936725107e-05,
+      "loss": 0.5483,
+      "step": 22090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2971355280034204e-05,
+      "loss": 0.5393,
+      "step": 22100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.294997862334331e-05,
+      "loss": 0.594,
+      "step": 22110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.292860196665241e-05,
+      "loss": 0.5376,
+      "step": 22120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2907225309961525e-05,
+      "loss": 0.5964,
+      "step": 22130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.288584865327063e-05,
+      "loss": 0.5449,
+      "step": 22140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.2864471996579734e-05,
+      "loss": 0.5306,
+      "step": 22150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.284309533988885e-05,
+      "loss": 0.6048,
+      "step": 22160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.282171868319795e-05,
+      "loss": 0.5321,
+      "step": 22170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2800342026507054e-05,
+      "loss": 0.5871,
+      "step": 22180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.277896536981616e-05,
+      "loss": 0.5379,
+      "step": 22190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.275758871312527e-05,
+      "loss": 0.5341,
+      "step": 22200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2736212056434374e-05,
+      "loss": 0.596,
+      "step": 22210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.271483539974348e-05,
+      "loss": 0.536,
+      "step": 22220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.269345874305259e-05,
+      "loss": 0.5889,
+      "step": 22230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2672082086361695e-05,
+      "loss": 0.5507,
+      "step": 22240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.26507054296708e-05,
+      "loss": 0.5277,
+      "step": 22250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.262932877297991e-05,
+      "loss": 0.6025,
+      "step": 22260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2607952116289015e-05,
+      "loss": 0.5354,
+      "step": 22270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.258657545959812e-05,
+      "loss": 0.5819,
+      "step": 22280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2565198802907224e-05,
+      "loss": 0.5543,
+      "step": 22290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2543822146216335e-05,
+      "loss": 0.5331,
+      "step": 22300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.252244548952544e-05,
+      "loss": 0.6056,
+      "step": 22310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2501068832834544e-05,
+      "loss": 0.5403,
+      "step": 22320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2479692176143656e-05,
+      "loss": 0.5814,
+      "step": 22330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.245831551945276e-05,
+      "loss": 0.5445,
+      "step": 22340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2436938862761865e-05,
+      "loss": 0.5346,
+      "step": 22350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2415562206070976e-05,
+      "loss": 0.6059,
+      "step": 22360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.239418554938008e-05,
+      "loss": 0.5263,
+      "step": 22370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 5.2372808892689185e-05,
+      "loss": 0.5958,
+      "step": 22380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.235143223599829e-05,
+      "loss": 0.5538,
+      "step": 22390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.23300555793074e-05,
+      "loss": 0.5383,
+      "step": 22400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2308678922616505e-05,
+      "loss": 0.589,
+      "step": 22410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.228730226592561e-05,
+      "loss": 0.5347,
+      "step": 22420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.226592560923472e-05,
+      "loss": 0.5917,
+      "step": 22430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2244548952543825e-05,
+      "loss": 0.5486,
+      "step": 22440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.222317229585293e-05,
+      "loss": 0.5283,
+      "step": 22450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2201795639162034e-05,
+      "loss": 0.6053,
+      "step": 22460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2180418982471146e-05,
+      "loss": 0.5451,
+      "step": 22470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.215904232578025e-05,
+      "loss": 0.5886,
+      "step": 22480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2137665669089355e-05,
+      "loss": 0.5372,
+      "step": 22490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2116289012398466e-05,
+      "loss": 0.5437,
+      "step": 22500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.209491235570757e-05,
+      "loss": 0.6081,
+      "step": 22510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2073535699016675e-05,
+      "loss": 0.5416,
+      "step": 22520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2052159042325786e-05,
+      "loss": 0.5899,
+      "step": 22530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.203078238563489e-05,
+      "loss": 0.5455,
+      "step": 22540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.2009405728943995e-05,
+      "loss": 0.5375,
+      "step": 22550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.19880290722531e-05,
+      "loss": 0.5954,
+      "step": 22560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.196665241556221e-05,
+      "loss": 0.535,
+      "step": 22570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1945275758871316e-05,
+      "loss": 0.603,
+      "step": 22580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.192389910218042e-05,
+      "loss": 0.5452,
+      "step": 22590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.190252244548953e-05,
+      "loss": 0.5336,
+      "step": 22600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 5.1881145788798636e-05,
+      "loss": 0.6026,
+      "step": 22610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.185976913210774e-05,
+      "loss": 0.5373,
+      "step": 22620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.183839247541685e-05,
+      "loss": 0.5843,
+      "step": 22630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1817015818725956e-05,
+      "loss": 0.5488,
+      "step": 22640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.179563916203506e-05,
+      "loss": 0.5317,
+      "step": 22650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1774262505344165e-05,
+      "loss": 0.5959,
+      "step": 22660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.175288584865328e-05,
+      "loss": 0.5265,
+      "step": 22670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.173150919196238e-05,
+      "loss": 0.5899,
+      "step": 22680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1710132535271486e-05,
+      "loss": 0.5491,
+      "step": 22690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.16887558785806e-05,
+      "loss": 0.5309,
+      "step": 22700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.16673792218897e-05,
+      "loss": 0.5978,
+      "step": 22710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1646002565198806e-05,
+      "loss": 0.5329,
+      "step": 22720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1624625908507904e-05,
+      "loss": 0.5925,
+      "step": 22730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.160324925181702e-05,
+      "loss": 0.5447,
+      "step": 22740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1581872595126126e-05,
+      "loss": 0.5336,
+      "step": 22750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1560495938435224e-05,
+      "loss": 0.5929,
+      "step": 22760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.153911928174434e-05,
+      "loss": 0.542,
+      "step": 22770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1517742625053447e-05,
+      "loss": 0.5914,
+      "step": 22780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1496365968362544e-05,
+      "loss": 0.5471,
+      "step": 22790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.147498931167166e-05,
+      "loss": 0.5384,
+      "step": 22800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.145361265498077e-05,
+      "loss": 0.5976,
+      "step": 22810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.143223599828987e-05,
+      "loss": 0.5349,
+      "step": 22820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.141085934159897e-05,
+      "loss": 0.5939,
+      "step": 22830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.138948268490809e-05,
+      "loss": 0.5451,
+      "step": 22840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.136810602821719e-05,
+      "loss": 0.529,
+      "step": 22850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.134672937152629e-05,
+      "loss": 0.5986,
+      "step": 22860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.132535271483541e-05,
+      "loss": 0.5293,
+      "step": 22870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.130397605814451e-05,
+      "loss": 0.5804,
+      "step": 22880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.128259940145361e-05,
+      "loss": 0.5472,
+      "step": 22890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.126122274476273e-05,
+      "loss": 0.5312,
+      "step": 22900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.123984608807183e-05,
+      "loss": 0.5972,
+      "step": 22910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.121846943138093e-05,
+      "loss": 0.5313,
+      "step": 22920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.1197092774690035e-05,
+      "loss": 0.5808,
+      "step": 22930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.117571611799915e-05,
+      "loss": 0.5413,
+      "step": 22940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.115433946130826e-05,
+      "loss": 0.5328,
+      "step": 22950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.1132962804617355e-05,
+      "loss": 0.604,
+      "step": 22960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.111158614792647e-05,
+      "loss": 0.5273,
+      "step": 22970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.109020949123558e-05,
+      "loss": 0.5921,
+      "step": 22980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.1068832834544675e-05,
+      "loss": 0.5493,
+      "step": 22990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.104745617785378e-05,
+      "loss": 0.5316,
+      "step": 23000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.10260795211629e-05,
+      "loss": 0.5947,
+      "step": 23010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.1004702864471996e-05,
+      "loss": 0.5234,
+      "step": 23020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.09833262077811e-05,
+      "loss": 0.5956,
+      "step": 23030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.096194955109022e-05,
+      "loss": 0.536,
+      "step": 23040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.0940572894399316e-05,
+      "loss": 0.5252,
+      "step": 23050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.091919623770842e-05,
+      "loss": 0.5896,
+      "step": 23060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.089781958101754e-05,
+      "loss": 0.5313,
+      "step": 23070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.0876442924326636e-05,
+      "loss": 0.5927,
+      "step": 23080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.085506626763574e-05,
+      "loss": 0.5454,
+      "step": 23090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0833689610944845e-05,
+      "loss": 0.5203,
+      "step": 23100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.081231295425396e-05,
+      "loss": 0.5987,
+      "step": 23110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.079093629756306e-05,
+      "loss": 0.5267,
+      "step": 23120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0769559640872165e-05,
+      "loss": 0.5954,
+      "step": 23130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0748182984181284e-05,
+      "loss": 0.5473,
+      "step": 23140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.072680632749038e-05,
+      "loss": 0.5311,
+      "step": 23150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0705429670799486e-05,
+      "loss": 0.6048,
+      "step": 23160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0684053014108604e-05,
+      "loss": 0.5327,
+      "step": 23170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.06626763574177e-05,
+      "loss": 0.588,
+      "step": 23180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0641299700726806e-05,
+      "loss": 0.5481,
+      "step": 23190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.061992304403591e-05,
+      "loss": 0.5284,
+      "step": 23200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.059854638734502e-05,
+      "loss": 0.594,
+      "step": 23210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0577169730654126e-05,
+      "loss": 0.5426,
+      "step": 23220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.055579307396323e-05,
+      "loss": 0.5884,
+      "step": 23230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.053441641727234e-05,
+      "loss": 0.5538,
+      "step": 23240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.051303976058145e-05,
+      "loss": 0.5268,
+      "step": 23250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.049166310389055e-05,
+      "loss": 0.5914,
+      "step": 23260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0470286447199656e-05,
+      "loss": 0.5324,
+      "step": 23270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.044890979050877e-05,
+      "loss": 0.5883,
+      "step": 23280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.042753313381787e-05,
+      "loss": 0.5476,
+      "step": 23290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0406156477126976e-05,
+      "loss": 0.5272,
+      "step": 23300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.038477982043609e-05,
+      "loss": 0.5893,
+      "step": 23310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.036340316374519e-05,
+      "loss": 0.5306,
+      "step": 23320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.0342026507054296e-05,
+      "loss": 0.5917,
+      "step": 23330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.032064985036341e-05,
+      "loss": 0.5453,
+      "step": 23340
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.029927319367251e-05,
+      "loss": 0.5221,
+      "step": 23350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.027789653698162e-05,
+      "loss": 0.5987,
+      "step": 23360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.025651988029072e-05,
+      "loss": 0.5334,
+      "step": 23370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.023514322359983e-05,
+      "loss": 0.586,
+      "step": 23380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.021376656690894e-05,
+      "loss": 0.5444,
+      "step": 23390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.019238991021804e-05,
+      "loss": 0.5238,
+      "step": 23400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.017101325352715e-05,
+      "loss": 0.5938,
+      "step": 23410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.014963659683626e-05,
+      "loss": 0.5414,
+      "step": 23420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.012825994014536e-05,
+      "loss": 0.5831,
+      "step": 23430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.010688328345447e-05,
+      "loss": 0.5288,
+      "step": 23440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.008550662676358e-05,
+      "loss": 0.5849,
+      "step": 23450
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.006412997007268e-05,
+      "loss": 0.5317,
+      "step": 23460
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.0042753313381787e-05,
+      "loss": 0.5916,
+      "step": 23470
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.00213766566909e-05,
+      "loss": 0.549,
+      "step": 23480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5e-05,
+      "loss": 0.5417,
+      "step": 23490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.9978623343309114e-05,
+      "loss": 0.5777,
+      "step": 23500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.995724668661821e-05,
+      "loss": 0.5373,
+      "step": 23510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.993587002992732e-05,
+      "loss": 0.598,
+      "step": 23520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.991449337323643e-05,
+      "loss": 0.5405,
+      "step": 23530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.989311671654553e-05,
+      "loss": 0.5587,
+      "step": 23540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.987174005985464e-05,
+      "loss": 0.5787,
+      "step": 23550
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.985036340316375e-05,
+      "loss": 0.5273,
+      "step": 23560
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.982898674647286e-05,
+      "loss": 0.5845,
+      "step": 23570
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.9807610089781957e-05,
+      "loss": 0.5433,
+      "step": 23580
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.978623343309107e-05,
+      "loss": 0.5581,
+      "step": 23590
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.976485677640018e-05,
+      "loss": 0.5769,
+      "step": 23600
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.974348011970928e-05,
+      "loss": 0.5288,
+      "step": 23610
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.972210346301839e-05,
+      "loss": 0.585,
+      "step": 23620
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.970072680632749e-05,
+      "loss": 0.5348,
+      "step": 23630
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.96793501496366e-05,
+      "loss": 0.5482,
+      "step": 23640
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.96579734929457e-05,
+      "loss": 0.583,
+      "step": 23650
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.963659683625481e-05,
+      "loss": 0.5376,
+      "step": 23660
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.961522017956392e-05,
+      "loss": 0.5879,
+      "step": 23670
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.959384352287302e-05,
+      "loss": 0.532,
+      "step": 23680
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.957246686618213e-05,
+      "loss": 0.5575,
+      "step": 23690
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.955109020949124e-05,
+      "loss": 0.5753,
+      "step": 23700
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.952971355280034e-05,
+      "loss": 0.5337,
+      "step": 23710
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.9508336896109454e-05,
+      "loss": 0.5888,
+      "step": 23720
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.948696023941856e-05,
+      "loss": 0.5491,
+      "step": 23730
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.946558358272766e-05,
+      "loss": 0.5513,
+      "step": 23740
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.944420692603677e-05,
+      "loss": 0.5789,
+      "step": 23750
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.942283026934588e-05,
+      "loss": 0.5258,
+      "step": 23760
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.940145361265498e-05,
+      "loss": 0.5841,
+      "step": 23770
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.938007695596409e-05,
+      "loss": 0.5401,
+      "step": 23780
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.93587002992732e-05,
+      "loss": 0.55,
+      "step": 23790
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.93373236425823e-05,
+      "loss": 0.5778,
+      "step": 23800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.931594698589141e-05,
+      "loss": 0.528,
+      "step": 23810
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.929457032920052e-05,
+      "loss": 0.5883,
+      "step": 23820
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.9273193672509624e-05,
+      "loss": 0.537,
+      "step": 23830
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.925181701581873e-05,
+      "loss": 0.5454,
+      "step": 23840
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.923044035912783e-05,
+      "loss": 0.5789,
+      "step": 23850
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.9209063702436944e-05,
+      "loss": 0.5159,
+      "step": 23860
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.918768704574605e-05,
+      "loss": 0.5968,
+      "step": 23870
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.916631038905515e-05,
+      "loss": 0.5452,
+      "step": 23880
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.9144933732364264e-05,
+      "loss": 0.5529,
+      "step": 23890
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.912355707567337e-05,
+      "loss": 0.58,
+      "step": 23900
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.910218041898247e-05,
+      "loss": 0.5335,
+      "step": 23910
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.908080376229158e-05,
+      "loss": 0.5914,
+      "step": 23920
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.905942710560069e-05,
+      "loss": 0.5343,
+      "step": 23930
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.9038050448909793e-05,
+      "loss": 0.5404,
+      "step": 23940
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.90166737922189e-05,
+      "loss": 0.5808,
+      "step": 23950
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.899529713552801e-05,
+      "loss": 0.5242,
+      "step": 23960
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.897392047883711e-05,
+      "loss": 0.5949,
+      "step": 23970
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.895254382214622e-05,
+      "loss": 0.5383,
+      "step": 23980
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.893116716545533e-05,
+      "loss": 0.5473,
+      "step": 23990
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.890979050876443e-05,
+      "loss": 0.5743,
+      "step": 24000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.888841385207354e-05,
+      "loss": 0.5273,
+      "step": 24010
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.886703719538264e-05,
+      "loss": 0.5894,
+      "step": 24020
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.884566053869175e-05,
+      "loss": 0.5348,
+      "step": 24030
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.882428388200086e-05,
+      "loss": 0.5482,
+      "step": 24040
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8802907225309963e-05,
+      "loss": 0.5778,
+      "step": 24050
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8781530568619075e-05,
+      "loss": 0.5201,
+      "step": 24060
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.876015391192817e-05,
+      "loss": 0.593,
+      "step": 24070
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8738777255237284e-05,
+      "loss": 0.5403,
+      "step": 24080
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8717400598546395e-05,
+      "loss": 0.5416,
+      "step": 24090
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.869602394185549e-05,
+      "loss": 0.5837,
+      "step": 24100
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8674647285164604e-05,
+      "loss": 0.5294,
+      "step": 24110
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.865327062847371e-05,
+      "loss": 0.5952,
+      "step": 24120
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.863189397178281e-05,
+      "loss": 0.5371,
+      "step": 24130
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8610517315091924e-05,
+      "loss": 0.546,
+      "step": 24140
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.858914065840103e-05,
+      "loss": 0.5801,
+      "step": 24150
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.856776400171013e-05,
+      "loss": 0.5211,
+      "step": 24160
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.854638734501924e-05,
+      "loss": 0.5887,
+      "step": 24170
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.852501068832835e-05,
+      "loss": 0.5422,
+      "step": 24180
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8503634031637454e-05,
+      "loss": 0.5523,
+      "step": 24190
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.848225737494656e-05,
+      "loss": 0.5763,
+      "step": 24200
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.846088071825567e-05,
+      "loss": 0.527,
+      "step": 24210
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8439504061564774e-05,
+      "loss": 0.5969,
+      "step": 24220
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.841812740487388e-05,
+      "loss": 0.5448,
+      "step": 24230
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.839675074818298e-05,
+      "loss": 0.548,
+      "step": 24240
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.8375374091492094e-05,
+      "loss": 0.5769,
+      "step": 24250
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.83539974348012e-05,
+      "loss": 0.5333,
+      "step": 24260
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.83326207781103e-05,
+      "loss": 0.5889,
+      "step": 24270
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8311244121419415e-05,
+      "loss": 0.5393,
+      "step": 24280
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.828986746472852e-05,
+      "loss": 0.5529,
+      "step": 24290
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8268490808037624e-05,
+      "loss": 0.5698,
+      "step": 24300
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8247114151346735e-05,
+      "loss": 0.5255,
+      "step": 24310
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.822573749465584e-05,
+      "loss": 0.5868,
+      "step": 24320
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8204360837964944e-05,
+      "loss": 0.5432,
+      "step": 24330
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.818298418127405e-05,
+      "loss": 0.5575,
+      "step": 24340
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.816160752458316e-05,
+      "loss": 0.575,
+      "step": 24350
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8140230867892264e-05,
+      "loss": 0.5231,
+      "step": 24360
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.811885421120137e-05,
+      "loss": 0.5865,
+      "step": 24370
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.809747755451048e-05,
+      "loss": 0.5369,
+      "step": 24380
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8076100897819585e-05,
+      "loss": 0.5546,
+      "step": 24390
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.805472424112869e-05,
+      "loss": 0.5707,
+      "step": 24400
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.80333475844378e-05,
+      "loss": 0.5367,
+      "step": 24410
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.8011970927746905e-05,
+      "loss": 0.5872,
+      "step": 24420
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.799059427105601e-05,
+      "loss": 0.544,
+      "step": 24430
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7969217614365114e-05,
+      "loss": 0.5423,
+      "step": 24440
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7947840957674225e-05,
+      "loss": 0.5802,
+      "step": 24450
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.792646430098332e-05,
+      "loss": 0.5309,
+      "step": 24460
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7905087644292434e-05,
+      "loss": 0.5887,
+      "step": 24470
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7883710987601545e-05,
+      "loss": 0.5405,
+      "step": 24480
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.786233433091064e-05,
+      "loss": 0.5565,
+      "step": 24490
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7840957674219754e-05,
+      "loss": 0.5757,
+      "step": 24500
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.781958101752886e-05,
+      "loss": 0.533,
+      "step": 24510
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.779820436083797e-05,
+      "loss": 0.589,
+      "step": 24520
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7776827704147075e-05,
+      "loss": 0.5461,
+      "step": 24530
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.775545104745618e-05,
+      "loss": 0.5465,
+      "step": 24540
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.773407439076529e-05,
+      "loss": 0.5839,
+      "step": 24550
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.771269773407439e-05,
+      "loss": 0.517,
+      "step": 24560
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.76913210773835e-05,
+      "loss": 0.5962,
+      "step": 24570
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.766994442069261e-05,
+      "loss": 0.5357,
+      "step": 24580
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.764856776400171e-05,
+      "loss": 0.5511,
+      "step": 24590
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.762719110731082e-05,
+      "loss": 0.581,
+      "step": 24600
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7605814450619924e-05,
+      "loss": 0.5222,
+      "step": 24610
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.758443779392903e-05,
+      "loss": 0.5949,
+      "step": 24620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.756306113723814e-05,
+      "loss": 0.535,
+      "step": 24630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7541684480547245e-05,
+      "loss": 0.5442,
+      "step": 24640
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.752030782385635e-05,
+      "loss": 0.567,
+      "step": 24650
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7498931167165454e-05,
+      "loss": 0.5259,
+      "step": 24660
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7477554510474565e-05,
+      "loss": 0.5927,
+      "step": 24670
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7456177853783676e-05,
+      "loss": 0.5418,
+      "step": 24680
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7434801197092774e-05,
+      "loss": 0.5488,
+      "step": 24690
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7413424540401885e-05,
+      "loss": 0.577,
+      "step": 24700
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.739204788371099e-05,
+      "loss": 0.533,
+      "step": 24710
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7370671227020094e-05,
+      "loss": 0.5874,
+      "step": 24720
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.73492945703292e-05,
+      "loss": 0.5453,
+      "step": 24730
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.732791791363831e-05,
+      "loss": 0.5532,
+      "step": 24740
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7306541256947415e-05,
+      "loss": 0.5677,
+      "step": 24750
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.728516460025652e-05,
+      "loss": 0.5225,
+      "step": 24760
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.726378794356563e-05,
+      "loss": 0.5891,
+      "step": 24770
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7242411286874735e-05,
+      "loss": 0.5469,
+      "step": 24780
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.722103463018384e-05,
+      "loss": 0.5421,
+      "step": 24790
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.719965797349295e-05,
+      "loss": 0.5721,
+      "step": 24800
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7178281316802055e-05,
+      "loss": 0.5333,
+      "step": 24810
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.715690466011116e-05,
+      "loss": 0.5845,
+      "step": 24820
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7135528003420264e-05,
+      "loss": 0.5392,
+      "step": 24830
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7114151346729376e-05,
+      "loss": 0.5587,
+      "step": 24840
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.709277469003848e-05,
+      "loss": 0.5754,
+      "step": 24850
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7071398033347585e-05,
+      "loss": 0.5245,
+      "step": 24860
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7050021376656696e-05,
+      "loss": 0.5847,
+      "step": 24870
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.70286447199658e-05,
+      "loss": 0.537,
+      "step": 24880
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7007268063274905e-05,
+      "loss": 0.5571,
+      "step": 24890
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.6985891406584016e-05,
+      "loss": 0.5724,
+      "step": 24900
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.696451474989312e-05,
+      "loss": 0.5285,
+      "step": 24910
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.6943138093202225e-05,
+      "loss": 0.5838,
+      "step": 24920
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.692176143651133e-05,
+      "loss": 0.5317,
+      "step": 24930
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.690038477982044e-05,
+      "loss": 0.5509,
+      "step": 24940
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.6879008123129546e-05,
+      "loss": 0.574,
+      "step": 24950
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.685763146643865e-05,
+      "loss": 0.5316,
+      "step": 24960
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.683625480974776e-05,
+      "loss": 0.5902,
+      "step": 24970
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6814878153056866e-05,
+      "loss": 0.5352,
+      "step": 24980
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.679350149636597e-05,
+      "loss": 0.5433,
+      "step": 24990
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6772124839675075e-05,
+      "loss": 0.5714,
+      "step": 25000
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6750748182984186e-05,
+      "loss": 0.5218,
+      "step": 25010
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.672937152629329e-05,
+      "loss": 0.5883,
+      "step": 25020
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6707994869602395e-05,
+      "loss": 0.5308,
+      "step": 25030
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6686618212911506e-05,
+      "loss": 0.5534,
+      "step": 25040
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6665241556220604e-05,
+      "loss": 0.5727,
+      "step": 25050
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6643864899529716e-05,
+      "loss": 0.5192,
+      "step": 25060
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.662248824283883e-05,
+      "loss": 0.592,
+      "step": 25070
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6601111586147925e-05,
+      "loss": 0.5351,
+      "step": 25080
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6579734929457036e-05,
+      "loss": 0.5477,
+      "step": 25090
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.655835827276614e-05,
+      "loss": 0.5653,
+      "step": 25100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6536981616075245e-05,
+      "loss": 0.5237,
+      "step": 25110
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.6515604959384356e-05,
+      "loss": 0.5817,
+      "step": 25120
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.649422830269346e-05,
+      "loss": 0.538,
+      "step": 25130
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.647285164600257e-05,
+      "loss": 0.542,
+      "step": 25140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.645147498931167e-05,
+      "loss": 0.5679,
+      "step": 25150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.643009833262078e-05,
+      "loss": 0.5198,
+      "step": 25160
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.640872167592989e-05,
+      "loss": 0.5858,
+      "step": 25170
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.638734501923899e-05,
+      "loss": 0.5407,
+      "step": 25180
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.63659683625481e-05,
+      "loss": 0.5444,
+      "step": 25190
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6344591705857206e-05,
+      "loss": 0.5826,
+      "step": 25200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.632321504916631e-05,
+      "loss": 0.5236,
+      "step": 25210
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.630183839247542e-05,
+      "loss": 0.5757,
+      "step": 25220
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6280461735784526e-05,
+      "loss": 0.5303,
+      "step": 25230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.625908507909363e-05,
+      "loss": 0.5476,
+      "step": 25240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6237708422402735e-05,
+      "loss": 0.5724,
+      "step": 25250
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6216331765711846e-05,
+      "loss": 0.5213,
+      "step": 25260
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.619495510902095e-05,
+      "loss": 0.5897,
+      "step": 25270
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6173578452330055e-05,
+      "loss": 0.537,
+      "step": 25280
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.615220179563917e-05,
+      "loss": 0.5486,
+      "step": 25290
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.613082513894827e-05,
+      "loss": 0.5758,
+      "step": 25300
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6109448482257376e-05,
+      "loss": 0.5252,
+      "step": 25310
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.608807182556648e-05,
+      "loss": 0.5948,
+      "step": 25320
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.606669516887559e-05,
+      "loss": 0.5435,
+      "step": 25330
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.6045318512184696e-05,
+      "loss": 0.546,
+      "step": 25340
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.60239418554938e-05,
+      "loss": 0.5653,
+      "step": 25350
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.600256519880291e-05,
+      "loss": 0.5221,
+      "step": 25360
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.5981188542112016e-05,
+      "loss": 0.5843,
+      "step": 25370
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.595981188542112e-05,
+      "loss": 0.5438,
+      "step": 25380
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.593843522873023e-05,
+      "loss": 0.5431,
+      "step": 25390
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.591705857203934e-05,
+      "loss": 0.5682,
+      "step": 25400
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.589568191534844e-05,
+      "loss": 0.5266,
+      "step": 25410
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.5874305258657546e-05,
+      "loss": 0.5849,
+      "step": 25420
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.585292860196666e-05,
+      "loss": 0.5282,
+      "step": 25430
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.583155194527576e-05,
+      "loss": 0.5434,
+      "step": 25440
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5810175288584866e-05,
+      "loss": 0.5651,
+      "step": 25450
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.578879863189398e-05,
+      "loss": 0.5248,
+      "step": 25460
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.576742197520308e-05,
+      "loss": 0.5971,
+      "step": 25470
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5746045318512186e-05,
+      "loss": 0.5333,
+      "step": 25480
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.57246686618213e-05,
+      "loss": 0.553,
+      "step": 25490
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.57032920051304e-05,
+      "loss": 0.5732,
+      "step": 25500
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5681915348439507e-05,
+      "loss": 0.5294,
+      "step": 25510
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.566053869174861e-05,
+      "loss": 0.5921,
+      "step": 25520
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.563916203505772e-05,
+      "loss": 0.5386,
+      "step": 25530
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.561778537836682e-05,
+      "loss": 0.5424,
+      "step": 25540
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.559640872167593e-05,
+      "loss": 0.5693,
+      "step": 25550
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.557503206498504e-05,
+      "loss": 0.5183,
+      "step": 25560
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.555365540829414e-05,
+      "loss": 0.5889,
+      "step": 25570
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.553227875160325e-05,
+      "loss": 0.5398,
+      "step": 25580
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5510902094912356e-05,
+      "loss": 0.5508,
+      "step": 25590
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.548952543822147e-05,
+      "loss": 0.572,
+      "step": 25600
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.546814878153057e-05,
+      "loss": 0.5295,
+      "step": 25610
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5446772124839677e-05,
+      "loss": 0.5871,
+      "step": 25620
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.542539546814879e-05,
+      "loss": 0.5449,
+      "step": 25630
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5404018811457886e-05,
+      "loss": 0.5512,
+      "step": 25640
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5382642154767e-05,
+      "loss": 0.5712,
+      "step": 25650
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.536126549807611e-05,
+      "loss": 0.5219,
+      "step": 25660
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.5339888841385206e-05,
+      "loss": 0.5876,
+      "step": 25670
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.531851218469432e-05,
+      "loss": 0.545,
+      "step": 25680
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.529713552800342e-05,
+      "loss": 0.5442,
+      "step": 25690
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.5275758871312526e-05,
+      "loss": 0.5719,
+      "step": 25700
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.525438221462164e-05,
+      "loss": 0.528,
+      "step": 25710
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.523300555793074e-05,
+      "loss": 0.589,
+      "step": 25720
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.5211628901239846e-05,
+      "loss": 0.5308,
+      "step": 25730
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.519025224454895e-05,
+      "loss": 0.5585,
+      "step": 25740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.516887558785806e-05,
+      "loss": 0.5839,
+      "step": 25750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.5147498931167174e-05,
+      "loss": 0.5289,
+      "step": 25760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.512612227447627e-05,
+      "loss": 0.5823,
+      "step": 25770
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.510474561778538e-05,
+      "loss": 0.5308,
+      "step": 25780
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.508336896109449e-05,
+      "loss": 0.5367,
+      "step": 25790
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.506199230440359e-05,
+      "loss": 0.5829,
+      "step": 25800
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.5040615647712696e-05,
+      "loss": 0.5231,
+      "step": 25810
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.501923899102181e-05,
+      "loss": 0.5893,
+      "step": 25820
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.499786233433091e-05,
+      "loss": 0.537,
+      "step": 25830
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.4976485677640016e-05,
+      "loss": 0.5455,
+      "step": 25840
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.495510902094913e-05,
+      "loss": 0.5768,
+      "step": 25850
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.493373236425823e-05,
+      "loss": 0.524,
+      "step": 25860
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.491235570756734e-05,
+      "loss": 0.5896,
+      "step": 25870
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.489097905087645e-05,
+      "loss": 0.5389,
+      "step": 25880
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.486960239418555e-05,
+      "loss": 0.5403,
+      "step": 25890
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.484822573749466e-05,
+      "loss": 0.5707,
+      "step": 25900
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.482684908080376e-05,
+      "loss": 0.5161,
+      "step": 25910
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.480547242411287e-05,
+      "loss": 0.5924,
+      "step": 25920
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.478409576742198e-05,
+      "loss": 0.5455,
+      "step": 25930
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.476271911073108e-05,
+      "loss": 0.5506,
+      "step": 25940
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.474134245404019e-05,
+      "loss": 0.5725,
+      "step": 25950
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.47199657973493e-05,
+      "loss": 0.529,
+      "step": 25960
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.46985891406584e-05,
+      "loss": 0.5901,
+      "step": 25970
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.4677212483967513e-05,
+      "loss": 0.5406,
+      "step": 25980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.465583582727662e-05,
+      "loss": 0.5513,
+      "step": 25990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.463445917058572e-05,
+      "loss": 0.5732,
+      "step": 26000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.461308251389483e-05,
+      "loss": 0.5324,
+      "step": 26010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.459170585720394e-05,
+      "loss": 0.5808,
+      "step": 26020
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.457032920051304e-05,
+      "loss": 0.5364,
+      "step": 26030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.454895254382215e-05,
+      "loss": 0.5452,
+      "step": 26040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.452757588713126e-05,
+      "loss": 0.5685,
+      "step": 26050
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.4506199230440356e-05,
+      "loss": 0.5225,
+      "step": 26060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.448482257374947e-05,
+      "loss": 0.5888,
+      "step": 26070
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.446344591705857e-05,
+      "loss": 0.5358,
+      "step": 26080
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.4442069260367683e-05,
+      "loss": 0.5433,
+      "step": 26090
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.442069260367679e-05,
+      "loss": 0.5739,
+      "step": 26100
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.439931594698589e-05,
+      "loss": 0.5268,
+      "step": 26110
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.4377939290295004e-05,
+      "loss": 0.589,
+      "step": 26120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.43565626336041e-05,
+      "loss": 0.5357,
+      "step": 26130
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.433518597691321e-05,
+      "loss": 0.5451,
+      "step": 26140
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.4313809320222324e-05,
+      "loss": 0.5656,
+      "step": 26150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.429243266353142e-05,
+      "loss": 0.5108,
+      "step": 26160
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.427105600684053e-05,
+      "loss": 0.5869,
+      "step": 26170
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.424967935014964e-05,
+      "loss": 0.5323,
+      "step": 26180
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.422830269345874e-05,
+      "loss": 0.5413,
+      "step": 26190
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.420692603676785e-05,
+      "loss": 0.5806,
+      "step": 26200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.418554938007696e-05,
+      "loss": 0.522,
+      "step": 26210
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.416417272338607e-05,
+      "loss": 0.5835,
+      "step": 26220
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.414279606669517e-05,
+      "loss": 0.5362,
+      "step": 26230
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.412141941000428e-05,
+      "loss": 0.5523,
+      "step": 26240
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.410004275331339e-05,
+      "loss": 0.5596,
+      "step": 26250
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.407866609662249e-05,
+      "loss": 0.5192,
+      "step": 26260
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.40572894399316e-05,
+      "loss": 0.5905,
+      "step": 26270
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.40359127832407e-05,
+      "loss": 0.5376,
+      "step": 26280
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.401453612654981e-05,
+      "loss": 0.5393,
+      "step": 26290
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.399315946985892e-05,
+      "loss": 0.5745,
+      "step": 26300
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.397178281316802e-05,
+      "loss": 0.523,
+      "step": 26310
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.395040615647713e-05,
+      "loss": 0.5895,
+      "step": 26320
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.392902949978623e-05,
+      "loss": 0.5351,
+      "step": 26330
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.3907652843095344e-05,
+      "loss": 0.5443,
+      "step": 26340
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.388627618640445e-05,
+      "loss": 0.5748,
+      "step": 26350
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.386489952971355e-05,
+      "loss": 0.521,
+      "step": 26360
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.3843522873022664e-05,
+      "loss": 0.5919,
+      "step": 26370
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.382214621633177e-05,
+      "loss": 0.5437,
+      "step": 26380
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.380076955964087e-05,
+      "loss": 0.5418,
+      "step": 26390
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.377939290294998e-05,
+      "loss": 0.5696,
+      "step": 26400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.375801624625909e-05,
+      "loss": 0.5231,
+      "step": 26410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.373663958956819e-05,
+      "loss": 0.5862,
+      "step": 26420
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.37152629328773e-05,
+      "loss": 0.5265,
+      "step": 26430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.369388627618641e-05,
+      "loss": 0.5446,
+      "step": 26440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3672509619495514e-05,
+      "loss": 0.5674,
+      "step": 26450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.365113296280462e-05,
+      "loss": 0.5266,
+      "step": 26460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.362975630611373e-05,
+      "loss": 0.5901,
+      "step": 26470
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3608379649422834e-05,
+      "loss": 0.5415,
+      "step": 26480
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.358700299273194e-05,
+      "loss": 0.5379,
+      "step": 26490
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.356562633604104e-05,
+      "loss": 0.5656,
+      "step": 26500
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3544249679350154e-05,
+      "loss": 0.529,
+      "step": 26510
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.352287302265926e-05,
+      "loss": 0.5908,
+      "step": 26520
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.350149636596836e-05,
+      "loss": 0.5325,
+      "step": 26530
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3480119709277474e-05,
+      "loss": 0.5465,
+      "step": 26540
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.345874305258658e-05,
+      "loss": 0.5705,
+      "step": 26550
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3437366395895684e-05,
+      "loss": 0.5226,
+      "step": 26560
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3415989739204795e-05,
+      "loss": 0.5808,
+      "step": 26570
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.33946130825139e-05,
+      "loss": 0.5355,
+      "step": 26580
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.3373236425823004e-05,
+      "loss": 0.5443,
+      "step": 26590
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.335185976913211e-05,
+      "loss": 0.5762,
+      "step": 26600
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.333048311244122e-05,
+      "loss": 0.5319,
+      "step": 26610
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.330910645575032e-05,
+      "loss": 0.5782,
+      "step": 26620
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.328772979905943e-05,
+      "loss": 0.5356,
+      "step": 26630
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.326635314236854e-05,
+      "loss": 0.5447,
+      "step": 26640
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.324497648567764e-05,
+      "loss": 0.5645,
+      "step": 26650
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.322359982898675e-05,
+      "loss": 0.5224,
+      "step": 26660
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3202223172295853e-05,
+      "loss": 0.5835,
+      "step": 26670
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.318084651560496e-05,
+      "loss": 0.5366,
+      "step": 26680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.315946985891407e-05,
+      "loss": 0.527,
+      "step": 26690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3138093202223174e-05,
+      "loss": 0.5644,
+      "step": 26700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3116716545532285e-05,
+      "loss": 0.5258,
+      "step": 26710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.309533988884138e-05,
+      "loss": 0.5855,
+      "step": 26720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3073963232150494e-05,
+      "loss": 0.5336,
+      "step": 26730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3052586575459605e-05,
+      "loss": 0.5415,
+      "step": 26740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.30312099187687e-05,
+      "loss": 0.573,
+      "step": 26750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.3009833262077814e-05,
+      "loss": 0.5242,
+      "step": 26760
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.298845660538692e-05,
+      "loss": 0.5835,
+      "step": 26770
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.2967079948696023e-05,
+      "loss": 0.5314,
+      "step": 26780
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.2945703292005135e-05,
+      "loss": 0.5343,
+      "step": 26790
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.292432663531424e-05,
+      "loss": 0.57,
+      "step": 26800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.2902949978623344e-05,
+      "loss": 0.5261,
+      "step": 26810
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.288157332193245e-05,
+      "loss": 0.5876,
+      "step": 26820
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.286019666524156e-05,
+      "loss": 0.5324,
+      "step": 26830
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.283882000855067e-05,
+      "loss": 0.5412,
+      "step": 26840
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.281744335185977e-05,
+      "loss": 0.5632,
+      "step": 26850
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.279606669516888e-05,
+      "loss": 0.5334,
+      "step": 26860
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2774690038477984e-05,
+      "loss": 0.5869,
+      "step": 26870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.275331338178709e-05,
+      "loss": 0.5311,
+      "step": 26880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.273193672509619e-05,
+      "loss": 0.5373,
+      "step": 26890
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2710560068405305e-05,
+      "loss": 0.5723,
+      "step": 26900
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.268918341171441e-05,
+      "loss": 0.5192,
+      "step": 26910
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2667806755023514e-05,
+      "loss": 0.5985,
+      "step": 26920
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2646430098332625e-05,
+      "loss": 0.5409,
+      "step": 26930
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.262505344164173e-05,
+      "loss": 0.549,
+      "step": 26940
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2603676784950834e-05,
+      "loss": 0.5693,
+      "step": 26950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2582300128259945e-05,
+      "loss": 0.522,
+      "step": 26960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.256092347156905e-05,
+      "loss": 0.5799,
+      "step": 26970
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2539546814878154e-05,
+      "loss": 0.5247,
+      "step": 26980
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.251817015818726e-05,
+      "loss": 0.5404,
+      "step": 26990
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.249679350149637e-05,
+      "loss": 0.5747,
+      "step": 27000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2475416844805475e-05,
+      "loss": 0.5198,
+      "step": 27010
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.245404018811458e-05,
+      "loss": 0.5877,
+      "step": 27020
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.243266353142369e-05,
+      "loss": 0.5352,
+      "step": 27030
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2411286874732795e-05,
+      "loss": 0.5382,
+      "step": 27040
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.23899102180419e-05,
+      "loss": 0.5681,
+      "step": 27050
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.236853356135101e-05,
+      "loss": 0.5219,
+      "step": 27060
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.2347156904660115e-05,
+      "loss": 0.5936,
+      "step": 27070
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.232578024796922e-05,
+      "loss": 0.5226,
+      "step": 27080
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2304403591278324e-05,
+      "loss": 0.5422,
+      "step": 27090
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2283026934587436e-05,
+      "loss": 0.5731,
+      "step": 27100
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.226165027789654e-05,
+      "loss": 0.5172,
+      "step": 27110
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2240273621205645e-05,
+      "loss": 0.5876,
+      "step": 27120
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2218896964514756e-05,
+      "loss": 0.5325,
+      "step": 27130
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2197520307823854e-05,
+      "loss": 0.5501,
+      "step": 27140
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2176143651132965e-05,
+      "loss": 0.5705,
+      "step": 27150
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.215476699444207e-05,
+      "loss": 0.5238,
+      "step": 27160
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.213339033775118e-05,
+      "loss": 0.586,
+      "step": 27170
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.2112013681060285e-05,
+      "loss": 0.5315,
+      "step": 27180
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.209063702436939e-05,
+      "loss": 0.5471,
+      "step": 27190
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.20692603676785e-05,
+      "loss": 0.5693,
+      "step": 27200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.20478837109876e-05,
+      "loss": 0.5162,
+      "step": 27210
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.202650705429671e-05,
+      "loss": 0.5898,
+      "step": 27220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.200513039760582e-05,
+      "loss": 0.5339,
+      "step": 27230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.198375374091492e-05,
+      "loss": 0.5419,
+      "step": 27240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.196237708422403e-05,
+      "loss": 0.5705,
+      "step": 27250
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.1941000427533135e-05,
+      "loss": 0.5172,
+      "step": 27260
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.191962377084224e-05,
+      "loss": 0.5822,
+      "step": 27270
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.189824711415135e-05,
+      "loss": 0.5297,
+      "step": 27280
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.1876870457460455e-05,
+      "loss": 0.5421,
+      "step": 27290
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.185549380076956e-05,
+      "loss": 0.5586,
+      "step": 27300
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1834117144078664e-05,
+      "loss": 0.5268,
+      "step": 27310
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1812740487387775e-05,
+      "loss": 0.5768,
+      "step": 27320
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.179136383069689e-05,
+      "loss": 0.5315,
+      "step": 27330
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1769987174005984e-05,
+      "loss": 0.5365,
+      "step": 27340
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1748610517315096e-05,
+      "loss": 0.5702,
+      "step": 27350
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.17272338606242e-05,
+      "loss": 0.537,
+      "step": 27360
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1705857203933305e-05,
+      "loss": 0.5854,
+      "step": 27370
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1684480547242416e-05,
+      "loss": 0.5334,
+      "step": 27380
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.166310389055152e-05,
+      "loss": 0.5387,
+      "step": 27390
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1641727233860625e-05,
+      "loss": 0.5669,
+      "step": 27400
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.162035057716973e-05,
+      "loss": 0.529,
+      "step": 27410
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.159897392047884e-05,
+      "loss": 0.5782,
+      "step": 27420
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1577597263787945e-05,
+      "loss": 0.5451,
+      "step": 27430
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.155622060709705e-05,
+      "loss": 0.5383,
+      "step": 27440
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.153484395040616e-05,
+      "loss": 0.565,
+      "step": 27450
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1513467293715266e-05,
+      "loss": 0.5274,
+      "step": 27460
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.149209063702437e-05,
+      "loss": 0.5822,
+      "step": 27470
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1470713980333475e-05,
+      "loss": 0.5401,
+      "step": 27480
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1449337323642586e-05,
+      "loss": 0.5455,
+      "step": 27490
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.142796066695169e-05,
+      "loss": 0.5746,
+      "step": 27500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1406584010260795e-05,
+      "loss": 0.5214,
+      "step": 27510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1385207353569906e-05,
+      "loss": 0.5706,
+      "step": 27520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.136383069687901e-05,
+      "loss": 0.5383,
+      "step": 27530
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.1342454040188115e-05,
+      "loss": 0.5477,
+      "step": 27540
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1321077383497227e-05,
+      "loss": 0.5602,
+      "step": 27550
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.129970072680633e-05,
+      "loss": 0.5158,
+      "step": 27560
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1278324070115436e-05,
+      "loss": 0.5895,
+      "step": 27570
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.125694741342454e-05,
+      "loss": 0.5286,
+      "step": 27580
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.123557075673365e-05,
+      "loss": 0.5382,
+      "step": 27590
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1214194100042756e-05,
+      "loss": 0.5638,
+      "step": 27600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.119281744335186e-05,
+      "loss": 0.5323,
+      "step": 27610
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.117144078666097e-05,
+      "loss": 0.5834,
+      "step": 27620
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1150064129970076e-05,
+      "loss": 0.5383,
+      "step": 27630
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.112868747327918e-05,
+      "loss": 0.5395,
+      "step": 27640
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.110731081658829e-05,
+      "loss": 0.5681,
+      "step": 27650
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1085934159897397e-05,
+      "loss": 0.5255,
+      "step": 27660
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.10645575032065e-05,
+      "loss": 0.5883,
+      "step": 27670
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1043180846515606e-05,
+      "loss": 0.5428,
+      "step": 27680
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.102180418982472e-05,
+      "loss": 0.5416,
+      "step": 27690
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.1000427533133815e-05,
+      "loss": 0.5681,
+      "step": 27700
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.0979050876442926e-05,
+      "loss": 0.5196,
+      "step": 27710
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.095767421975204e-05,
+      "loss": 0.5879,
+      "step": 27720
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.0936297563061135e-05,
+      "loss": 0.5448,
+      "step": 27730
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.0914920906370246e-05,
+      "loss": 0.5473,
+      "step": 27740
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.089354424967935e-05,
+      "loss": 0.5635,
+      "step": 27750
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.0872167592988455e-05,
+      "loss": 0.5224,
+      "step": 27760
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.0850790936297566e-05,
+      "loss": 0.5835,
+      "step": 27770
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.082941427960667e-05,
+      "loss": 0.5303,
+      "step": 27780
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.080803762291578e-05,
+      "loss": 0.5364,
+      "step": 27790
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.078666096622488e-05,
+      "loss": 0.5728,
+      "step": 27800
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.076528430953399e-05,
+      "loss": 0.5251,
+      "step": 27810
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.07439076528431e-05,
+      "loss": 0.5798,
+      "step": 27820
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.07225309961522e-05,
+      "loss": 0.5331,
+      "step": 27830
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.070115433946131e-05,
+      "loss": 0.5344,
+      "step": 27840
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0679777682770416e-05,
+      "loss": 0.5721,
+      "step": 27850
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.065840102607952e-05,
+      "loss": 0.5264,
+      "step": 27860
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.063702436938863e-05,
+      "loss": 0.5951,
+      "step": 27870
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0615647712697736e-05,
+      "loss": 0.5346,
+      "step": 27880
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.059427105600684e-05,
+      "loss": 0.5429,
+      "step": 27890
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0572894399315945e-05,
+      "loss": 0.56,
+      "step": 27900
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.055151774262506e-05,
+      "loss": 0.5207,
+      "step": 27910
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.053014108593416e-05,
+      "loss": 0.5894,
+      "step": 27920
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0508764429243266e-05,
+      "loss": 0.5352,
+      "step": 27930
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.048738777255238e-05,
+      "loss": 0.541,
+      "step": 27940
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.046601111586148e-05,
+      "loss": 0.5732,
+      "step": 27950
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0444634459170586e-05,
+      "loss": 0.5221,
+      "step": 27960
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.042325780247969e-05,
+      "loss": 0.5849,
+      "step": 27970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.04018811457888e-05,
+      "loss": 0.5314,
+      "step": 27980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.0380504489097906e-05,
+      "loss": 0.5391,
+      "step": 27990
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.035912783240701e-05,
+      "loss": 0.5665,
+      "step": 28000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.033775117571612e-05,
+      "loss": 0.5222,
+      "step": 28010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.031637451902523e-05,
+      "loss": 0.5772,
+      "step": 28020
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.029499786233433e-05,
+      "loss": 0.521,
+      "step": 28030
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.027362120564344e-05,
+      "loss": 0.5398,
+      "step": 28040
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.025224454895255e-05,
+      "loss": 0.583,
+      "step": 28050
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.023086789226165e-05,
+      "loss": 0.524,
+      "step": 28060
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.0209491235570756e-05,
+      "loss": 0.5916,
+      "step": 28070
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.018811457887987e-05,
+      "loss": 0.5414,
+      "step": 28080
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.016673792218897e-05,
+      "loss": 0.5315,
+      "step": 28090
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.0145361265498076e-05,
+      "loss": 0.5714,
+      "step": 28100
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.012398460880719e-05,
+      "loss": 0.5265,
+      "step": 28110
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.010260795211629e-05,
+      "loss": 0.5774,
+      "step": 28120
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.00812312954254e-05,
+      "loss": 0.5286,
+      "step": 28130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.005985463873451e-05,
+      "loss": 0.5389,
+      "step": 28140
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.003847798204361e-05,
+      "loss": 0.5623,
+      "step": 28150
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.001710132535272e-05,
+      "loss": 0.5251,
+      "step": 28160
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.999572466866182e-05,
+      "loss": 0.5805,
+      "step": 28170
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.997434801197093e-05,
+      "loss": 0.5365,
+      "step": 28180
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.995297135528004e-05,
+      "loss": 0.5396,
+      "step": 28190
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.993159469858914e-05,
+      "loss": 0.5647,
+      "step": 28200
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.991021804189825e-05,
+      "loss": 0.5197,
+      "step": 28210
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.988884138520735e-05,
+      "loss": 0.5787,
+      "step": 28220
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.986746472851646e-05,
+      "loss": 0.5316,
+      "step": 28230
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.9846088071825567e-05,
+      "loss": 0.5351,
+      "step": 28240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.982471141513468e-05,
+      "loss": 0.5673,
+      "step": 28250
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.980333475844378e-05,
+      "loss": 0.5287,
+      "step": 28260
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.978195810175289e-05,
+      "loss": 0.5749,
+      "step": 28270
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.9760581445062e-05,
+      "loss": 0.5258,
+      "step": 28280
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.9739204788371096e-05,
+      "loss": 0.5367,
+      "step": 28290
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.971782813168021e-05,
+      "loss": 0.5621,
+      "step": 28300
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.969645147498932e-05,
+      "loss": 0.5232,
+      "step": 28310
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.9675074818298416e-05,
+      "loss": 0.5871,
+      "step": 28320
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.965369816160753e-05,
+      "loss": 0.5368,
+      "step": 28330
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.963232150491663e-05,
+      "loss": 0.5434,
+      "step": 28340
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.9610944848225737e-05,
+      "loss": 0.5687,
+      "step": 28350
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.958956819153485e-05,
+      "loss": 0.5161,
+      "step": 28360
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.956819153484395e-05,
+      "loss": 0.5788,
+      "step": 28370
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.954681487815306e-05,
+      "loss": 0.5281,
+      "step": 28380
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.952543822146216e-05,
+      "loss": 0.5363,
+      "step": 28390
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.950406156477127e-05,
+      "loss": 0.5578,
+      "step": 28400
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.9482684908080384e-05,
+      "loss": 0.5191,
+      "step": 28410
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.946130825138948e-05,
+      "loss": 0.5766,
+      "step": 28420
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.943993159469859e-05,
+      "loss": 0.5241,
+      "step": 28430
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.94185549380077e-05,
+      "loss": 0.5439,
+      "step": 28440
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.93971782813168e-05,
+      "loss": 0.5761,
+      "step": 28450
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.937580162462591e-05,
+      "loss": 0.52,
+      "step": 28460
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.935442496793502e-05,
+      "loss": 0.5846,
+      "step": 28470
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.933304831124412e-05,
+      "loss": 0.5384,
+      "step": 28480
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.931167165455323e-05,
+      "loss": 0.54,
+      "step": 28490
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.929029499786234e-05,
+      "loss": 0.5724,
+      "step": 28500
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.926891834117144e-05,
+      "loss": 0.5318,
+      "step": 28510
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.924754168448055e-05,
+      "loss": 0.5892,
+      "step": 28520
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.922616502778966e-05,
+      "loss": 0.5398,
+      "step": 28530
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.920478837109876e-05,
+      "loss": 0.5405,
+      "step": 28540
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.918341171440787e-05,
+      "loss": 0.5772,
+      "step": 28550
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.916203505771697e-05,
+      "loss": 0.5195,
+      "step": 28560
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.914065840102608e-05,
+      "loss": 0.5777,
+      "step": 28570
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.911928174433519e-05,
+      "loss": 0.5357,
+      "step": 28580
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.909790508764429e-05,
+      "loss": 0.5342,
+      "step": 28590
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.9076528430953404e-05,
+      "loss": 0.5688,
+      "step": 28600
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.905515177426251e-05,
+      "loss": 0.517,
+      "step": 28610
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.903377511757161e-05,
+      "loss": 0.5809,
+      "step": 28620
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.9012398460880724e-05,
+      "loss": 0.5345,
+      "step": 28630
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.899102180418983e-05,
+      "loss": 0.536,
+      "step": 28640
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.896964514749893e-05,
+      "loss": 0.5751,
+      "step": 28650
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.894826849080804e-05,
+      "loss": 0.5207,
+      "step": 28660
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.892689183411715e-05,
+      "loss": 0.5897,
+      "step": 28670
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.890551517742625e-05,
+      "loss": 0.5367,
+      "step": 28680
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.888413852073536e-05,
+      "loss": 0.5404,
+      "step": 28690
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.886276186404447e-05,
+      "loss": 0.5679,
+      "step": 28700
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.884138520735357e-05,
+      "loss": 0.515,
+      "step": 28710
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.882000855066268e-05,
+      "loss": 0.5826,
+      "step": 28720
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.879863189397179e-05,
+      "loss": 0.5246,
+      "step": 28730
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8777255237280894e-05,
+      "loss": 0.5357,
+      "step": 28740
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.875587858059e-05,
+      "loss": 0.5704,
+      "step": 28750
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.87345019238991e-05,
+      "loss": 0.5219,
+      "step": 28760
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8713125267208214e-05,
+      "loss": 0.5877,
+      "step": 28770
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.869174861051731e-05,
+      "loss": 0.5208,
+      "step": 28780
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.867037195382642e-05,
+      "loss": 0.545,
+      "step": 28790
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8648995297135534e-05,
+      "loss": 0.5647,
+      "step": 28800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.862761864044463e-05,
+      "loss": 0.5228,
+      "step": 28810
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8606241983753743e-05,
+      "loss": 0.5823,
+      "step": 28820
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.858486532706285e-05,
+      "loss": 0.5309,
+      "step": 28830
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.856348867037195e-05,
+      "loss": 0.5351,
+      "step": 28840
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8542112013681064e-05,
+      "loss": 0.5734,
+      "step": 28850
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.852073535699017e-05,
+      "loss": 0.5169,
+      "step": 28860
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.849935870029928e-05,
+      "loss": 0.5788,
+      "step": 28870
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.847798204360838e-05,
+      "loss": 0.5346,
+      "step": 28880
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.845660538691749e-05,
+      "loss": 0.545,
+      "step": 28890
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.84352287302266e-05,
+      "loss": 0.5654,
+      "step": 28900
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.84138520735357e-05,
+      "loss": 0.523,
+      "step": 28910
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.839247541684481e-05,
+      "loss": 0.5912,
+      "step": 28920
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.837109876015391e-05,
+      "loss": 0.5321,
+      "step": 28930
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.834972210346302e-05,
+      "loss": 0.5423,
+      "step": 28940
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.832834544677213e-05,
+      "loss": 0.562,
+      "step": 28950
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8306968790081234e-05,
+      "loss": 0.5268,
+      "step": 28960
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.828559213339034e-05,
+      "loss": 0.5792,
+      "step": 28970
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.826421547669944e-05,
+      "loss": 0.5309,
+      "step": 28980
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8242838820008554e-05,
+      "loss": 0.5498,
+      "step": 28990
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.822146216331766e-05,
+      "loss": 0.5709,
+      "step": 29000
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.820008550662676e-05,
+      "loss": 0.5182,
+      "step": 29010
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8178708849935874e-05,
+      "loss": 0.5773,
+      "step": 29020
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.815733219324498e-05,
+      "loss": 0.5353,
+      "step": 29030
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.813595553655408e-05,
+      "loss": 0.5403,
+      "step": 29040
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.811457887986319e-05,
+      "loss": 0.5621,
+      "step": 29050
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.80932022231723e-05,
+      "loss": 0.5292,
+      "step": 29060
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8071825566481404e-05,
+      "loss": 0.5733,
+      "step": 29070
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.805044890979051e-05,
+      "loss": 0.5387,
+      "step": 29080
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.802907225309962e-05,
+      "loss": 0.5361,
+      "step": 29090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8007695596408724e-05,
+      "loss": 0.5673,
+      "step": 29100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.798631893971783e-05,
+      "loss": 0.5223,
+      "step": 29110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.796494228302694e-05,
+      "loss": 0.5827,
+      "step": 29120
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.7943565626336044e-05,
+      "loss": 0.536,
+      "step": 29130
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.792218896964515e-05,
+      "loss": 0.5403,
+      "step": 29140
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.790081231295425e-05,
+      "loss": 0.5643,
+      "step": 29150
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.7879435656263365e-05,
+      "loss": 0.5214,
+      "step": 29160
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.785805899957247e-05,
+      "loss": 0.5833,
+      "step": 29170
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.7836682342881574e-05,
+      "loss": 0.5306,
+      "step": 29180
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7815305686190685e-05,
+      "loss": 0.5416,
+      "step": 29190
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.779392902949979e-05,
+      "loss": 0.5692,
+      "step": 29200
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7772552372808894e-05,
+      "loss": 0.5235,
+      "step": 29210
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7751175716118005e-05,
+      "loss": 0.5803,
+      "step": 29220
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.772979905942711e-05,
+      "loss": 0.5411,
+      "step": 29230
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7708422402736214e-05,
+      "loss": 0.5404,
+      "step": 29240
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.768704574604532e-05,
+      "loss": 0.5681,
+      "step": 29250
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.766566908935443e-05,
+      "loss": 0.5234,
+      "step": 29260
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7644292432663534e-05,
+      "loss": 0.586,
+      "step": 29270
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.762291577597264e-05,
+      "loss": 0.5316,
+      "step": 29280
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.760153911928175e-05,
+      "loss": 0.537,
+      "step": 29290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.758016246259085e-05,
+      "loss": 0.5613,
+      "step": 29300
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.755878580589996e-05,
+      "loss": 0.522,
+      "step": 29310
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7537409149209064e-05,
+      "loss": 0.578,
+      "step": 29320
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.751603249251817e-05,
+      "loss": 0.5344,
+      "step": 29330
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.749465583582728e-05,
+      "loss": 0.5324,
+      "step": 29340
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7473279179136384e-05,
+      "loss": 0.5662,
+      "step": 29350
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7451902522445495e-05,
+      "loss": 0.5243,
+      "step": 29360
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.743052586575459e-05,
+      "loss": 0.5759,
+      "step": 29370
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7409149209063704e-05,
+      "loss": 0.5256,
+      "step": 29380
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7387772552372816e-05,
+      "loss": 0.5412,
+      "step": 29390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7366395895681913e-05,
+      "loss": 0.5732,
+      "step": 29400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.7345019238991025e-05,
+      "loss": 0.5222,
+      "step": 29410
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.732364258230013e-05,
+      "loss": 0.5817,
+      "step": 29420
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7302265925609234e-05,
+      "loss": 0.5352,
+      "step": 29430
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7280889268918345e-05,
+      "loss": 0.5358,
+      "step": 29440
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.725951261222745e-05,
+      "loss": 0.5681,
+      "step": 29450
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7238135955536554e-05,
+      "loss": 0.5245,
+      "step": 29460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.721675929884566e-05,
+      "loss": 0.579,
+      "step": 29470
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.719538264215477e-05,
+      "loss": 0.5244,
+      "step": 29480
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.717400598546388e-05,
+      "loss": 0.5407,
+      "step": 29490
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.715262932877298e-05,
+      "loss": 0.5686,
+      "step": 29500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.713125267208209e-05,
+      "loss": 0.5219,
+      "step": 29510
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7109876015391195e-05,
+      "loss": 0.5819,
+      "step": 29520
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.70884993587003e-05,
+      "loss": 0.5306,
+      "step": 29530
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7067122702009404e-05,
+      "loss": 0.5316,
+      "step": 29540
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7045746045318515e-05,
+      "loss": 0.5641,
+      "step": 29550
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.702436938862762e-05,
+      "loss": 0.5161,
+      "step": 29560
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.7002992731936724e-05,
+      "loss": 0.5775,
+      "step": 29570
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.6981616075245835e-05,
+      "loss": 0.5274,
+      "step": 29580
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.696023941855494e-05,
+      "loss": 0.5353,
+      "step": 29590
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.6938862761864044e-05,
+      "loss": 0.566,
+      "step": 29600
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.6917486105173156e-05,
+      "loss": 0.5192,
+      "step": 29610
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.689610944848226e-05,
+      "loss": 0.5761,
+      "step": 29620
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.6874732791791365e-05,
+      "loss": 0.5263,
+      "step": 29630
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.685335613510047e-05,
+      "loss": 0.542,
+      "step": 29640
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.683197947840958e-05,
+      "loss": 0.5681,
+      "step": 29650
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6810602821718685e-05,
+      "loss": 0.5194,
+      "step": 29660
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.678922616502779e-05,
+      "loss": 0.5802,
+      "step": 29670
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.67678495083369e-05,
+      "loss": 0.5291,
+      "step": 29680
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6746472851646005e-05,
+      "loss": 0.5354,
+      "step": 29690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.672509619495511e-05,
+      "loss": 0.5848,
+      "step": 29700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.670371953826422e-05,
+      "loss": 0.5268,
+      "step": 29710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6682342881573326e-05,
+      "loss": 0.5855,
+      "step": 29720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.666096622488243e-05,
+      "loss": 0.5323,
+      "step": 29730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6639589568191535e-05,
+      "loss": 0.5465,
+      "step": 29740
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6618212911500646e-05,
+      "loss": 0.5684,
+      "step": 29750
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.659683625480975e-05,
+      "loss": 0.5217,
+      "step": 29760
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6575459598118855e-05,
+      "loss": 0.5807,
+      "step": 29770
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6554082941427966e-05,
+      "loss": 0.5363,
+      "step": 29780
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6532706284737064e-05,
+      "loss": 0.5413,
+      "step": 29790
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6511329628046175e-05,
+      "loss": 0.5759,
+      "step": 29800
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.648995297135528e-05,
+      "loss": 0.5165,
+      "step": 29810
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.646857631466439e-05,
+      "loss": 0.5839,
+      "step": 29820
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.6447199657973496e-05,
+      "loss": 0.5301,
+      "step": 29830
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.64258230012826e-05,
+      "loss": 0.535,
+      "step": 29840
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.640444634459171e-05,
+      "loss": 0.5741,
+      "step": 29850
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.638306968790081e-05,
+      "loss": 0.517,
+      "step": 29860
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.636169303120992e-05,
+      "loss": 0.5797,
+      "step": 29870
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.634031637451903e-05,
+      "loss": 0.533,
+      "step": 29880
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.631893971782813e-05,
+      "loss": 0.5356,
+      "step": 29890
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.629756306113724e-05,
+      "loss": 0.5618,
+      "step": 29900
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6276186404446345e-05,
+      "loss": 0.5155,
+      "step": 29910
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.625480974775545e-05,
+      "loss": 0.576,
+      "step": 29920
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.623343309106456e-05,
+      "loss": 0.5279,
+      "step": 29930
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6212056434373665e-05,
+      "loss": 0.5424,
+      "step": 29940
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.619067977768277e-05,
+      "loss": 0.5668,
+      "step": 29950
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6169303120991874e-05,
+      "loss": 0.5161,
+      "step": 29960
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6147926464300986e-05,
+      "loss": 0.5808,
+      "step": 29970
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.61265498076101e-05,
+      "loss": 0.5319,
+      "step": 29980
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6105173150919195e-05,
+      "loss": 0.5384,
+      "step": 29990
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6083796494228306e-05,
+      "loss": 0.5704,
+      "step": 30000
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.606241983753741e-05,
+      "loss": 0.5184,
+      "step": 30010
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6041043180846515e-05,
+      "loss": 0.5757,
+      "step": 30020
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.6019666524155626e-05,
+      "loss": 0.5345,
+      "step": 30030
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.599828986746473e-05,
+      "loss": 0.5386,
+      "step": 30040
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.5976913210773835e-05,
+      "loss": 0.5742,
+      "step": 30050
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.595553655408294e-05,
+      "loss": 0.5198,
+      "step": 30060
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.593415989739205e-05,
+      "loss": 0.594,
+      "step": 30070
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.5912783240701156e-05,
+      "loss": 0.5332,
+      "step": 30080
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.589140658401026e-05,
+      "loss": 0.5505,
+      "step": 30090
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.587002992731937e-05,
+      "loss": 0.564,
+      "step": 30100
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.5848653270628476e-05,
+      "loss": 0.5089,
+      "step": 30110
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.582727661393758e-05,
+      "loss": 0.574,
+      "step": 30120
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5805899957246685e-05,
+      "loss": 0.53,
+      "step": 30130
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5784523300555796e-05,
+      "loss": 0.542,
+      "step": 30140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.57631466438649e-05,
+      "loss": 0.5549,
+      "step": 30150
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5741769987174005e-05,
+      "loss": 0.5212,
+      "step": 30160
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.572039333048312e-05,
+      "loss": 0.5834,
+      "step": 30170
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.569901667379222e-05,
+      "loss": 0.5249,
+      "step": 30180
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5677640017101326e-05,
+      "loss": 0.5381,
+      "step": 30190
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.565626336041044e-05,
+      "loss": 0.5712,
+      "step": 30200
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.563488670371954e-05,
+      "loss": 0.5291,
+      "step": 30210
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5613510047028646e-05,
+      "loss": 0.5785,
+      "step": 30220
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.559213339033775e-05,
+      "loss": 0.5319,
+      "step": 30230
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.557075673364686e-05,
+      "loss": 0.5481,
+      "step": 30240
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5549380076955966e-05,
+      "loss": 0.5599,
+      "step": 30250
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.552800342026507e-05,
+      "loss": 0.5166,
+      "step": 30260
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.550662676357418e-05,
+      "loss": 0.5815,
+      "step": 30270
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5485250106883287e-05,
+      "loss": 0.5281,
+      "step": 30280
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.546387345019239e-05,
+      "loss": 0.5324,
+      "step": 30290
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.54424967935015e-05,
+      "loss": 0.5662,
+      "step": 30300
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.542112013681061e-05,
+      "loss": 0.5237,
+      "step": 30310
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.539974348011971e-05,
+      "loss": 0.5843,
+      "step": 30320
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5378366823428816e-05,
+      "loss": 0.5284,
+      "step": 30330
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.535699016673793e-05,
+      "loss": 0.5364,
+      "step": 30340
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.5335613510047025e-05,
+      "loss": 0.5696,
+      "step": 30350
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.5314236853356136e-05,
+      "loss": 0.5198,
+      "step": 30360
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.529286019666525e-05,
+      "loss": 0.5781,
+      "step": 30370
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.5271483539974345e-05,
+      "loss": 0.5357,
+      "step": 30380
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.5250106883283457e-05,
+      "loss": 0.5332,
+      "step": 30390
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.522873022659256e-05,
+      "loss": 0.5574,
+      "step": 30400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.5207353569901666e-05,
+      "loss": 0.522,
+      "step": 30410
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.518597691321078e-05,
+      "loss": 0.5884,
+      "step": 30420
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.516460025651988e-05,
+      "loss": 0.5352,
+      "step": 30430
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.514322359982899e-05,
+      "loss": 0.5359,
+      "step": 30440
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.512184694313809e-05,
+      "loss": 0.5688,
+      "step": 30450
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.51004702864472e-05,
+      "loss": 0.5187,
+      "step": 30460
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.507909362975631e-05,
+      "loss": 0.5805,
+      "step": 30470
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.505771697306541e-05,
+      "loss": 0.5217,
+      "step": 30480
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.503634031637452e-05,
+      "loss": 0.5394,
+      "step": 30490
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.5014963659683626e-05,
+      "loss": 0.5767,
+      "step": 30500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.499358700299273e-05,
+      "loss": 0.525,
+      "step": 30510
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.497221034630184e-05,
+      "loss": 0.5718,
+      "step": 30520
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.495083368961095e-05,
+      "loss": 0.5236,
+      "step": 30530
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.492945703292005e-05,
+      "loss": 0.5318,
+      "step": 30540
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.4908080376229156e-05,
+      "loss": 0.5633,
+      "step": 30550
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.488670371953827e-05,
+      "loss": 0.5176,
+      "step": 30560
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.486532706284737e-05,
+      "loss": 0.5812,
+      "step": 30570
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.4843950406156476e-05,
+      "loss": 0.529,
+      "step": 30580
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.482257374946559e-05,
+      "loss": 0.5428,
+      "step": 30590
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.480119709277469e-05,
+      "loss": 0.5757,
+      "step": 30600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.4779820436083796e-05,
+      "loss": 0.5186,
+      "step": 30610
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.47584437793929e-05,
+      "loss": 0.5784,
+      "step": 30620
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.473706712270201e-05,
+      "loss": 0.534,
+      "step": 30630
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.471569046601112e-05,
+      "loss": 0.5342,
+      "step": 30640
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.469431380932022e-05,
+      "loss": 0.5708,
+      "step": 30650
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.467293715262933e-05,
+      "loss": 0.5124,
+      "step": 30660
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.465156049593844e-05,
+      "loss": 0.5683,
+      "step": 30670
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.463018383924754e-05,
+      "loss": 0.5317,
+      "step": 30680
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.460880718255665e-05,
+      "loss": 0.534,
+      "step": 30690
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.458743052586576e-05,
+      "loss": 0.5692,
+      "step": 30700
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.456605386917486e-05,
+      "loss": 0.5226,
+      "step": 30710
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.4544677212483966e-05,
+      "loss": 0.5819,
+      "step": 30720
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.452330055579308e-05,
+      "loss": 0.5248,
+      "step": 30730
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.450192389910218e-05,
+      "loss": 0.5306,
+      "step": 30740
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.448054724241129e-05,
+      "loss": 0.5615,
+      "step": 30750
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.44591705857204e-05,
+      "loss": 0.5122,
+      "step": 30760
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.44377939290295e-05,
+      "loss": 0.5765,
+      "step": 30770
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.441641727233861e-05,
+      "loss": 0.531,
+      "step": 30780
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.439504061564772e-05,
+      "loss": 0.5342,
+      "step": 30790
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.437366395895682e-05,
+      "loss": 0.557,
+      "step": 30800
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.435228730226593e-05,
+      "loss": 0.5038,
+      "step": 30810
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.433091064557503e-05,
+      "loss": 0.5785,
+      "step": 30820
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.430953398888414e-05,
+      "loss": 0.5338,
+      "step": 30830
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.428815733219325e-05,
+      "loss": 0.5405,
+      "step": 30840
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.426678067550235e-05,
+      "loss": 0.5637,
+      "step": 30850
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.4245404018811463e-05,
+      "loss": 0.5148,
+      "step": 30860
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.422402736212056e-05,
+      "loss": 0.5776,
+      "step": 30870
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.420265070542967e-05,
+      "loss": 0.5283,
+      "step": 30880
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.418127404873878e-05,
+      "loss": 0.5327,
+      "step": 30890
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.415989739204789e-05,
+      "loss": 0.5741,
+      "step": 30900
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.413852073535699e-05,
+      "loss": 0.5134,
+      "step": 30910
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.41171440786661e-05,
+      "loss": 0.5905,
+      "step": 30920
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.409576742197521e-05,
+      "loss": 0.5325,
+      "step": 30930
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.4074390765284306e-05,
+      "loss": 0.5404,
+      "step": 30940
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.405301410859342e-05,
+      "loss": 0.5671,
+      "step": 30950
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.403163745190253e-05,
+      "loss": 0.5129,
+      "step": 30960
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.4010260795211627e-05,
+      "loss": 0.5845,
+      "step": 30970
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.398888413852074e-05,
+      "loss": 0.5284,
+      "step": 30980
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.396750748182984e-05,
+      "loss": 0.5381,
+      "step": 30990
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.394613082513895e-05,
+      "loss": 0.56,
+      "step": 31000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.392475416844806e-05,
+      "loss": 0.523,
+      "step": 31010
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.390337751175716e-05,
+      "loss": 0.5794,
+      "step": 31020
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.388200085506627e-05,
+      "loss": 0.5228,
+      "step": 31030
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.386062419837537e-05,
+      "loss": 0.5308,
+      "step": 31040
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.383924754168448e-05,
+      "loss": 0.5606,
+      "step": 31050
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.3817870884993594e-05,
+      "loss": 0.5142,
+      "step": 31060
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.379649422830269e-05,
+      "loss": 0.5742,
+      "step": 31070
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.37751175716118e-05,
+      "loss": 0.5238,
+      "step": 31080
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.375374091492091e-05,
+      "loss": 0.5374,
+      "step": 31090
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.373236425823001e-05,
+      "loss": 0.5757,
+      "step": 31100
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.3710987601539124e-05,
+      "loss": 0.514,
+      "step": 31110
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.368961094484823e-05,
+      "loss": 0.5831,
+      "step": 31120
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.366823428815733e-05,
+      "loss": 0.5306,
+      "step": 31130
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.364685763146644e-05,
+      "loss": 0.5369,
+      "step": 31140
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.362548097477555e-05,
+      "loss": 0.5649,
+      "step": 31150
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.360410431808465e-05,
+      "loss": 0.5145,
+      "step": 31160
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.358272766139376e-05,
+      "loss": 0.5787,
+      "step": 31170
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.356135100470287e-05,
+      "loss": 0.5311,
+      "step": 31180
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.353997434801197e-05,
+      "loss": 0.5353,
+      "step": 31190
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.351859769132108e-05,
+      "loss": 0.5614,
+      "step": 31200
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.349722103463018e-05,
+      "loss": 0.5189,
+      "step": 31210
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.3475844377939294e-05,
+      "loss": 0.5776,
+      "step": 31220
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.34544677212484e-05,
+      "loss": 0.5321,
+      "step": 31230
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.34330910645575e-05,
+      "loss": 0.5382,
+      "step": 31240
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.3411714407866614e-05,
+      "loss": 0.5576,
+      "step": 31250
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.339033775117572e-05,
+      "loss": 0.5191,
+      "step": 31260
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.336896109448482e-05,
+      "loss": 0.5794,
+      "step": 31270
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.3347584437793934e-05,
+      "loss": 0.5326,
+      "step": 31280
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.332620778110304e-05,
+      "loss": 0.5384,
+      "step": 31290
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.330483112441214e-05,
+      "loss": 0.5657,
+      "step": 31300
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.328345446772125e-05,
+      "loss": 0.5207,
+      "step": 31310
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.326207781103036e-05,
+      "loss": 0.5721,
+      "step": 31320
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.3240701154339464e-05,
+      "loss": 0.5246,
+      "step": 31330
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.321932449764857e-05,
+      "loss": 0.5388,
+      "step": 31340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.319794784095768e-05,
+      "loss": 0.5709,
+      "step": 31350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.317657118426678e-05,
+      "loss": 0.5168,
+      "step": 31360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.315519452757589e-05,
+      "loss": 0.587,
+      "step": 31370
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.3133817870885e-05,
+      "loss": 0.5253,
+      "step": 31380
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.3112441214194104e-05,
+      "loss": 0.5419,
+      "step": 31390
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.309106455750321e-05,
+      "loss": 0.5612,
+      "step": 31400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.306968790081231e-05,
+      "loss": 0.5262,
+      "step": 31410
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.3048311244121424e-05,
+      "loss": 0.5812,
+      "step": 31420
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.302693458743052e-05,
+      "loss": 0.5289,
+      "step": 31430
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.3005557930739633e-05,
+      "loss": 0.5448,
+      "step": 31440
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.2984181274048745e-05,
+      "loss": 0.5624,
+      "step": 31450
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.296280461735784e-05,
+      "loss": 0.5226,
+      "step": 31460
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.2941427960666954e-05,
+      "loss": 0.5722,
+      "step": 31470
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.292005130397606e-05,
+      "loss": 0.5289,
+      "step": 31480
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.289867464728516e-05,
+      "loss": 0.5348,
+      "step": 31490
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.2877297990594274e-05,
+      "loss": 0.5652,
+      "step": 31500
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.285592133390338e-05,
+      "loss": 0.5143,
+      "step": 31510
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.283454467721249e-05,
+      "loss": 0.5674,
+      "step": 31520
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.281316802052159e-05,
+      "loss": 0.5198,
+      "step": 31530
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.27917913638307e-05,
+      "loss": 0.5416,
+      "step": 31540
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.277041470713981e-05,
+      "loss": 0.5736,
+      "step": 31550
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.274903805044891e-05,
+      "loss": 0.5047,
+      "step": 31560
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.272766139375802e-05,
+      "loss": 0.58,
+      "step": 31570
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2706284737067124e-05,
+      "loss": 0.5262,
+      "step": 31580
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.268490808037623e-05,
+      "loss": 0.5333,
+      "step": 31590
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.266353142368534e-05,
+      "loss": 0.5536,
+      "step": 31600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2642154766994444e-05,
+      "loss": 0.519,
+      "step": 31610
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.262077811030355e-05,
+      "loss": 0.57,
+      "step": 31620
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.259940145361265e-05,
+      "loss": 0.5244,
+      "step": 31630
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2578024796921764e-05,
+      "loss": 0.5211,
+      "step": 31640
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.255664814023087e-05,
+      "loss": 0.5688,
+      "step": 31650
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.253527148353997e-05,
+      "loss": 0.5162,
+      "step": 31660
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2513894826849085e-05,
+      "loss": 0.5749,
+      "step": 31670
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.249251817015819e-05,
+      "loss": 0.5242,
+      "step": 31680
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2471141513467294e-05,
+      "loss": 0.5234,
+      "step": 31690
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.24497648567764e-05,
+      "loss": 0.5676,
+      "step": 31700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.242838820008551e-05,
+      "loss": 0.5186,
+      "step": 31710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2407011543394614e-05,
+      "loss": 0.5789,
+      "step": 31720
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.238563488670372e-05,
+      "loss": 0.5277,
+      "step": 31730
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.236425823001283e-05,
+      "loss": 0.5412,
+      "step": 31740
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.2342881573321934e-05,
+      "loss": 0.5709,
+      "step": 31750
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.232150491663104e-05,
+      "loss": 0.5199,
+      "step": 31760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.230012825994015e-05,
+      "loss": 0.5772,
+      "step": 31770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2278751603249255e-05,
+      "loss": 0.5289,
+      "step": 31780
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.225737494655836e-05,
+      "loss": 0.5375,
+      "step": 31790
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2235998289867464e-05,
+      "loss": 0.5603,
+      "step": 31800
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2214621633176575e-05,
+      "loss": 0.5236,
+      "step": 31810
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.219324497648568e-05,
+      "loss": 0.5784,
+      "step": 31820
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2171868319794784e-05,
+      "loss": 0.5291,
+      "step": 31830
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2150491663103895e-05,
+      "loss": 0.5332,
+      "step": 31840
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2129115006413e-05,
+      "loss": 0.5531,
+      "step": 31850
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2107738349722104e-05,
+      "loss": 0.5193,
+      "step": 31860
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2086361693031216e-05,
+      "loss": 0.5864,
+      "step": 31870
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.206498503634032e-05,
+      "loss": 0.5257,
+      "step": 31880
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.2043608379649425e-05,
+      "loss": 0.5277,
+      "step": 31890
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.202223172295853e-05,
+      "loss": 0.5615,
+      "step": 31900
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.200085506626764e-05,
+      "loss": 0.5131,
+      "step": 31910
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.1979478409576745e-05,
+      "loss": 0.5847,
+      "step": 31920
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.195810175288585e-05,
+      "loss": 0.5307,
+      "step": 31930
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.193672509619496e-05,
+      "loss": 0.5336,
+      "step": 31940
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.191534843950406e-05,
+      "loss": 0.5623,
+      "step": 31950
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.189397178281317e-05,
+      "loss": 0.5168,
+      "step": 31960
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.1872595126122274e-05,
+      "loss": 0.5748,
+      "step": 31970
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.185121846943138e-05,
+      "loss": 0.5342,
+      "step": 31980
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.182984181274049e-05,
+      "loss": 0.5413,
+      "step": 31990
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1808465156049594e-05,
+      "loss": 0.5689,
+      "step": 32000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1787088499358706e-05,
+      "loss": 0.5097,
+      "step": 32010
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1765711842667804e-05,
+      "loss": 0.5798,
+      "step": 32020
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1744335185976915e-05,
+      "loss": 0.5389,
+      "step": 32030
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1722958529286026e-05,
+      "loss": 0.5394,
+      "step": 32040
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1701581872595124e-05,
+      "loss": 0.5643,
+      "step": 32050
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1680205215904235e-05,
+      "loss": 0.513,
+      "step": 32060
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.165882855921334e-05,
+      "loss": 0.5777,
+      "step": 32070
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1637451902522444e-05,
+      "loss": 0.5335,
+      "step": 32080
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1616075245831555e-05,
+      "loss": 0.5365,
+      "step": 32090
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.159469858914066e-05,
+      "loss": 0.5718,
+      "step": 32100
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1573321932449764e-05,
+      "loss": 0.5076,
+      "step": 32110
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.155194527575887e-05,
+      "loss": 0.5851,
+      "step": 32120
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.153056861906798e-05,
+      "loss": 0.528,
+      "step": 32130
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.150919196237709e-05,
+      "loss": 0.5316,
+      "step": 32140
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.148781530568619e-05,
+      "loss": 0.5575,
+      "step": 32150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.14664386489953e-05,
+      "loss": 0.5165,
+      "step": 32160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1445061992304405e-05,
+      "loss": 0.5719,
+      "step": 32170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.142368533561351e-05,
+      "loss": 0.5301,
+      "step": 32180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.140230867892262e-05,
+      "loss": 0.5373,
+      "step": 32190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1380932022231725e-05,
+      "loss": 0.5671,
+      "step": 32200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.135955536554083e-05,
+      "loss": 0.5187,
+      "step": 32210
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1338178708849934e-05,
+      "loss": 0.5788,
+      "step": 32220
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.1316802052159046e-05,
+      "loss": 0.5255,
+      "step": 32230
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.129542539546815e-05,
+      "loss": 0.5345,
+      "step": 32240
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1274048738777255e-05,
+      "loss": 0.566,
+      "step": 32250
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1252672082086366e-05,
+      "loss": 0.5161,
+      "step": 32260
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.123129542539547e-05,
+      "loss": 0.5795,
+      "step": 32270
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1209918768704575e-05,
+      "loss": 0.5289,
+      "step": 32280
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.118854211201368e-05,
+      "loss": 0.5391,
+      "step": 32290
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.116716545532279e-05,
+      "loss": 0.5624,
+      "step": 32300
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1145788798631895e-05,
+      "loss": 0.5146,
+      "step": 32310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1124412141941e-05,
+      "loss": 0.5868,
+      "step": 32320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.110303548525011e-05,
+      "loss": 0.5275,
+      "step": 32330
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1081658828559216e-05,
+      "loss": 0.5313,
+      "step": 32340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.106028217186832e-05,
+      "loss": 0.5648,
+      "step": 32350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.103890551517743e-05,
+      "loss": 0.5076,
+      "step": 32360
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.1017528858486536e-05,
+      "loss": 0.5637,
+      "step": 32370
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.099615220179564e-05,
+      "loss": 0.5254,
+      "step": 32380
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0974775545104745e-05,
+      "loss": 0.526,
+      "step": 32390
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0953398888413856e-05,
+      "loss": 0.5633,
+      "step": 32400
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.093202223172296e-05,
+      "loss": 0.5195,
+      "step": 32410
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0910645575032065e-05,
+      "loss": 0.575,
+      "step": 32420
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0889268918341177e-05,
+      "loss": 0.5296,
+      "step": 32430
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0867892261650274e-05,
+      "loss": 0.539,
+      "step": 32440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.0846515604959386e-05,
+      "loss": 0.5581,
+      "step": 32450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.08251389482685e-05,
+      "loss": 0.5106,
+      "step": 32460
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.08037622915776e-05,
+      "loss": 0.5786,
+      "step": 32470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0782385634886706e-05,
+      "loss": 0.5268,
+      "step": 32480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.076100897819581e-05,
+      "loss": 0.5383,
+      "step": 32490
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.073963232150492e-05,
+      "loss": 0.5595,
+      "step": 32500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.071825566481402e-05,
+      "loss": 0.5094,
+      "step": 32510
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.069687900812313e-05,
+      "loss": 0.5772,
+      "step": 32520
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.067550235143224e-05,
+      "loss": 0.5309,
+      "step": 32530
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.065412569474134e-05,
+      "loss": 0.5325,
+      "step": 32540
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.063274903805045e-05,
+      "loss": 0.5613,
+      "step": 32550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0611372381359556e-05,
+      "loss": 0.5116,
+      "step": 32560
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.058999572466866e-05,
+      "loss": 0.569,
+      "step": 32570
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.056861906797777e-05,
+      "loss": 0.5261,
+      "step": 32580
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0547242411286876e-05,
+      "loss": 0.5367,
+      "step": 32590
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.052586575459598e-05,
+      "loss": 0.5703,
+      "step": 32600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0504489097905088e-05,
+      "loss": 0.5205,
+      "step": 32610
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0483112441214196e-05,
+      "loss": 0.5781,
+      "step": 32620
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0461735784523304e-05,
+      "loss": 0.5207,
+      "step": 32630
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.044035912783241e-05,
+      "loss": 0.5354,
+      "step": 32640
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0418982471141516e-05,
+      "loss": 0.5669,
+      "step": 32650
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.039760581445062e-05,
+      "loss": 0.5098,
+      "step": 32660
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.037622915775973e-05,
+      "loss": 0.569,
+      "step": 32670
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.0354852501068837e-05,
+      "loss": 0.5232,
+      "step": 32680
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.033347584437794e-05,
+      "loss": 0.5257,
+      "step": 32690
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.031209918768705e-05,
+      "loss": 0.5668,
+      "step": 32700
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.029072253099615e-05,
+      "loss": 0.5144,
+      "step": 32710
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.026934587430526e-05,
+      "loss": 0.5834,
+      "step": 32720
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.024796921761437e-05,
+      "loss": 0.5206,
+      "step": 32730
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0226592560923474e-05,
+      "loss": 0.5319,
+      "step": 32740
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0205215904232582e-05,
+      "loss": 0.5594,
+      "step": 32750
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0183839247541683e-05,
+      "loss": 0.5162,
+      "step": 32760
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0162462590850794e-05,
+      "loss": 0.578,
+      "step": 32770
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0141085934159895e-05,
+      "loss": 0.5254,
+      "step": 32780
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0119709277469003e-05,
+      "loss": 0.5366,
+      "step": 32790
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0098332620778115e-05,
+      "loss": 0.5704,
+      "step": 32800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0076955964087216e-05,
+      "loss": 0.5144,
+      "step": 32810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0055579307396327e-05,
+      "loss": 0.5798,
+      "step": 32820
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0034202650705428e-05,
+      "loss": 0.5268,
+      "step": 32830
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.0012825994014536e-05,
+      "loss": 0.5343,
+      "step": 32840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9991449337323647e-05,
+      "loss": 0.5715,
+      "step": 32850
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.997007268063275e-05,
+      "loss": 0.5162,
+      "step": 32860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9948696023941856e-05,
+      "loss": 0.5786,
+      "step": 32870
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.992731936725096e-05,
+      "loss": 0.5312,
+      "step": 32880
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.990594271056007e-05,
+      "loss": 0.5359,
+      "step": 32890
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.988456605386918e-05,
+      "loss": 0.5594,
+      "step": 32900
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.986318939717828e-05,
+      "loss": 0.5146,
+      "step": 32910
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.984181274048739e-05,
+      "loss": 0.5782,
+      "step": 32920
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.9820436083796494e-05,
+      "loss": 0.5227,
+      "step": 32930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.97990594271056e-05,
+      "loss": 0.5331,
+      "step": 32940
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.977768277041471e-05,
+      "loss": 0.5606,
+      "step": 32950
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9756306113723814e-05,
+      "loss": 0.5165,
+      "step": 32960
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9734929457032922e-05,
+      "loss": 0.5736,
+      "step": 32970
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9713552800342026e-05,
+      "loss": 0.5159,
+      "step": 32980
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9692176143651134e-05,
+      "loss": 0.5341,
+      "step": 32990
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9670799486960242e-05,
+      "loss": 0.5564,
+      "step": 33000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9649422830269347e-05,
+      "loss": 0.5148,
+      "step": 33010
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9628046173578454e-05,
+      "loss": 0.5668,
+      "step": 33020
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.960666951688756e-05,
+      "loss": 0.531,
+      "step": 33030
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9585292860196667e-05,
+      "loss": 0.5342,
+      "step": 33040
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.956391620350577e-05,
+      "loss": 0.5591,
+      "step": 33050
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.954253954681488e-05,
+      "loss": 0.504,
+      "step": 33060
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9521162890123987e-05,
+      "loss": 0.5767,
+      "step": 33070
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9499786233433092e-05,
+      "loss": 0.5295,
+      "step": 33080
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.94784095767422e-05,
+      "loss": 0.5222,
+      "step": 33090
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9457032920051304e-05,
+      "loss": 0.5504,
+      "step": 33100
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9435656263360412e-05,
+      "loss": 0.5076,
+      "step": 33110
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.941427960666952e-05,
+      "loss": 0.5766,
+      "step": 33120
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9392902949978624e-05,
+      "loss": 0.5286,
+      "step": 33130
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9371526293287732e-05,
+      "loss": 0.543,
+      "step": 33140
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9350149636596837e-05,
+      "loss": 0.5627,
+      "step": 33150
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9328772979905945e-05,
+      "loss": 0.5135,
+      "step": 33160
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9307396323215053e-05,
+      "loss": 0.5714,
+      "step": 33170
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9286019666524157e-05,
+      "loss": 0.5228,
+      "step": 33180
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9264643009833265e-05,
+      "loss": 0.5269,
+      "step": 33190
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9243266353142366e-05,
+      "loss": 0.5536,
+      "step": 33200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9221889696451477e-05,
+      "loss": 0.5151,
+      "step": 33210
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9200513039760585e-05,
+      "loss": 0.573,
+      "step": 33220
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.917913638306969e-05,
+      "loss": 0.5246,
+      "step": 33230
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9157759726378798e-05,
+      "loss": 0.5243,
+      "step": 33240
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.91363830696879e-05,
+      "loss": 0.5613,
+      "step": 33250
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.911500641299701e-05,
+      "loss": 0.5062,
+      "step": 33260
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9093629756306118e-05,
+      "loss": 0.5688,
+      "step": 33270
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9072253099615223e-05,
+      "loss": 0.5271,
+      "step": 33280
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.905087644292433e-05,
+      "loss": 0.5306,
+      "step": 33290
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.902949978623343e-05,
+      "loss": 0.5638,
+      "step": 33300
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9008123129542543e-05,
+      "loss": 0.5098,
+      "step": 33310
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8986746472851644e-05,
+      "loss": 0.5746,
+      "step": 33320
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8965369816160752e-05,
+      "loss": 0.5254,
+      "step": 33330
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8943993159469863e-05,
+      "loss": 0.5336,
+      "step": 33340
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8922616502778964e-05,
+      "loss": 0.564,
+      "step": 33350
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8901239846088076e-05,
+      "loss": 0.5116,
+      "step": 33360
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8879863189397177e-05,
+      "loss": 0.5709,
+      "step": 33370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8858486532706285e-05,
+      "loss": 0.5363,
+      "step": 33380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8837109876015396e-05,
+      "loss": 0.5337,
+      "step": 33390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8815733219324497e-05,
+      "loss": 0.5698,
+      "step": 33400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8794356562633605e-05,
+      "loss": 0.5178,
+      "step": 33410
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.877297990594271e-05,
+      "loss": 0.5746,
+      "step": 33420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8751603249251817e-05,
+      "loss": 0.5183,
+      "step": 33430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.873022659256093e-05,
+      "loss": 0.5262,
+      "step": 33440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.870884993587003e-05,
+      "loss": 0.5651,
+      "step": 33450
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8687473279179138e-05,
+      "loss": 0.5155,
+      "step": 33460
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8666096622488242e-05,
+      "loss": 0.5805,
+      "step": 33470
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.864471996579735e-05,
+      "loss": 0.5248,
+      "step": 33480
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8623343309106458e-05,
+      "loss": 0.5344,
+      "step": 33490
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8601966652415562e-05,
+      "loss": 0.5593,
+      "step": 33500
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.858058999572467e-05,
+      "loss": 0.5115,
+      "step": 33510
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8559213339033775e-05,
+      "loss": 0.5824,
+      "step": 33520
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8537836682342883e-05,
+      "loss": 0.5263,
+      "step": 33530
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.851646002565199e-05,
+      "loss": 0.5296,
+      "step": 33540
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8495083368961095e-05,
+      "loss": 0.5629,
+      "step": 33550
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8473706712270203e-05,
+      "loss": 0.5067,
+      "step": 33560
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8452330055579308e-05,
+      "loss": 0.5842,
+      "step": 33570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8430953398888416e-05,
+      "loss": 0.5331,
+      "step": 33580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.840957674219752e-05,
+      "loss": 0.533,
+      "step": 33590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8388200085506628e-05,
+      "loss": 0.564,
+      "step": 33600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8366823428815736e-05,
+      "loss": 0.512,
+      "step": 33610
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.834544677212484e-05,
+      "loss": 0.5764,
+      "step": 33620
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.8324070115433948e-05,
+      "loss": 0.5182,
+      "step": 33630
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8302693458743053e-05,
+      "loss": 0.5308,
+      "step": 33640
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.828131680205216e-05,
+      "loss": 0.5644,
+      "step": 33650
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.825994014536127e-05,
+      "loss": 0.5179,
+      "step": 33660
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8238563488670373e-05,
+      "loss": 0.5764,
+      "step": 33670
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.821718683197948e-05,
+      "loss": 0.5243,
+      "step": 33680
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8195810175288585e-05,
+      "loss": 0.5313,
+      "step": 33690
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8174433518597693e-05,
+      "loss": 0.55,
+      "step": 33700
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.81530568619068e-05,
+      "loss": 0.5085,
+      "step": 33710
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8131680205215906e-05,
+      "loss": 0.5797,
+      "step": 33720
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8110303548525014e-05,
+      "loss": 0.5286,
+      "step": 33730
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8088926891834115e-05,
+      "loss": 0.5265,
+      "step": 33740
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8067550235143226e-05,
+      "loss": 0.563,
+      "step": 33750
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8046173578452334e-05,
+      "loss": 0.5141,
+      "step": 33760
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.802479692176144e-05,
+      "loss": 0.5775,
+      "step": 33770
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.8003420265070546e-05,
+      "loss": 0.5321,
+      "step": 33780
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7982043608379648e-05,
+      "loss": 0.5328,
+      "step": 33790
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.796066695168876e-05,
+      "loss": 0.5606,
+      "step": 33800
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7939290294997867e-05,
+      "loss": 0.5112,
+      "step": 33810
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7917913638306968e-05,
+      "loss": 0.5636,
+      "step": 33820
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.789653698161608e-05,
+      "loss": 0.5164,
+      "step": 33830
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.787516032492518e-05,
+      "loss": 0.5254,
+      "step": 33840
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.785378366823429e-05,
+      "loss": 0.5659,
+      "step": 33850
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7832407011543393e-05,
+      "loss": 0.5072,
+      "step": 33860
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.78110303548525e-05,
+      "loss": 0.5785,
+      "step": 33870
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7789653698161612e-05,
+      "loss": 0.5266,
+      "step": 33880
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7768277041470713e-05,
+      "loss": 0.5279,
+      "step": 33890
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7746900384779824e-05,
+      "loss": 0.5596,
+      "step": 33900
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7725523728088925e-05,
+      "loss": 0.5146,
+      "step": 33910
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7704147071398033e-05,
+      "loss": 0.5722,
+      "step": 33920
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7682770414707145e-05,
+      "loss": 0.5152,
+      "step": 33930
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7661393758016246e-05,
+      "loss": 0.5321,
+      "step": 33940
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7640017101325354e-05,
+      "loss": 0.5576,
+      "step": 33950
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7618640444634458e-05,
+      "loss": 0.5005,
+      "step": 33960
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7597263787943566e-05,
+      "loss": 0.5738,
+      "step": 33970
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7575887131252677e-05,
+      "loss": 0.5172,
+      "step": 33980
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.755451047456178e-05,
+      "loss": 0.5321,
+      "step": 33990
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7533133817870886e-05,
+      "loss": 0.5577,
+      "step": 34000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.751175716117999e-05,
+      "loss": 0.5174,
+      "step": 34010
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.74903805044891e-05,
+      "loss": 0.5681,
+      "step": 34020
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7469003847798207e-05,
+      "loss": 0.5274,
+      "step": 34030
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.744762719110731e-05,
+      "loss": 0.5365,
+      "step": 34040
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.742625053441642e-05,
+      "loss": 0.5657,
+      "step": 34050
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7404873877725524e-05,
+      "loss": 0.5049,
+      "step": 34060
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.738349722103463e-05,
+      "loss": 0.58,
+      "step": 34070
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.736212056434374e-05,
+      "loss": 0.5251,
+      "step": 34080
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7340743907652844e-05,
+      "loss": 0.5245,
+      "step": 34090
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7319367250961952e-05,
+      "loss": 0.5592,
+      "step": 34100
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7297990594271056e-05,
+      "loss": 0.5078,
+      "step": 34110
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7276613937580164e-05,
+      "loss": 0.5709,
+      "step": 34120
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.725523728088927e-05,
+      "loss": 0.5226,
+      "step": 34130
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7233860624198377e-05,
+      "loss": 0.5295,
+      "step": 34140
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7212483967507484e-05,
+      "loss": 0.5535,
+      "step": 34150
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.719110731081659e-05,
+      "loss": 0.5152,
+      "step": 34160
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7169730654125697e-05,
+      "loss": 0.5658,
+      "step": 34170
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.71483539974348e-05,
+      "loss": 0.5261,
+      "step": 34180
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.712697734074391e-05,
+      "loss": 0.5375,
+      "step": 34190
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7105600684053017e-05,
+      "loss": 0.5523,
+      "step": 34200
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.708422402736212e-05,
+      "loss": 0.5182,
+      "step": 34210
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.706284737067123e-05,
+      "loss": 0.578,
+      "step": 34220
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7041470713980334e-05,
+      "loss": 0.5227,
+      "step": 34230
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7020094057289442e-05,
+      "loss": 0.5195,
+      "step": 34240
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.699871740059855e-05,
+      "loss": 0.5665,
+      "step": 34250
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6977340743907654e-05,
+      "loss": 0.5152,
+      "step": 34260
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6955964087216762e-05,
+      "loss": 0.5786,
+      "step": 34270
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6934587430525863e-05,
+      "loss": 0.5187,
+      "step": 34280
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6913210773834975e-05,
+      "loss": 0.5336,
+      "step": 34290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6891834117144083e-05,
+      "loss": 0.5545,
+      "step": 34300
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6870457460453187e-05,
+      "loss": 0.5191,
+      "step": 34310
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6849080803762295e-05,
+      "loss": 0.5709,
+      "step": 34320
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6827704147071396e-05,
+      "loss": 0.5237,
+      "step": 34330
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.6806327490380507e-05,
+      "loss": 0.531,
+      "step": 34340
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6784950833689615e-05,
+      "loss": 0.5636,
+      "step": 34350
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6763574176998716e-05,
+      "loss": 0.519,
+      "step": 34360
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6742197520307828e-05,
+      "loss": 0.5738,
+      "step": 34370
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.672082086361693e-05,
+      "loss": 0.5341,
+      "step": 34380
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.669944420692604e-05,
+      "loss": 0.5372,
+      "step": 34390
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.667806755023514e-05,
+      "loss": 0.5558,
+      "step": 34400
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.665669089354425e-05,
+      "loss": 0.5109,
+      "step": 34410
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.663531423685336e-05,
+      "loss": 0.5724,
+      "step": 34420
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.661393758016246e-05,
+      "loss": 0.5174,
+      "step": 34430
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.659256092347157e-05,
+      "loss": 0.5345,
+      "step": 34440
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6571184266780674e-05,
+      "loss": 0.5633,
+      "step": 34450
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6549807610089782e-05,
+      "loss": 0.517,
+      "step": 34460
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6528430953398893e-05,
+      "loss": 0.5758,
+      "step": 34470
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6507054296707994e-05,
+      "loss": 0.5154,
+      "step": 34480
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6485677640017102e-05,
+      "loss": 0.534,
+      "step": 34490
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6464300983326207e-05,
+      "loss": 0.5624,
+      "step": 34500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6442924326635315e-05,
+      "loss": 0.5152,
+      "step": 34510
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6421547669944426e-05,
+      "loss": 0.5726,
+      "step": 34520
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6400171013253527e-05,
+      "loss": 0.5289,
+      "step": 34530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6378794356562635e-05,
+      "loss": 0.5271,
+      "step": 34540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.635741769987174e-05,
+      "loss": 0.5541,
+      "step": 34550
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6336041043180847e-05,
+      "loss": 0.5079,
+      "step": 34560
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6314664386489955e-05,
+      "loss": 0.5677,
+      "step": 34570
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.629328772979906e-05,
+      "loss": 0.5311,
+      "step": 34580
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6271911073108168e-05,
+      "loss": 0.5256,
+      "step": 34590
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6250534416417272e-05,
+      "loss": 0.5566,
+      "step": 34600
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.622915775972638e-05,
+      "loss": 0.5126,
+      "step": 34610
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6207781103035488e-05,
+      "loss": 0.5754,
+      "step": 34620
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6186404446344592e-05,
+      "loss": 0.5214,
+      "step": 34630
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.61650277896537e-05,
+      "loss": 0.5213,
+      "step": 34640
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6143651132962805e-05,
+      "loss": 0.5616,
+      "step": 34650
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6122274476271913e-05,
+      "loss": 0.5136,
+      "step": 34660
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6100897819581017e-05,
+      "loss": 0.5736,
+      "step": 34670
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6079521162890125e-05,
+      "loss": 0.5249,
+      "step": 34680
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6058144506199233e-05,
+      "loss": 0.5254,
+      "step": 34690
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6036767849508338e-05,
+      "loss": 0.5509,
+      "step": 34700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6015391192817445e-05,
+      "loss": 0.5115,
+      "step": 34710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.599401453612655e-05,
+      "loss": 0.5785,
+      "step": 34720
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5972637879435658e-05,
+      "loss": 0.525,
+      "step": 34730
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5951261222744766e-05,
+      "loss": 0.5265,
+      "step": 34740
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.592988456605387e-05,
+      "loss": 0.5641,
+      "step": 34750
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5908507909362978e-05,
+      "loss": 0.5076,
+      "step": 34760
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5887131252672083e-05,
+      "loss": 0.5688,
+      "step": 34770
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.586575459598119e-05,
+      "loss": 0.5368,
+      "step": 34780
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.58443779392903e-05,
+      "loss": 0.5348,
+      "step": 34790
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.5823001282599403e-05,
+      "loss": 0.5631,
+      "step": 34800
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.580162462590851e-05,
+      "loss": 0.5054,
+      "step": 34810
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5780247969217612e-05,
+      "loss": 0.5692,
+      "step": 34820
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5758871312526723e-05,
+      "loss": 0.5298,
+      "step": 34830
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.573749465583583e-05,
+      "loss": 0.5358,
+      "step": 34840
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5716117999144936e-05,
+      "loss": 0.557,
+      "step": 34850
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5694741342454044e-05,
+      "loss": 0.5155,
+      "step": 34860
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5673364685763145e-05,
+      "loss": 0.5732,
+      "step": 34870
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5651988029072256e-05,
+      "loss": 0.5141,
+      "step": 34880
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5630611372381364e-05,
+      "loss": 0.5327,
+      "step": 34890
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5609234715690465e-05,
+      "loss": 0.5547,
+      "step": 34900
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5587858058999576e-05,
+      "loss": 0.5121,
+      "step": 34910
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5566481402308677e-05,
+      "loss": 0.5634,
+      "step": 34920
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.554510474561779e-05,
+      "loss": 0.5217,
+      "step": 34930
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.552372808892689e-05,
+      "loss": 0.5303,
+      "step": 34940
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5502351432235998e-05,
+      "loss": 0.5592,
+      "step": 34950
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.548097477554511e-05,
+      "loss": 0.5127,
+      "step": 34960
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.545959811885421e-05,
+      "loss": 0.5743,
+      "step": 34970
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5438221462163318e-05,
+      "loss": 0.5207,
+      "step": 34980
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5416844805472423e-05,
+      "loss": 0.5298,
+      "step": 34990
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.539546814878153e-05,
+      "loss": 0.554,
+      "step": 35000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5374091492090642e-05,
+      "loss": 0.5228,
+      "step": 35010
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5352714835399743e-05,
+      "loss": 0.5705,
+      "step": 35020
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.533133817870885e-05,
+      "loss": 0.5192,
+      "step": 35030
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.5309961522017955e-05,
+      "loss": 0.5324,
+      "step": 35040
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5288584865327063e-05,
+      "loss": 0.563,
+      "step": 35050
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.526720820863617e-05,
+      "loss": 0.5167,
+      "step": 35060
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5245831551945276e-05,
+      "loss": 0.5698,
+      "step": 35070
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5224454895254384e-05,
+      "loss": 0.5215,
+      "step": 35080
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5203078238563488e-05,
+      "loss": 0.5319,
+      "step": 35090
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5181701581872596e-05,
+      "loss": 0.561,
+      "step": 35100
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5160324925181704e-05,
+      "loss": 0.5154,
+      "step": 35110
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.513894826849081e-05,
+      "loss": 0.5704,
+      "step": 35120
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5117571611799916e-05,
+      "loss": 0.5174,
+      "step": 35130
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.509619495510902e-05,
+      "loss": 0.5323,
+      "step": 35140
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.507481829841813e-05,
+      "loss": 0.5601,
+      "step": 35150
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5053441641727237e-05,
+      "loss": 0.5,
+      "step": 35160
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.503206498503634e-05,
+      "loss": 0.564,
+      "step": 35170
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.501068832834545e-05,
+      "loss": 0.5187,
+      "step": 35180
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4989311671654557e-05,
+      "loss": 0.5252,
+      "step": 35190
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.496793501496366e-05,
+      "loss": 0.5566,
+      "step": 35200
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4946558358272766e-05,
+      "loss": 0.5152,
+      "step": 35210
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4925181701581874e-05,
+      "loss": 0.5704,
+      "step": 35220
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4903805044890978e-05,
+      "loss": 0.5178,
+      "step": 35230
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.488242838820009e-05,
+      "loss": 0.5296,
+      "step": 35240
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4861051731509194e-05,
+      "loss": 0.5649,
+      "step": 35250
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.48396750748183e-05,
+      "loss": 0.507,
+      "step": 35260
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4818298418127406e-05,
+      "loss": 0.5789,
+      "step": 35270
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.479692176143651e-05,
+      "loss": 0.53,
+      "step": 35280
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.477554510474562e-05,
+      "loss": 0.5291,
+      "step": 35290
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4754168448054727e-05,
+      "loss": 0.5581,
+      "step": 35300
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.473279179136383e-05,
+      "loss": 0.5144,
+      "step": 35310
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.471141513467294e-05,
+      "loss": 0.578,
+      "step": 35320
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4690038477982044e-05,
+      "loss": 0.5249,
+      "step": 35330
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.466866182129115e-05,
+      "loss": 0.5376,
+      "step": 35340
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.464728516460026e-05,
+      "loss": 0.5595,
+      "step": 35350
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4625908507909364e-05,
+      "loss": 0.5096,
+      "step": 35360
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4604531851218472e-05,
+      "loss": 0.5741,
+      "step": 35370
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4583155194527576e-05,
+      "loss": 0.5225,
+      "step": 35380
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4561778537836684e-05,
+      "loss": 0.5297,
+      "step": 35390
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.454040188114579e-05,
+      "loss": 0.5614,
+      "step": 35400
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4519025224454897e-05,
+      "loss": 0.516,
+      "step": 35410
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4497648567764005e-05,
+      "loss": 0.5679,
+      "step": 35420
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.447627191107311e-05,
+      "loss": 0.5222,
+      "step": 35430
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4454895254382214e-05,
+      "loss": 0.5355,
+      "step": 35440
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.443351859769132e-05,
+      "loss": 0.5556,
+      "step": 35450
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.441214194100043e-05,
+      "loss": 0.5109,
+      "step": 35460
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4390765284309537e-05,
+      "loss": 0.5739,
+      "step": 35470
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4369388627618642e-05,
+      "loss": 0.5287,
+      "step": 35480
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4348011970927746e-05,
+      "loss": 0.5242,
+      "step": 35490
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4326635314236854e-05,
+      "loss": 0.547,
+      "step": 35500
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.4305258657545962e-05,
+      "loss": 0.4988,
+      "step": 35510
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4283882000855067e-05,
+      "loss": 0.5665,
+      "step": 35520
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4262505344164175e-05,
+      "loss": 0.5159,
+      "step": 35530
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.424112868747328e-05,
+      "loss": 0.5279,
+      "step": 35540
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4219752030782387e-05,
+      "loss": 0.5495,
+      "step": 35550
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.419837537409149e-05,
+      "loss": 0.5146,
+      "step": 35560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.41769987174006e-05,
+      "loss": 0.5709,
+      "step": 35570
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4155622060709707e-05,
+      "loss": 0.5151,
+      "step": 35580
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4134245404018812e-05,
+      "loss": 0.5375,
+      "step": 35590
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.411286874732792e-05,
+      "loss": 0.5589,
+      "step": 35600
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4091492090637024e-05,
+      "loss": 0.5115,
+      "step": 35610
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4070115433946132e-05,
+      "loss": 0.5673,
+      "step": 35620
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.404873877725524e-05,
+      "loss": 0.5278,
+      "step": 35630
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4027362120564345e-05,
+      "loss": 0.5312,
+      "step": 35640
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.4005985463873452e-05,
+      "loss": 0.5608,
+      "step": 35650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3984608807182557e-05,
+      "loss": 0.5019,
+      "step": 35660
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.396323215049166e-05,
+      "loss": 0.5585,
+      "step": 35670
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3941855493800773e-05,
+      "loss": 0.5185,
+      "step": 35680
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3920478837109877e-05,
+      "loss": 0.5291,
+      "step": 35690
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3899102180418985e-05,
+      "loss": 0.5532,
+      "step": 35700
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.387772552372809e-05,
+      "loss": 0.5016,
+      "step": 35710
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3856348867037194e-05,
+      "loss": 0.5744,
+      "step": 35720
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.3834972210346305e-05,
+      "loss": 0.5189,
+      "step": 35730
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.381359555365541e-05,
+      "loss": 0.5365,
+      "step": 35740
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3792218896964514e-05,
+      "loss": 0.5557,
+      "step": 35750
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3770842240273622e-05,
+      "loss": 0.5181,
+      "step": 35760
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3749465583582727e-05,
+      "loss": 0.5725,
+      "step": 35770
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3728088926891838e-05,
+      "loss": 0.5176,
+      "step": 35780
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3706712270200943e-05,
+      "loss": 0.5292,
+      "step": 35790
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3685335613510047e-05,
+      "loss": 0.5636,
+      "step": 35800
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3663958956819155e-05,
+      "loss": 0.5078,
+      "step": 35810
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.364258230012826e-05,
+      "loss": 0.5733,
+      "step": 35820
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3621205643437368e-05,
+      "loss": 0.5194,
+      "step": 35830
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3599828986746475e-05,
+      "loss": 0.524,
+      "step": 35840
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.357845233005558e-05,
+      "loss": 0.5647,
+      "step": 35850
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3557075673364688e-05,
+      "loss": 0.5045,
+      "step": 35860
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3535699016673792e-05,
+      "loss": 0.5716,
+      "step": 35870
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.35143223599829e-05,
+      "loss": 0.5268,
+      "step": 35880
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3492945703292008e-05,
+      "loss": 0.5302,
+      "step": 35890
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3471569046601113e-05,
+      "loss": 0.561,
+      "step": 35900
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.345019238991022e-05,
+      "loss": 0.5131,
+      "step": 35910
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3428815733219325e-05,
+      "loss": 0.566,
+      "step": 35920
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3407439076528433e-05,
+      "loss": 0.5198,
+      "step": 35930
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3386062419837537e-05,
+      "loss": 0.5239,
+      "step": 35940
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3364685763146645e-05,
+      "loss": 0.5678,
+      "step": 35950
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3343309106455753e-05,
+      "loss": 0.5085,
+      "step": 35960
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3321932449764858e-05,
+      "loss": 0.5696,
+      "step": 35970
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.3300555793073962e-05,
+      "loss": 0.5329,
+      "step": 35980
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.327917913638307e-05,
+      "loss": 0.5315,
+      "step": 35990
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3257802479692178e-05,
+      "loss": 0.5494,
+      "step": 36000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3236425823001286e-05,
+      "loss": 0.5076,
+      "step": 36010
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.321504916631039e-05,
+      "loss": 0.5642,
+      "step": 36020
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3193672509619495e-05,
+      "loss": 0.5173,
+      "step": 36030
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3172295852928603e-05,
+      "loss": 0.531,
+      "step": 36040
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.315091919623771e-05,
+      "loss": 0.5557,
+      "step": 36050
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3129542539546815e-05,
+      "loss": 0.5031,
+      "step": 36060
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3108165882855923e-05,
+      "loss": 0.5672,
+      "step": 36070
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3086789226165028e-05,
+      "loss": 0.5232,
+      "step": 36080
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3065412569474136e-05,
+      "loss": 0.5402,
+      "step": 36090
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.304403591278324e-05,
+      "loss": 0.5547,
+      "step": 36100
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3022659256092348e-05,
+      "loss": 0.5044,
+      "step": 36110
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.3001282599401456e-05,
+      "loss": 0.5757,
+      "step": 36120
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.297990594271056e-05,
+      "loss": 0.5126,
+      "step": 36130
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.295852928601967e-05,
+      "loss": 0.5331,
+      "step": 36140
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.2937152629328773e-05,
+      "loss": 0.5594,
+      "step": 36150
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.291577597263788e-05,
+      "loss": 0.5237,
+      "step": 36160
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.289439931594699e-05,
+      "loss": 0.5747,
+      "step": 36170
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.2873022659256093e-05,
+      "loss": 0.523,
+      "step": 36180
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.28516460025652e-05,
+      "loss": 0.534,
+      "step": 36190
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.2830269345874306e-05,
+      "loss": 0.5549,
+      "step": 36200
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.280889268918341e-05,
+      "loss": 0.5067,
+      "step": 36210
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.278751603249252e-05,
+      "loss": 0.5756,
+      "step": 36220
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2766139375801626e-05,
+      "loss": 0.5255,
+      "step": 36230
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2744762719110734e-05,
+      "loss": 0.523,
+      "step": 36240
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2723386062419838e-05,
+      "loss": 0.5588,
+      "step": 36250
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2702009405728943e-05,
+      "loss": 0.5225,
+      "step": 36260
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2680632749038054e-05,
+      "loss": 0.5595,
+      "step": 36270
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.265925609234716e-05,
+      "loss": 0.5207,
+      "step": 36280
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2637879435656263e-05,
+      "loss": 0.5341,
+      "step": 36290
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.261650277896537e-05,
+      "loss": 0.5537,
+      "step": 36300
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2595126122274476e-05,
+      "loss": 0.5081,
+      "step": 36310
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2573749465583587e-05,
+      "loss": 0.5695,
+      "step": 36320
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.255237280889269e-05,
+      "loss": 0.5212,
+      "step": 36330
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2530996152201796e-05,
+      "loss": 0.5268,
+      "step": 36340
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2509619495510904e-05,
+      "loss": 0.549,
+      "step": 36350
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2488242838820008e-05,
+      "loss": 0.51,
+      "step": 36360
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2466866182129116e-05,
+      "loss": 0.5736,
+      "step": 36370
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2445489525438224e-05,
+      "loss": 0.5176,
+      "step": 36380
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.242411286874733e-05,
+      "loss": 0.5193,
+      "step": 36390
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2402736212056436e-05,
+      "loss": 0.567,
+      "step": 36400
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.238135955536554e-05,
+      "loss": 0.5099,
+      "step": 36410
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.235998289867465e-05,
+      "loss": 0.5755,
+      "step": 36420
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.2338606241983757e-05,
+      "loss": 0.518,
+      "step": 36430
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.231722958529286e-05,
+      "loss": 0.5304,
+      "step": 36440
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.229585292860197e-05,
+      "loss": 0.5602,
+      "step": 36450
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2274476271911074e-05,
+      "loss": 0.5131,
+      "step": 36460
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2253099615220178e-05,
+      "loss": 0.5645,
+      "step": 36470
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2231722958529286e-05,
+      "loss": 0.5124,
+      "step": 36480
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2210346301838394e-05,
+      "loss": 0.5274,
+      "step": 36490
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2188969645147502e-05,
+      "loss": 0.5584,
+      "step": 36500
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2167592988456606e-05,
+      "loss": 0.5214,
+      "step": 36510
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.214621633176571e-05,
+      "loss": 0.5685,
+      "step": 36520
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.212483967507482e-05,
+      "loss": 0.529,
+      "step": 36530
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2103463018383927e-05,
+      "loss": 0.5272,
+      "step": 36540
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2082086361693035e-05,
+      "loss": 0.5577,
+      "step": 36550
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.206070970500214e-05,
+      "loss": 0.509,
+      "step": 36560
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.2039333048311244e-05,
+      "loss": 0.5645,
+      "step": 36570
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.201795639162035e-05,
+      "loss": 0.5213,
+      "step": 36580
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.199657973492946e-05,
+      "loss": 0.5372,
+      "step": 36590
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1975203078238564e-05,
+      "loss": 0.5591,
+      "step": 36600
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1953826421547672e-05,
+      "loss": 0.5049,
+      "step": 36610
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1932449764856776e-05,
+      "loss": 0.5707,
+      "step": 36620
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1911073108165884e-05,
+      "loss": 0.5234,
+      "step": 36630
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.188969645147499e-05,
+      "loss": 0.5317,
+      "step": 36640
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1868319794784097e-05,
+      "loss": 0.5657,
+      "step": 36650
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1846943138093205e-05,
+      "loss": 0.5075,
+      "step": 36660
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.182556648140231e-05,
+      "loss": 0.5682,
+      "step": 36670
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.1804189824711417e-05,
+      "loss": 0.5286,
+      "step": 36680
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.178281316802052e-05,
+      "loss": 0.5222,
+      "step": 36690
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.176143651132963e-05,
+      "loss": 0.5531,
+      "step": 36700
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1740059854638737e-05,
+      "loss": 0.508,
+      "step": 36710
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1718683197947842e-05,
+      "loss": 0.5625,
+      "step": 36720
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.169730654125695e-05,
+      "loss": 0.5214,
+      "step": 36730
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1675929884566054e-05,
+      "loss": 0.5263,
+      "step": 36740
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.165455322787516e-05,
+      "loss": 0.5592,
+      "step": 36750
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.163317657118427e-05,
+      "loss": 0.5034,
+      "step": 36760
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1611799914493374e-05,
+      "loss": 0.5541,
+      "step": 36770
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.159042325780248e-05,
+      "loss": 0.5204,
+      "step": 36780
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1569046601111587e-05,
+      "loss": 0.5261,
+      "step": 36790
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.154766994442069e-05,
+      "loss": 0.5638,
+      "step": 36800
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1526293287729803e-05,
+      "loss": 0.5069,
+      "step": 36810
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1504916631038907e-05,
+      "loss": 0.5717,
+      "step": 36820
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1483539974348012e-05,
+      "loss": 0.5233,
+      "step": 36830
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.146216331765712e-05,
+      "loss": 0.5286,
+      "step": 36840
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1440786660966224e-05,
+      "loss": 0.5526,
+      "step": 36850
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1419410004275335e-05,
+      "loss": 0.5133,
+      "step": 36860
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.139803334758444e-05,
+      "loss": 0.5688,
+      "step": 36870
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1376656690893544e-05,
+      "loss": 0.5229,
+      "step": 36880
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1355280034202652e-05,
+      "loss": 0.523,
+      "step": 36890
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1333903377511757e-05,
+      "loss": 0.5482,
+      "step": 36900
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.1312526720820865e-05,
+      "loss": 0.5119,
+      "step": 36910
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1291150064129973e-05,
+      "loss": 0.5647,
+      "step": 36920
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1269773407439077e-05,
+      "loss": 0.5249,
+      "step": 36930
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1248396750748185e-05,
+      "loss": 0.5304,
+      "step": 36940
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.122702009405729e-05,
+      "loss": 0.5616,
+      "step": 36950
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1205643437366397e-05,
+      "loss": 0.5084,
+      "step": 36960
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1184266780675505e-05,
+      "loss": 0.5712,
+      "step": 36970
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.116289012398461e-05,
+      "loss": 0.5205,
+      "step": 36980
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1141513467293718e-05,
+      "loss": 0.5253,
+      "step": 36990
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1120136810602822e-05,
+      "loss": 0.5599,
+      "step": 37000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1098760153911927e-05,
+      "loss": 0.4978,
+      "step": 37010
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1077383497221035e-05,
+      "loss": 0.572,
+      "step": 37020
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1056006840530143e-05,
+      "loss": 0.5189,
+      "step": 37030
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.103463018383925e-05,
+      "loss": 0.5354,
+      "step": 37040
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.1013253527148355e-05,
+      "loss": 0.5606,
+      "step": 37050
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.099187687045746e-05,
+      "loss": 0.5094,
+      "step": 37060
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0970500213766567e-05,
+      "loss": 0.5725,
+      "step": 37070
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0949123557075675e-05,
+      "loss": 0.518,
+      "step": 37080
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.092774690038478e-05,
+      "loss": 0.5284,
+      "step": 37090
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0906370243693888e-05,
+      "loss": 0.5517,
+      "step": 37100
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0884993587002992e-05,
+      "loss": 0.5121,
+      "step": 37110
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.08636169303121e-05,
+      "loss": 0.5735,
+      "step": 37120
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0842240273621208e-05,
+      "loss": 0.5225,
+      "step": 37130
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.0820863616930313e-05,
+      "loss": 0.5281,
+      "step": 37140
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.079948696023942e-05,
+      "loss": 0.562,
+      "step": 37150
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0778110303548525e-05,
+      "loss": 0.5082,
+      "step": 37160
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0756733646857633e-05,
+      "loss": 0.564,
+      "step": 37170
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0735356990166737e-05,
+      "loss": 0.5248,
+      "step": 37180
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0713980333475845e-05,
+      "loss": 0.52,
+      "step": 37190
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0692603676784953e-05,
+      "loss": 0.5596,
+      "step": 37200
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0671227020094058e-05,
+      "loss": 0.5103,
+      "step": 37210
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0649850363403166e-05,
+      "loss": 0.5729,
+      "step": 37220
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.062847370671227e-05,
+      "loss": 0.5222,
+      "step": 37230
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0607097050021378e-05,
+      "loss": 0.5309,
+      "step": 37240
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0585720393330486e-05,
+      "loss": 0.5519,
+      "step": 37250
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.056434373663959e-05,
+      "loss": 0.5073,
+      "step": 37260
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0542967079948698e-05,
+      "loss": 0.5591,
+      "step": 37270
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0521590423257803e-05,
+      "loss": 0.5271,
+      "step": 37280
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0500213766566907e-05,
+      "loss": 0.5256,
+      "step": 37290
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.047883710987602e-05,
+      "loss": 0.5645,
+      "step": 37300
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0457460453185123e-05,
+      "loss": 0.517,
+      "step": 37310
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0436083796494228e-05,
+      "loss": 0.5666,
+      "step": 37320
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0414707139803336e-05,
+      "loss": 0.5096,
+      "step": 37330
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.039333048311244e-05,
+      "loss": 0.5355,
+      "step": 37340
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.037195382642155e-05,
+      "loss": 0.5594,
+      "step": 37350
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0350577169730656e-05,
+      "loss": 0.503,
+      "step": 37360
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.032920051303976e-05,
+      "loss": 0.5716,
+      "step": 37370
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.0307823856348868e-05,
+      "loss": 0.5254,
+      "step": 37380
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0286447199657973e-05,
+      "loss": 0.5338,
+      "step": 37390
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.026507054296708e-05,
+      "loss": 0.557,
+      "step": 37400
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.024369388627619e-05,
+      "loss": 0.5114,
+      "step": 37410
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0222317229585293e-05,
+      "loss": 0.5682,
+      "step": 37420
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.02009405728944e-05,
+      "loss": 0.5182,
+      "step": 37430
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0179563916203505e-05,
+      "loss": 0.5255,
+      "step": 37440
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0158187259512613e-05,
+      "loss": 0.5603,
+      "step": 37450
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.013681060282172e-05,
+      "loss": 0.5068,
+      "step": 37460
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0115433946130826e-05,
+      "loss": 0.5704,
+      "step": 37470
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0094057289439934e-05,
+      "loss": 0.5153,
+      "step": 37480
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0072680632749038e-05,
+      "loss": 0.5272,
+      "step": 37490
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0051303976058146e-05,
+      "loss": 0.5513,
+      "step": 37500
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.0029927319367254e-05,
+      "loss": 0.5076,
+      "step": 37510
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.000855066267636e-05,
+      "loss": 0.5853,
+      "step": 37520
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9987174005985466e-05,
+      "loss": 0.5215,
+      "step": 37530
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.996579734929457e-05,
+      "loss": 0.5277,
+      "step": 37540
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9944420692603675e-05,
+      "loss": 0.5589,
+      "step": 37550
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9923044035912783e-05,
+      "loss": 0.5087,
+      "step": 37560
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.990166737922189e-05,
+      "loss": 0.5612,
+      "step": 37570
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9880290722531e-05,
+      "loss": 0.5153,
+      "step": 37580
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9858914065840104e-05,
+      "loss": 0.5333,
+      "step": 37590
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9837537409149208e-05,
+      "loss": 0.5608,
+      "step": 37600
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9816160752458316e-05,
+      "loss": 0.5017,
+      "step": 37610
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9794784095767424e-05,
+      "loss": 0.5673,
+      "step": 37620
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.977340743907653e-05,
+      "loss": 0.5236,
+      "step": 37630
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9752030782385636e-05,
+      "loss": 0.5221,
+      "step": 37640
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.973065412569474e-05,
+      "loss": 0.5548,
+      "step": 37650
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.970927746900385e-05,
+      "loss": 0.5003,
+      "step": 37660
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9687900812312957e-05,
+      "loss": 0.5672,
+      "step": 37670
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.966652415562206e-05,
+      "loss": 0.5219,
+      "step": 37680
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.964514749893117e-05,
+      "loss": 0.5354,
+      "step": 37690
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9623770842240274e-05,
+      "loss": 0.5589,
+      "step": 37700
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.960239418554938e-05,
+      "loss": 0.5161,
+      "step": 37710
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9581017528858486e-05,
+      "loss": 0.5671,
+      "step": 37720
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9559640872167594e-05,
+      "loss": 0.5205,
+      "step": 37730
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9538264215476702e-05,
+      "loss": 0.5299,
+      "step": 37740
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9516887558785806e-05,
+      "loss": 0.5636,
+      "step": 37750
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9495510902094914e-05,
+      "loss": 0.5004,
+      "step": 37760
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.947413424540402e-05,
+      "loss": 0.5564,
+      "step": 37770
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9452757588713127e-05,
+      "loss": 0.5259,
+      "step": 37780
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9431380932022234e-05,
+      "loss": 0.5273,
+      "step": 37790
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.941000427533134e-05,
+      "loss": 0.5627,
+      "step": 37800
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9388627618640447e-05,
+      "loss": 0.4996,
+      "step": 37810
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.936725096194955e-05,
+      "loss": 0.5716,
+      "step": 37820
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9345874305258656e-05,
+      "loss": 0.5161,
+      "step": 37830
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9324497648567767e-05,
+      "loss": 0.5245,
+      "step": 37840
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.9303120991876872e-05,
+      "loss": 0.5642,
+      "step": 37850
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9281744335185976e-05,
+      "loss": 0.5128,
+      "step": 37860
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9260367678495084e-05,
+      "loss": 0.5704,
+      "step": 37870
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.923899102180419e-05,
+      "loss": 0.526,
+      "step": 37880
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.92176143651133e-05,
+      "loss": 0.5337,
+      "step": 37890
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9196237708422404e-05,
+      "loss": 0.5522,
+      "step": 37900
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.917486105173151e-05,
+      "loss": 0.508,
+      "step": 37910
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9153484395040617e-05,
+      "loss": 0.5649,
+      "step": 37920
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.913210773834972e-05,
+      "loss": 0.5195,
+      "step": 37930
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.911073108165883e-05,
+      "loss": 0.5259,
+      "step": 37940
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9089354424967937e-05,
+      "loss": 0.5615,
+      "step": 37950
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.906797776827704e-05,
+      "loss": 0.5151,
+      "step": 37960
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.904660111158615e-05,
+      "loss": 0.5667,
+      "step": 37970
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9025224454895254e-05,
+      "loss": 0.5242,
+      "step": 37980
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.9003847798204362e-05,
+      "loss": 0.5269,
+      "step": 37990
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.898247114151347e-05,
+      "loss": 0.549,
+      "step": 38000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8961094484822574e-05,
+      "loss": 0.5042,
+      "step": 38010
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8939717828131682e-05,
+      "loss": 0.5665,
+      "step": 38020
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8918341171440787e-05,
+      "loss": 0.5068,
+      "step": 38030
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8896964514749895e-05,
+      "loss": 0.5193,
+      "step": 38040
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8875587858059003e-05,
+      "loss": 0.5512,
+      "step": 38050
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8854211201368107e-05,
+      "loss": 0.5185,
+      "step": 38060
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8832834544677215e-05,
+      "loss": 0.5581,
+      "step": 38070
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.881145788798632e-05,
+      "loss": 0.5225,
+      "step": 38080
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.8790081231295424e-05,
+      "loss": 0.5277,
+      "step": 38090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8768704574604532e-05,
+      "loss": 0.5649,
+      "step": 38100
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.874732791791364e-05,
+      "loss": 0.4986,
+      "step": 38110
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8725951261222748e-05,
+      "loss": 0.5674,
+      "step": 38120
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8704574604531852e-05,
+      "loss": 0.5197,
+      "step": 38130
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8683197947840957e-05,
+      "loss": 0.5308,
+      "step": 38140
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8661821291150065e-05,
+      "loss": 0.5578,
+      "step": 38150
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8640444634459173e-05,
+      "loss": 0.5063,
+      "step": 38160
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8619067977768277e-05,
+      "loss": 0.5711,
+      "step": 38170
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8597691321077385e-05,
+      "loss": 0.5165,
+      "step": 38180
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.857631466438649e-05,
+      "loss": 0.5249,
+      "step": 38190
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8554938007695597e-05,
+      "loss": 0.5477,
+      "step": 38200
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8533561351004702e-05,
+      "loss": 0.51,
+      "step": 38210
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.851218469431381e-05,
+      "loss": 0.5727,
+      "step": 38220
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8490808037622918e-05,
+      "loss": 0.513,
+      "step": 38230
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8469431380932022e-05,
+      "loss": 0.5305,
+      "step": 38240
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.844805472424113e-05,
+      "loss": 0.5605,
+      "step": 38250
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8426678067550235e-05,
+      "loss": 0.5104,
+      "step": 38260
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8405301410859342e-05,
+      "loss": 0.5694,
+      "step": 38270
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.838392475416845e-05,
+      "loss": 0.5148,
+      "step": 38280
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8362548097477555e-05,
+      "loss": 0.5296,
+      "step": 38290
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8341171440786663e-05,
+      "loss": 0.5514,
+      "step": 38300
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8319794784095767e-05,
+      "loss": 0.508,
+      "step": 38310
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.8298418127404875e-05,
+      "loss": 0.5728,
+      "step": 38320
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8277041470713983e-05,
+      "loss": 0.5156,
+      "step": 38330
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8255664814023088e-05,
+      "loss": 0.5198,
+      "step": 38340
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8234288157332196e-05,
+      "loss": 0.5434,
+      "step": 38350
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.82129115006413e-05,
+      "loss": 0.5131,
+      "step": 38360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8191534843950405e-05,
+      "loss": 0.568,
+      "step": 38370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8170158187259516e-05,
+      "loss": 0.5121,
+      "step": 38380
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.814878153056862e-05,
+      "loss": 0.5255,
+      "step": 38390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8127404873877725e-05,
+      "loss": 0.5673,
+      "step": 38400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8106028217186833e-05,
+      "loss": 0.5118,
+      "step": 38410
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8084651560495937e-05,
+      "loss": 0.5715,
+      "step": 38420
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.806327490380505e-05,
+      "loss": 0.5241,
+      "step": 38430
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8041898247114153e-05,
+      "loss": 0.5356,
+      "step": 38440
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.8020521590423258e-05,
+      "loss": 0.5496,
+      "step": 38450
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7999144933732365e-05,
+      "loss": 0.5135,
+      "step": 38460
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.797776827704147e-05,
+      "loss": 0.5605,
+      "step": 38470
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7956391620350578e-05,
+      "loss": 0.5171,
+      "step": 38480
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7935014963659686e-05,
+      "loss": 0.5216,
+      "step": 38490
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.791363830696879e-05,
+      "loss": 0.5654,
+      "step": 38500
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7892261650277898e-05,
+      "loss": 0.5102,
+      "step": 38510
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7870884993587003e-05,
+      "loss": 0.5682,
+      "step": 38520
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.784950833689611e-05,
+      "loss": 0.5161,
+      "step": 38530
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.782813168020522e-05,
+      "loss": 0.5299,
+      "step": 38540
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.7806755023514323e-05,
+      "loss": 0.5659,
+      "step": 38550
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.778537836682343e-05,
+      "loss": 0.5038,
+      "step": 38560
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7764001710132535e-05,
+      "loss": 0.5591,
+      "step": 38570
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7742625053441643e-05,
+      "loss": 0.5197,
+      "step": 38580
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.772124839675075e-05,
+      "loss": 0.5212,
+      "step": 38590
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7699871740059856e-05,
+      "loss": 0.5545,
+      "step": 38600
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7678495083368964e-05,
+      "loss": 0.5122,
+      "step": 38610
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7657118426678068e-05,
+      "loss": 0.5693,
+      "step": 38620
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7635741769987173e-05,
+      "loss": 0.5204,
+      "step": 38630
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.761436511329628e-05,
+      "loss": 0.5229,
+      "step": 38640
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.759298845660539e-05,
+      "loss": 0.5528,
+      "step": 38650
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7571611799914496e-05,
+      "loss": 0.5099,
+      "step": 38660
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.75502351432236e-05,
+      "loss": 0.5809,
+      "step": 38670
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7528858486532705e-05,
+      "loss": 0.5174,
+      "step": 38680
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7507481829841813e-05,
+      "loss": 0.5295,
+      "step": 38690
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.748610517315092e-05,
+      "loss": 0.5602,
+      "step": 38700
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7464728516460026e-05,
+      "loss": 0.5008,
+      "step": 38710
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7443351859769134e-05,
+      "loss": 0.5734,
+      "step": 38720
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7421975203078238e-05,
+      "loss": 0.5235,
+      "step": 38730
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7400598546387346e-05,
+      "loss": 0.5313,
+      "step": 38740
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.737922188969645e-05,
+      "loss": 0.56,
+      "step": 38750
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.735784523300556e-05,
+      "loss": 0.5049,
+      "step": 38760
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.7336468576314666e-05,
+      "loss": 0.5697,
+      "step": 38770
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.731509191962377e-05,
+      "loss": 0.52,
+      "step": 38780
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.729371526293288e-05,
+      "loss": 0.5214,
+      "step": 38790
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7272338606241983e-05,
+      "loss": 0.5637,
+      "step": 38800
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.725096194955109e-05,
+      "loss": 0.5022,
+      "step": 38810
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.72295852928602e-05,
+      "loss": 0.5754,
+      "step": 38820
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7208208636169303e-05,
+      "loss": 0.518,
+      "step": 38830
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.718683197947841e-05,
+      "loss": 0.5312,
+      "step": 38840
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7165455322787516e-05,
+      "loss": 0.5594,
+      "step": 38850
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7144078666096624e-05,
+      "loss": 0.5084,
+      "step": 38860
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7122702009405732e-05,
+      "loss": 0.5726,
+      "step": 38870
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7101325352714836e-05,
+      "loss": 0.5136,
+      "step": 38880
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7079948696023944e-05,
+      "loss": 0.5292,
+      "step": 38890
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.705857203933305e-05,
+      "loss": 0.5546,
+      "step": 38900
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7037195382642153e-05,
+      "loss": 0.5019,
+      "step": 38910
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.7015818725951264e-05,
+      "loss": 0.566,
+      "step": 38920
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.699444206926037e-05,
+      "loss": 0.5118,
+      "step": 38930
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6973065412569473e-05,
+      "loss": 0.5285,
+      "step": 38940
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.695168875587858e-05,
+      "loss": 0.558,
+      "step": 38950
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6930312099187686e-05,
+      "loss": 0.4989,
+      "step": 38960
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6908935442496797e-05,
+      "loss": 0.5581,
+      "step": 38970
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.68875587858059e-05,
+      "loss": 0.5153,
+      "step": 38980
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6866182129115006e-05,
+      "loss": 0.53,
+      "step": 38990
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6844805472424114e-05,
+      "loss": 0.5501,
+      "step": 39000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.682342881573322e-05,
+      "loss": 0.5112,
+      "step": 39010
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6802052159042326e-05,
+      "loss": 0.5584,
+      "step": 39020
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6780675502351434e-05,
+      "loss": 0.5208,
+      "step": 39030
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.675929884566054e-05,
+      "loss": 0.5123,
+      "step": 39040
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6737922188969647e-05,
+      "loss": 0.5547,
+      "step": 39050
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.671654553227875e-05,
+      "loss": 0.5048,
+      "step": 39060
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.669516887558786e-05,
+      "loss": 0.5687,
+      "step": 39070
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6673792218896967e-05,
+      "loss": 0.5187,
+      "step": 39080
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.665241556220607e-05,
+      "loss": 0.5316,
+      "step": 39090
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.663103890551518e-05,
+      "loss": 0.5485,
+      "step": 39100
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6609662248824284e-05,
+      "loss": 0.5114,
+      "step": 39110
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.658828559213339e-05,
+      "loss": 0.5725,
+      "step": 39120
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.65669089354425e-05,
+      "loss": 0.5265,
+      "step": 39130
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6545532278751604e-05,
+      "loss": 0.5188,
+      "step": 39140
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6524155622060712e-05,
+      "loss": 0.5499,
+      "step": 39150
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6502778965369817e-05,
+      "loss": 0.5027,
+      "step": 39160
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.648140230867892e-05,
+      "loss": 0.5696,
+      "step": 39170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.646002565198803e-05,
+      "loss": 0.5243,
+      "step": 39180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6438648995297137e-05,
+      "loss": 0.5253,
+      "step": 39190
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6417272338606245e-05,
+      "loss": 0.5628,
+      "step": 39200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.639589568191535e-05,
+      "loss": 0.5031,
+      "step": 39210
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6374519025224454e-05,
+      "loss": 0.5765,
+      "step": 39220
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6353142368533562e-05,
+      "loss": 0.5262,
+      "step": 39230
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.633176571184267e-05,
+      "loss": 0.5338,
+      "step": 39240
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6310389055151774e-05,
+      "loss": 0.5531,
+      "step": 39250
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6289012398460882e-05,
+      "loss": 0.5043,
+      "step": 39260
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6267635741769987e-05,
+      "loss": 0.57,
+      "step": 39270
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6246259085079095e-05,
+      "loss": 0.5234,
+      "step": 39280
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.62248824283882e-05,
+      "loss": 0.5198,
+      "step": 39290
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6203505771697307e-05,
+      "loss": 0.5573,
+      "step": 39300
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6182129115006415e-05,
+      "loss": 0.5148,
+      "step": 39310
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.616075245831552e-05,
+      "loss": 0.5646,
+      "step": 39320
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6139375801624627e-05,
+      "loss": 0.5127,
+      "step": 39330
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6117999144933732e-05,
+      "loss": 0.5251,
+      "step": 39340
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.609662248824284e-05,
+      "loss": 0.5552,
+      "step": 39350
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6075245831551948e-05,
+      "loss": 0.5125,
+      "step": 39360
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6053869174861052e-05,
+      "loss": 0.5655,
+      "step": 39370
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.603249251817016e-05,
+      "loss": 0.5103,
+      "step": 39380
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.6011115861479265e-05,
+      "loss": 0.5161,
+      "step": 39390
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5989739204788372e-05,
+      "loss": 0.5525,
+      "step": 39400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.596836254809748e-05,
+      "loss": 0.502,
+      "step": 39410
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5946985891406585e-05,
+      "loss": 0.5731,
+      "step": 39420
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.592560923471569e-05,
+      "loss": 0.5189,
+      "step": 39430
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5904232578024797e-05,
+      "loss": 0.5223,
+      "step": 39440
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5882855921333902e-05,
+      "loss": 0.5464,
+      "step": 39450
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5861479264643013e-05,
+      "loss": 0.5001,
+      "step": 39460
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5840102607952118e-05,
+      "loss": 0.5768,
+      "step": 39470
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.5818725951261222e-05,
+      "loss": 0.525,
+      "step": 39480
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.579734929457033e-05,
+      "loss": 0.5232,
+      "step": 39490
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5775972637879434e-05,
+      "loss": 0.5565,
+      "step": 39500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5754595981188546e-05,
+      "loss": 0.5087,
+      "step": 39510
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.573321932449765e-05,
+      "loss": 0.5669,
+      "step": 39520
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5711842667806755e-05,
+      "loss": 0.5226,
+      "step": 39530
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5690466011115863e-05,
+      "loss": 0.5239,
+      "step": 39540
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5669089354424967e-05,
+      "loss": 0.5559,
+      "step": 39550
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5647712697734075e-05,
+      "loss": 0.5064,
+      "step": 39560
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5626336041043183e-05,
+      "loss": 0.5658,
+      "step": 39570
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5604959384352287e-05,
+      "loss": 0.5179,
+      "step": 39580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5583582727661395e-05,
+      "loss": 0.5338,
+      "step": 39590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.55622060709705e-05,
+      "loss": 0.5614,
+      "step": 39600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5540829414279608e-05,
+      "loss": 0.514,
+      "step": 39610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5519452757588716e-05,
+      "loss": 0.5565,
+      "step": 39620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.549807610089782e-05,
+      "loss": 0.5097,
+      "step": 39630
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5476699444206928e-05,
+      "loss": 0.521,
+      "step": 39640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5455322787516033e-05,
+      "loss": 0.5607,
+      "step": 39650
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5433946130825137e-05,
+      "loss": 0.5021,
+      "step": 39660
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.541256947413425e-05,
+      "loss": 0.5605,
+      "step": 39670
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5391192817443353e-05,
+      "loss": 0.515,
+      "step": 39680
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.536981616075246e-05,
+      "loss": 0.5225,
+      "step": 39690
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5348439504061565e-05,
+      "loss": 0.5504,
+      "step": 39700
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.532706284737067e-05,
+      "loss": 0.5155,
+      "step": 39710
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5305686190679778e-05,
+      "loss": 0.5586,
+      "step": 39720
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.5284309533988886e-05,
+      "loss": 0.5223,
+      "step": 39730
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.526293287729799e-05,
+      "loss": 0.5211,
+      "step": 39740
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5241556220607098e-05,
+      "loss": 0.5536,
+      "step": 39750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5220179563916204e-05,
+      "loss": 0.5054,
+      "step": 39760
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.519880290722531e-05,
+      "loss": 0.5726,
+      "step": 39770
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5177426250534418e-05,
+      "loss": 0.5204,
+      "step": 39780
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5156049593843525e-05,
+      "loss": 0.5305,
+      "step": 39790
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.513467293715263e-05,
+      "loss": 0.5532,
+      "step": 39800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5113296280461737e-05,
+      "loss": 0.5084,
+      "step": 39810
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5091919623770841e-05,
+      "loss": 0.5689,
+      "step": 39820
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5070542967079948e-05,
+      "loss": 0.5179,
+      "step": 39830
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5049166310389057e-05,
+      "loss": 0.5303,
+      "step": 39840
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5027789653698163e-05,
+      "loss": 0.5508,
+      "step": 39850
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.5006412997007268e-05,
+      "loss": 0.4999,
+      "step": 39860
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4985036340316374e-05,
+      "loss": 0.5794,
+      "step": 39870
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.496365968362548e-05,
+      "loss": 0.5159,
+      "step": 39880
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.494228302693459e-05,
+      "loss": 0.5263,
+      "step": 39890
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4920906370243695e-05,
+      "loss": 0.5502,
+      "step": 39900
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.48995297135528e-05,
+      "loss": 0.5047,
+      "step": 39910
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4878153056861907e-05,
+      "loss": 0.5581,
+      "step": 39920
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4856776400171013e-05,
+      "loss": 0.5193,
+      "step": 39930
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4835399743480121e-05,
+      "loss": 0.5291,
+      "step": 39940
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4814023086789227e-05,
+      "loss": 0.5464,
+      "step": 39950
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.4792646430098333e-05,
+      "loss": 0.4991,
+      "step": 39960
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.477126977340744e-05,
+      "loss": 0.5594,
+      "step": 39970
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4749893116716546e-05,
+      "loss": 0.5084,
+      "step": 39980
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4728516460025652e-05,
+      "loss": 0.5331,
+      "step": 39990
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.470713980333476e-05,
+      "loss": 0.5491,
+      "step": 40000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4685763146643866e-05,
+      "loss": 0.5045,
+      "step": 40010
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4664386489952972e-05,
+      "loss": 0.5638,
+      "step": 40020
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4643009833262079e-05,
+      "loss": 0.5124,
+      "step": 40030
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4621633176571183e-05,
+      "loss": 0.525,
+      "step": 40040
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4600256519880293e-05,
+      "loss": 0.555,
+      "step": 40050
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4578879863189399e-05,
+      "loss": 0.5053,
+      "step": 40060
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4557503206498505e-05,
+      "loss": 0.5617,
+      "step": 40070
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4536126549807611e-05,
+      "loss": 0.53,
+      "step": 40080
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4514749893116716e-05,
+      "loss": 0.5287,
+      "step": 40090
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4493373236425822e-05,
+      "loss": 0.5579,
+      "step": 40100
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4471996579734932e-05,
+      "loss": 0.4981,
+      "step": 40110
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4450619923044038e-05,
+      "loss": 0.5615,
+      "step": 40120
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4429243266353142e-05,
+      "loss": 0.5165,
+      "step": 40130
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4407866609662249e-05,
+      "loss": 0.5177,
+      "step": 40140
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4386489952971355e-05,
+      "loss": 0.5567,
+      "step": 40150
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4365113296280464e-05,
+      "loss": 0.4999,
+      "step": 40160
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4343736639589569e-05,
+      "loss": 0.5627,
+      "step": 40170
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4322359982898675e-05,
+      "loss": 0.5102,
+      "step": 40180
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4300983326207781e-05,
+      "loss": 0.5206,
+      "step": 40190
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.4279606669516887e-05,
+      "loss": 0.5538,
+      "step": 40200
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4258230012825995e-05,
+      "loss": 0.509,
+      "step": 40210
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4236853356135102e-05,
+      "loss": 0.5642,
+      "step": 40220
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4215476699444208e-05,
+      "loss": 0.5163,
+      "step": 40230
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4194100042753314e-05,
+      "loss": 0.5198,
+      "step": 40240
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.417272338606242e-05,
+      "loss": 0.5468,
+      "step": 40250
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4151346729371526e-05,
+      "loss": 0.502,
+      "step": 40260
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4129970072680634e-05,
+      "loss": 0.5645,
+      "step": 40270
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.410859341598974e-05,
+      "loss": 0.5204,
+      "step": 40280
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4087216759298847e-05,
+      "loss": 0.5252,
+      "step": 40290
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4065840102607953e-05,
+      "loss": 0.5529,
+      "step": 40300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4044463445917057e-05,
+      "loss": 0.5043,
+      "step": 40310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4023086789226167e-05,
+      "loss": 0.5661,
+      "step": 40320
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.4001710132535273e-05,
+      "loss": 0.5141,
+      "step": 40330
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.398033347584438e-05,
+      "loss": 0.5242,
+      "step": 40340
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3958956819153484e-05,
+      "loss": 0.5575,
+      "step": 40350
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.393758016246259e-05,
+      "loss": 0.4986,
+      "step": 40360
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3916203505771696e-05,
+      "loss": 0.5641,
+      "step": 40370
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3894826849080806e-05,
+      "loss": 0.5157,
+      "step": 40380
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3873450192389912e-05,
+      "loss": 0.5217,
+      "step": 40390
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3852073535699017e-05,
+      "loss": 0.5534,
+      "step": 40400
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3830696879008123e-05,
+      "loss": 0.5028,
+      "step": 40410
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3809320222317229e-05,
+      "loss": 0.5713,
+      "step": 40420
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.3787943565626339e-05,
+      "loss": 0.5282,
+      "step": 40430
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3766566908935443e-05,
+      "loss": 0.5118,
+      "step": 40440
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.374519025224455e-05,
+      "loss": 0.5586,
+      "step": 40450
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3723813595553656e-05,
+      "loss": 0.5132,
+      "step": 40460
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3702436938862762e-05,
+      "loss": 0.5677,
+      "step": 40470
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.368106028217187e-05,
+      "loss": 0.5162,
+      "step": 40480
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3659683625480976e-05,
+      "loss": 0.5248,
+      "step": 40490
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3638306968790082e-05,
+      "loss": 0.5554,
+      "step": 40500
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3616930312099188e-05,
+      "loss": 0.5069,
+      "step": 40510
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3595553655408294e-05,
+      "loss": 0.5715,
+      "step": 40520
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.35741769987174e-05,
+      "loss": 0.5124,
+      "step": 40530
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3552800342026509e-05,
+      "loss": 0.5219,
+      "step": 40540
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3531423685335615e-05,
+      "loss": 0.5411,
+      "step": 40550
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3510047028644721e-05,
+      "loss": 0.5052,
+      "step": 40560
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3488670371953827e-05,
+      "loss": 0.5621,
+      "step": 40570
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3467293715262932e-05,
+      "loss": 0.5124,
+      "step": 40580
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3445917058572041e-05,
+      "loss": 0.5134,
+      "step": 40590
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3424540401881147e-05,
+      "loss": 0.5494,
+      "step": 40600
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3403163745190254e-05,
+      "loss": 0.5032,
+      "step": 40610
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3381787088499358e-05,
+      "loss": 0.566,
+      "step": 40620
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3360410431808464e-05,
+      "loss": 0.5154,
+      "step": 40630
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.333903377511757e-05,
+      "loss": 0.5252,
+      "step": 40640
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.331765711842668e-05,
+      "loss": 0.556,
+      "step": 40650
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.3296280461735785e-05,
+      "loss": 0.5067,
+      "step": 40660
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3274903805044891e-05,
+      "loss": 0.5683,
+      "step": 40670
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3253527148353997e-05,
+      "loss": 0.5116,
+      "step": 40680
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3232150491663103e-05,
+      "loss": 0.5171,
+      "step": 40690
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3210773834972213e-05,
+      "loss": 0.5522,
+      "step": 40700
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3189397178281317e-05,
+      "loss": 0.5101,
+      "step": 40710
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3168020521590424e-05,
+      "loss": 0.5683,
+      "step": 40720
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.314664386489953e-05,
+      "loss": 0.5176,
+      "step": 40730
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3125267208208636e-05,
+      "loss": 0.5228,
+      "step": 40740
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3103890551517744e-05,
+      "loss": 0.5439,
+      "step": 40750
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.308251389482685e-05,
+      "loss": 0.507,
+      "step": 40760
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3061137238135956e-05,
+      "loss": 0.5665,
+      "step": 40770
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3039760581445063e-05,
+      "loss": 0.5112,
+      "step": 40780
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.3018383924754169e-05,
+      "loss": 0.5214,
+      "step": 40790
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2997007268063275e-05,
+      "loss": 0.5534,
+      "step": 40800
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2975630611372383e-05,
+      "loss": 0.5018,
+      "step": 40810
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2954253954681489e-05,
+      "loss": 0.5679,
+      "step": 40820
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2932877297990595e-05,
+      "loss": 0.5183,
+      "step": 40830
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2911500641299701e-05,
+      "loss": 0.5148,
+      "step": 40840
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2890123984608806e-05,
+      "loss": 0.5451,
+      "step": 40850
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2868747327917916e-05,
+      "loss": 0.5003,
+      "step": 40860
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2847370671227022e-05,
+      "loss": 0.5627,
+      "step": 40870
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2825994014536128e-05,
+      "loss": 0.5117,
+      "step": 40880
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2804617357845233e-05,
+      "loss": 0.5263,
+      "step": 40890
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2783240701154339e-05,
+      "loss": 0.5484,
+      "step": 40900
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2761864044463445e-05,
+      "loss": 0.5056,
+      "step": 40910
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2740487387772555e-05,
+      "loss": 0.558,
+      "step": 40920
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2719110731081659e-05,
+      "loss": 0.5204,
+      "step": 40930
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2697734074390765e-05,
+      "loss": 0.5221,
+      "step": 40940
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2676357417699871e-05,
+      "loss": 0.5537,
+      "step": 40950
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2654980761008978e-05,
+      "loss": 0.5114,
+      "step": 40960
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2633604104318086e-05,
+      "loss": 0.5598,
+      "step": 40970
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2612227447627192e-05,
+      "loss": 0.5174,
+      "step": 40980
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2590850790936298e-05,
+      "loss": 0.5196,
+      "step": 40990
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2569474134245404e-05,
+      "loss": 0.5483,
+      "step": 41000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.254809747755451e-05,
+      "loss": 0.5085,
+      "step": 41010
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2526720820863618e-05,
+      "loss": 0.5646,
+      "step": 41020
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2505344164172724e-05,
+      "loss": 0.5127,
+      "step": 41030
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.248396750748183e-05,
+      "loss": 0.5171,
+      "step": 41040
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2462590850790937e-05,
+      "loss": 0.5522,
+      "step": 41050
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2441214194100045e-05,
+      "loss": 0.5076,
+      "step": 41060
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.241983753740915e-05,
+      "loss": 0.5602,
+      "step": 41070
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2398460880718255e-05,
+      "loss": 0.513,
+      "step": 41080
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2377084224027363e-05,
+      "loss": 0.5224,
+      "step": 41090
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.235570756733647e-05,
+      "loss": 0.5429,
+      "step": 41100
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2334330910645576e-05,
+      "loss": 0.5042,
+      "step": 41110
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2312954253954682e-05,
+      "loss": 0.5733,
+      "step": 41120
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.2291577597263788e-05,
+      "loss": 0.5147,
+      "step": 41130
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2270200940572894e-05,
+      "loss": 0.5268,
+      "step": 41140
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2248824283882002e-05,
+      "loss": 0.5561,
+      "step": 41150
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2227447627191107e-05,
+      "loss": 0.5125,
+      "step": 41160
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2206070970500215e-05,
+      "loss": 0.5639,
+      "step": 41170
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2184694313809321e-05,
+      "loss": 0.5106,
+      "step": 41180
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2163317657118427e-05,
+      "loss": 0.5236,
+      "step": 41190
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2141941000427533e-05,
+      "loss": 0.5443,
+      "step": 41200
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.212056434373664e-05,
+      "loss": 0.5085,
+      "step": 41210
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2099187687045746e-05,
+      "loss": 0.5681,
+      "step": 41220
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2077811030354854e-05,
+      "loss": 0.5077,
+      "step": 41230
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.205643437366396e-05,
+      "loss": 0.5215,
+      "step": 41240
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2035057716973066e-05,
+      "loss": 0.5513,
+      "step": 41250
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.2013681060282172e-05,
+      "loss": 0.5034,
+      "step": 41260
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1992304403591278e-05,
+      "loss": 0.5688,
+      "step": 41270
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1970927746900386e-05,
+      "loss": 0.5171,
+      "step": 41280
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1949551090209493e-05,
+      "loss": 0.525,
+      "step": 41290
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1928174433518597e-05,
+      "loss": 0.5471,
+      "step": 41300
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1906797776827705e-05,
+      "loss": 0.5052,
+      "step": 41310
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1885421120136811e-05,
+      "loss": 0.5683,
+      "step": 41320
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1864044463445919e-05,
+      "loss": 0.5166,
+      "step": 41330
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1842667806755024e-05,
+      "loss": 0.5127,
+      "step": 41340
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.182129115006413e-05,
+      "loss": 0.5636,
+      "step": 41350
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1799914493373238e-05,
+      "loss": 0.5107,
+      "step": 41360
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.1778537836682344e-05,
+      "loss": 0.5588,
+      "step": 41370
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.175716117999145e-05,
+      "loss": 0.5127,
+      "step": 41380
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1735784523300556e-05,
+      "loss": 0.5225,
+      "step": 41390
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1714407866609663e-05,
+      "loss": 0.5463,
+      "step": 41400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1693031209918769e-05,
+      "loss": 0.5081,
+      "step": 41410
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1671654553227877e-05,
+      "loss": 0.5612,
+      "step": 41420
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1650277896536981e-05,
+      "loss": 0.5154,
+      "step": 41430
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1628901239846089e-05,
+      "loss": 0.5222,
+      "step": 41440
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1607524583155195e-05,
+      "loss": 0.5473,
+      "step": 41450
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1586147926464301e-05,
+      "loss": 0.4975,
+      "step": 41460
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1564771269773408e-05,
+      "loss": 0.5628,
+      "step": 41470
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1543394613082514e-05,
+      "loss": 0.5163,
+      "step": 41480
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.152201795639162e-05,
+      "loss": 0.5241,
+      "step": 41490
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1500641299700728e-05,
+      "loss": 0.5412,
+      "step": 41500
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1479264643009834e-05,
+      "loss": 0.5064,
+      "step": 41510
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.145788798631894e-05,
+      "loss": 0.5605,
+      "step": 41520
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1436511329628047e-05,
+      "loss": 0.5161,
+      "step": 41530
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1415134672937153e-05,
+      "loss": 0.5293,
+      "step": 41540
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.139375801624626e-05,
+      "loss": 0.548,
+      "step": 41550
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1372381359555367e-05,
+      "loss": 0.505,
+      "step": 41560
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1351004702864471e-05,
+      "loss": 0.565,
+      "step": 41570
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.132962804617358e-05,
+      "loss": 0.5183,
+      "step": 41580
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1308251389482685e-05,
+      "loss": 0.5249,
+      "step": 41590
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1286874732791793e-05,
+      "loss": 0.5464,
+      "step": 41600
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1265498076100898e-05,
+      "loss": 0.5033,
+      "step": 41610
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1244121419410004e-05,
+      "loss": 0.5585,
+      "step": 41620
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1222744762719112e-05,
+      "loss": 0.52,
+      "step": 41630
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1201368106028218e-05,
+      "loss": 0.5215,
+      "step": 41640
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1179991449337324e-05,
+      "loss": 0.546,
+      "step": 41650
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.115861479264643e-05,
+      "loss": 0.5047,
+      "step": 41660
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1137238135955537e-05,
+      "loss": 0.5565,
+      "step": 41670
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1115861479264643e-05,
+      "loss": 0.5122,
+      "step": 41680
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1094484822573751e-05,
+      "loss": 0.5235,
+      "step": 41690
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1073108165882855e-05,
+      "loss": 0.5528,
+      "step": 41700
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1051731509191963e-05,
+      "loss": 0.5108,
+      "step": 41710
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.103035485250107e-05,
+      "loss": 0.5608,
+      "step": 41720
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.1008978195810176e-05,
+      "loss": 0.5076,
+      "step": 41730
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0987601539119282e-05,
+      "loss": 0.5264,
+      "step": 41740
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0966224882428388e-05,
+      "loss": 0.5519,
+      "step": 41750
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0944848225737494e-05,
+      "loss": 0.5026,
+      "step": 41760
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0923471569046602e-05,
+      "loss": 0.5612,
+      "step": 41770
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0902094912355708e-05,
+      "loss": 0.5182,
+      "step": 41780
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0880718255664815e-05,
+      "loss": 0.5251,
+      "step": 41790
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0859341598973921e-05,
+      "loss": 0.5528,
+      "step": 41800
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0837964942283027e-05,
+      "loss": 0.5013,
+      "step": 41810
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0816588285592135e-05,
+      "loss": 0.5561,
+      "step": 41820
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.079521162890124e-05,
+      "loss": 0.5142,
+      "step": 41830
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.0773834972210346e-05,
+      "loss": 0.5273,
+      "step": 41840
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0752458315519454e-05,
+      "loss": 0.5565,
+      "step": 41850
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.073108165882856e-05,
+      "loss": 0.5017,
+      "step": 41860
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0709705002137668e-05,
+      "loss": 0.5726,
+      "step": 41870
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0688328345446772e-05,
+      "loss": 0.5096,
+      "step": 41880
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0666951688755878e-05,
+      "loss": 0.5205,
+      "step": 41890
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0645575032064986e-05,
+      "loss": 0.5508,
+      "step": 41900
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0624198375374093e-05,
+      "loss": 0.5005,
+      "step": 41910
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0602821718683199e-05,
+      "loss": 0.5657,
+      "step": 41920
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0581445061992305e-05,
+      "loss": 0.5118,
+      "step": 41930
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0560068405301411e-05,
+      "loss": 0.5195,
+      "step": 41940
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0538691748610517e-05,
+      "loss": 0.5482,
+      "step": 41950
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0517315091919625e-05,
+      "loss": 0.4987,
+      "step": 41960
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.049593843522873e-05,
+      "loss": 0.5671,
+      "step": 41970
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0474561778537838e-05,
+      "loss": 0.5015,
+      "step": 41980
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0453185121846944e-05,
+      "loss": 0.5188,
+      "step": 41990
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.043180846515605e-05,
+      "loss": 0.5486,
+      "step": 42000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0410431808465156e-05,
+      "loss": 0.5005,
+      "step": 42010
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0389055151774262e-05,
+      "loss": 0.5609,
+      "step": 42020
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0367678495083369e-05,
+      "loss": 0.516,
+      "step": 42030
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0346301838392477e-05,
+      "loss": 0.5234,
+      "step": 42040
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0324925181701583e-05,
+      "loss": 0.5424,
+      "step": 42050
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0303548525010689e-05,
+      "loss": 0.4915,
+      "step": 42060
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.0282171868319795e-05,
+      "loss": 0.5652,
+      "step": 42070
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0260795211628901e-05,
+      "loss": 0.5156,
+      "step": 42080
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.023941855493801e-05,
+      "loss": 0.5266,
+      "step": 42090
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0218041898247114e-05,
+      "loss": 0.5479,
+      "step": 42100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.019666524155622e-05,
+      "loss": 0.5043,
+      "step": 42110
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0175288584865328e-05,
+      "loss": 0.5692,
+      "step": 42120
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0153911928174434e-05,
+      "loss": 0.5165,
+      "step": 42130
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.013253527148354e-05,
+      "loss": 0.5144,
+      "step": 42140
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0111158614792647e-05,
+      "loss": 0.5476,
+      "step": 42150
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0089781958101753e-05,
+      "loss": 0.5122,
+      "step": 42160
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.006840530141086e-05,
+      "loss": 0.5704,
+      "step": 42170
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0047028644719967e-05,
+      "loss": 0.5222,
+      "step": 42180
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.0025651988029073e-05,
+      "loss": 0.5131,
+      "step": 42190
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.000427533133818e-05,
+      "loss": 0.5537,
+      "step": 42200
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.982898674647285e-06,
+      "loss": 0.5093,
+      "step": 42210
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.961522017956392e-06,
+      "loss": 0.5591,
+      "step": 42220
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.9401453612655e-06,
+      "loss": 0.5137,
+      "step": 42230
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.918768704574604e-06,
+      "loss": 0.5159,
+      "step": 42240
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.897392047883712e-06,
+      "loss": 0.5404,
+      "step": 42250
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.876015391192818e-06,
+      "loss": 0.5019,
+      "step": 42260
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.854638734501924e-06,
+      "loss": 0.5679,
+      "step": 42270
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.83326207781103e-06,
+      "loss": 0.5113,
+      "step": 42280
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.811885421120137e-06,
+      "loss": 0.53,
+      "step": 42290
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.790508764429243e-06,
+      "loss": 0.5465,
+      "step": 42300
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.769132107738351e-06,
+      "loss": 0.493,
+      "step": 42310
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.747755451047457e-06,
+      "loss": 0.5612,
+      "step": 42320
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.726378794356563e-06,
+      "loss": 0.5223,
+      "step": 42330
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.70500213766567e-06,
+      "loss": 0.5253,
+      "step": 42340
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.683625480974776e-06,
+      "loss": 0.5391,
+      "step": 42350
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.662248824283884e-06,
+      "loss": 0.5063,
+      "step": 42360
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.640872167592988e-06,
+      "loss": 0.5604,
+      "step": 42370
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.619495510902094e-06,
+      "loss": 0.5145,
+      "step": 42380
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.598118854211202e-06,
+      "loss": 0.5214,
+      "step": 42390
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.576742197520308e-06,
+      "loss": 0.5511,
+      "step": 42400
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.555365540829415e-06,
+      "loss": 0.4929,
+      "step": 42410
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.53398888413852e-06,
+      "loss": 0.5647,
+      "step": 42420
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.512612227447627e-06,
+      "loss": 0.521,
+      "step": 42430
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.491235570756735e-06,
+      "loss": 0.5345,
+      "step": 42440
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.469858914065841e-06,
+      "loss": 0.5393,
+      "step": 42450
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.448482257374947e-06,
+      "loss": 0.5024,
+      "step": 42460
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.427105600684054e-06,
+      "loss": 0.5668,
+      "step": 42470
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.40572894399316e-06,
+      "loss": 0.516,
+      "step": 42480
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.384352287302266e-06,
+      "loss": 0.5233,
+      "step": 42490
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.362975630611374e-06,
+      "loss": 0.5451,
+      "step": 42500
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.341598973920478e-06,
+      "loss": 0.5121,
+      "step": 42510
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.320222317229586e-06,
+      "loss": 0.5625,
+      "step": 42520
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.298845660538692e-06,
+      "loss": 0.5146,
+      "step": 42530
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.277469003847799e-06,
+      "loss": 0.5248,
+      "step": 42540
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.256092347156905e-06,
+      "loss": 0.5458,
+      "step": 42550
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.234715690466011e-06,
+      "loss": 0.5039,
+      "step": 42560
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.213339033775117e-06,
+      "loss": 0.5661,
+      "step": 42570
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.191962377084225e-06,
+      "loss": 0.5119,
+      "step": 42580
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.170585720393331e-06,
+      "loss": 0.5182,
+      "step": 42590
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.149209063702438e-06,
+      "loss": 0.559,
+      "step": 42600
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.127832407011544e-06,
+      "loss": 0.5081,
+      "step": 42610
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.10645575032065e-06,
+      "loss": 0.5762,
+      "step": 42620
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.085079093629758e-06,
+      "loss": 0.5034,
+      "step": 42630
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.063702436938862e-06,
+      "loss": 0.5192,
+      "step": 42640
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.042325780247969e-06,
+      "loss": 0.5505,
+      "step": 42650
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 9.020949123557077e-06,
+      "loss": 0.5026,
+      "step": 42660
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.999572466866183e-06,
+      "loss": 0.5607,
+      "step": 42670
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.978195810175289e-06,
+      "loss": 0.5174,
+      "step": 42680
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.956819153484395e-06,
+      "loss": 0.5171,
+      "step": 42690
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.935442496793501e-06,
+      "loss": 0.5517,
+      "step": 42700
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.91406584010261e-06,
+      "loss": 0.5065,
+      "step": 42710
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.892689183411715e-06,
+      "loss": 0.5643,
+      "step": 42720
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.871312526720822e-06,
+      "loss": 0.5147,
+      "step": 42730
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.849935870029928e-06,
+      "loss": 0.5222,
+      "step": 42740
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.828559213339034e-06,
+      "loss": 0.5503,
+      "step": 42750
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.80718255664814e-06,
+      "loss": 0.4987,
+      "step": 42760
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.785805899957248e-06,
+      "loss": 0.5692,
+      "step": 42770
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.764429243266353e-06,
+      "loss": 0.52,
+      "step": 42780
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.74305258657546e-06,
+      "loss": 0.524,
+      "step": 42790
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.721675929884567e-06,
+      "loss": 0.5374,
+      "step": 42800
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.700299273193673e-06,
+      "loss": 0.4956,
+      "step": 42810
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.67892261650278e-06,
+      "loss": 0.559,
+      "step": 42820
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.657545959811885e-06,
+      "loss": 0.5066,
+      "step": 42830
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.636169303120992e-06,
+      "loss": 0.515,
+      "step": 42840
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.6147926464301e-06,
+      "loss": 0.5417,
+      "step": 42850
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.593415989739206e-06,
+      "loss": 0.4989,
+      "step": 42860
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.572039333048312e-06,
+      "loss": 0.557,
+      "step": 42870
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.550662676357418e-06,
+      "loss": 0.5143,
+      "step": 42880
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.529286019666524e-06,
+      "loss": 0.5236,
+      "step": 42890
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.507909362975632e-06,
+      "loss": 0.5459,
+      "step": 42900
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.486532706284737e-06,
+      "loss": 0.5086,
+      "step": 42910
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.465156049593843e-06,
+      "loss": 0.5672,
+      "step": 42920
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.44377939290295e-06,
+      "loss": 0.5162,
+      "step": 42930
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.422402736212057e-06,
+      "loss": 0.5181,
+      "step": 42940
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.401026079521163e-06,
+      "loss": 0.5462,
+      "step": 42950
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.37964942283027e-06,
+      "loss": 0.5016,
+      "step": 42960
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.358272766139376e-06,
+      "loss": 0.5667,
+      "step": 42970
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.336896109448484e-06,
+      "loss": 0.5093,
+      "step": 42980
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.31551945275759e-06,
+      "loss": 0.523,
+      "step": 42990
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.294142796066694e-06,
+      "loss": 0.5536,
+      "step": 43000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.272766139375802e-06,
+      "loss": 0.5025,
+      "step": 43010
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.251389482684908e-06,
+      "loss": 0.5581,
+      "step": 43020
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.230012825994015e-06,
+      "loss": 0.5187,
+      "step": 43030
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.208636169303122e-06,
+      "loss": 0.5189,
+      "step": 43040
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.187259512612227e-06,
+      "loss": 0.5461,
+      "step": 43050
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.165882855921335e-06,
+      "loss": 0.5043,
+      "step": 43060
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.144506199230441e-06,
+      "loss": 0.5663,
+      "step": 43070
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.123129542539547e-06,
+      "loss": 0.5221,
+      "step": 43080
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.101752885848653e-06,
+      "loss": 0.5205,
+      "step": 43090
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.08037622915776e-06,
+      "loss": 0.5404,
+      "step": 43100
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.058999572466866e-06,
+      "loss": 0.502,
+      "step": 43110
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.037622915775974e-06,
+      "loss": 0.5652,
+      "step": 43120
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.01624625908508e-06,
+      "loss": 0.5164,
+      "step": 43130
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.994869602394186e-06,
+      "loss": 0.5173,
+      "step": 43140
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.973492945703292e-06,
+      "loss": 0.5519,
+      "step": 43150
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.952116289012399e-06,
+      "loss": 0.5036,
+      "step": 43160
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.930739632321507e-06,
+      "loss": 0.563,
+      "step": 43170
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.909362975630611e-06,
+      "loss": 0.5159,
+      "step": 43180
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.887986318939717e-06,
+      "loss": 0.5354,
+      "step": 43190
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.866609662248825e-06,
+      "loss": 0.5548,
+      "step": 43200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.845233005557931e-06,
+      "loss": 0.5043,
+      "step": 43210
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.823856348867038e-06,
+      "loss": 0.5604,
+      "step": 43220
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.802479692176144e-06,
+      "loss": 0.5163,
+      "step": 43230
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.78110303548525e-06,
+      "loss": 0.5172,
+      "step": 43240
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.759726378794358e-06,
+      "loss": 0.5485,
+      "step": 43250
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.738349722103464e-06,
+      "loss": 0.5004,
+      "step": 43260
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.716973065412569e-06,
+      "loss": 0.5551,
+      "step": 43270
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.695596408721676e-06,
+      "loss": 0.5127,
+      "step": 43280
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.674219752030783e-06,
+      "loss": 0.5253,
+      "step": 43290
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.652843095339889e-06,
+      "loss": 0.5409,
+      "step": 43300
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.631466438648995e-06,
+      "loss": 0.5021,
+      "step": 43310
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.610089781958102e-06,
+      "loss": 0.5556,
+      "step": 43320
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.588713125267209e-06,
+      "loss": 0.5181,
+      "step": 43330
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.567336468576315e-06,
+      "loss": 0.5273,
+      "step": 43340
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.545959811885421e-06,
+      "loss": 0.5626,
+      "step": 43350
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.524583155194529e-06,
+      "loss": 0.4999,
+      "step": 43360
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.503206498503634e-06,
+      "loss": 0.5666,
+      "step": 43370
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.48182984181274e-06,
+      "loss": 0.5158,
+      "step": 43380
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.460453185121847e-06,
+      "loss": 0.524,
+      "step": 43390
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.4390765284309535e-06,
+      "loss": 0.5447,
+      "step": 43400
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.4176998717400605e-06,
+      "loss": 0.5059,
+      "step": 43410
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.396323215049167e-06,
+      "loss": 0.5517,
+      "step": 43420
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.374946558358273e-06,
+      "loss": 0.5059,
+      "step": 43430
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.35356990166738e-06,
+      "loss": 0.5145,
+      "step": 43440
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.332193244976486e-06,
+      "loss": 0.5574,
+      "step": 43450
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.3108165882855915e-06,
+      "loss": 0.5033,
+      "step": 43460
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.2894399315946994e-06,
+      "loss": 0.5653,
+      "step": 43470
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.268063274903806e-06,
+      "loss": 0.5059,
+      "step": 43480
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.246686618212911e-06,
+      "loss": 0.5133,
+      "step": 43490
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.225309961522019e-06,
+      "loss": 0.5418,
+      "step": 43500
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.203933304831124e-06,
+      "loss": 0.5009,
+      "step": 43510
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.182556648140232e-06,
+      "loss": 0.5598,
+      "step": 43520
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.1611799914493375e-06,
+      "loss": 0.517,
+      "step": 43530
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.139803334758444e-06,
+      "loss": 0.5139,
+      "step": 43540
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.118426678067551e-06,
+      "loss": 0.5522,
+      "step": 43550
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.097050021376657e-06,
+      "loss": 0.5017,
+      "step": 43560
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.075673364685763e-06,
+      "loss": 0.5556,
+      "step": 43570
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.05429670799487e-06,
+      "loss": 0.5191,
+      "step": 43580
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.0329200513039764e-06,
+      "loss": 0.5124,
+      "step": 43590
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.0115433946130835e-06,
+      "loss": 0.5473,
+      "step": 43600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.99016673792219e-06,
+      "loss": 0.5112,
+      "step": 43610
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.968790081231295e-06,
+      "loss": 0.5667,
+      "step": 43620
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.947413424540403e-06,
+      "loss": 0.5144,
+      "step": 43630
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.926036767849508e-06,
+      "loss": 0.5123,
+      "step": 43640
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.9046601111586145e-06,
+      "loss": 0.5561,
+      "step": 43650
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.8832834544677216e-06,
+      "loss": 0.5013,
+      "step": 43660
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.861906797776828e-06,
+      "loss": 0.5632,
+      "step": 43670
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.840530141085935e-06,
+      "loss": 0.5162,
+      "step": 43680
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.819153484395041e-06,
+      "loss": 0.522,
+      "step": 43690
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.797776827704147e-06,
+      "loss": 0.5467,
+      "step": 43700
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.776400171013254e-06,
+      "loss": 0.506,
+      "step": 43710
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.7550235143223605e-06,
+      "loss": 0.5666,
+      "step": 43720
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.733646857631466e-06,
+      "loss": 0.5142,
+      "step": 43730
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.712270200940574e-06,
+      "loss": 0.5148,
+      "step": 43740
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.690893544249679e-06,
+      "loss": 0.5491,
+      "step": 43750
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.669516887558785e-06,
+      "loss": 0.5048,
+      "step": 43760
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.648140230867892e-06,
+      "loss": 0.568,
+      "step": 43770
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.6267635741769986e-06,
+      "loss": 0.51,
+      "step": 43780
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.6053869174861065e-06,
+      "loss": 0.5169,
+      "step": 43790
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.584010260795212e-06,
+      "loss": 0.5524,
+      "step": 43800
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.562633604104318e-06,
+      "loss": 0.4972,
+      "step": 43810
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.541256947413425e-06,
+      "loss": 0.5626,
+      "step": 43820
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.519880290722531e-06,
+      "loss": 0.5136,
+      "step": 43830
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.4985036340316375e-06,
+      "loss": 0.5139,
+      "step": 43840
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.4771269773407445e-06,
+      "loss": 0.5448,
+      "step": 43850
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.455750320649851e-06,
+      "loss": 0.504,
+      "step": 43860
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.434373663958958e-06,
+      "loss": 0.5622,
+      "step": 43870
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.412997007268064e-06,
+      "loss": 0.5169,
+      "step": 43880
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.391620350577169e-06,
+      "loss": 0.5125,
+      "step": 43890
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.370243693886277e-06,
+      "loss": 0.5451,
+      "step": 43900
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.348867037195383e-06,
+      "loss": 0.5,
+      "step": 43910
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.327490380504489e-06,
+      "loss": 0.5554,
+      "step": 43920
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.306113723813596e-06,
+      "loss": 0.5153,
+      "step": 43930
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.284737067122702e-06,
+      "loss": 0.5142,
+      "step": 43940
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.263360410431809e-06,
+      "loss": 0.5491,
+      "step": 43950
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.241983753740915e-06,
+      "loss": 0.5065,
+      "step": 43960
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.220607097050022e-06,
+      "loss": 0.5663,
+      "step": 43970
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.199230440359128e-06,
+      "loss": 0.5183,
+      "step": 43980
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.177853783668235e-06,
+      "loss": 0.5178,
+      "step": 43990
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.156477126977341e-06,
+      "loss": 0.5539,
+      "step": 44000
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.135100470286447e-06,
+      "loss": 0.5025,
+      "step": 44010
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.113723813595553e-06,
+      "loss": 0.5581,
+      "step": 44020
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.0923471569046605e-06,
+      "loss": 0.5098,
+      "step": 44030
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.070970500213767e-06,
+      "loss": 0.5179,
+      "step": 44040
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.049593843522873e-06,
+      "loss": 0.5497,
+      "step": 44050
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.02821718683198e-06,
+      "loss": 0.5022,
+      "step": 44060
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.006840530141086e-06,
+      "loss": 0.5635,
+      "step": 44070
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.985463873450193e-06,
+      "loss": 0.5174,
+      "step": 44080
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.9640872167592985e-06,
+      "loss": 0.5158,
+      "step": 44090
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.942710560068406e-06,
+      "loss": 0.5526,
+      "step": 44100
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.921333903377512e-06,
+      "loss": 0.4952,
+      "step": 44110
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.899957246686619e-06,
+      "loss": 0.5581,
+      "step": 44120
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.878580589995725e-06,
+      "loss": 0.5146,
+      "step": 44130
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.857203933304831e-06,
+      "loss": 0.5283,
+      "step": 44140
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.835827276613938e-06,
+      "loss": 0.5506,
+      "step": 44150
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.8144506199230445e-06,
+      "loss": 0.501,
+      "step": 44160
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.793073963232151e-06,
+      "loss": 0.5568,
+      "step": 44170
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 5.771697306541257e-06,
+      "loss": 0.5214,
+      "step": 44180
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.750320649850364e-06,
+      "loss": 0.5237,
+      "step": 44190
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.72894399315947e-06,
+      "loss": 0.552,
+      "step": 44200
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.707567336468576e-06,
+      "loss": 0.5071,
+      "step": 44210
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.6861906797776834e-06,
+      "loss": 0.5658,
+      "step": 44220
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.66481402308679e-06,
+      "loss": 0.5065,
+      "step": 44230
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.643437366395897e-06,
+      "loss": 0.512,
+      "step": 44240
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.622060709705002e-06,
+      "loss": 0.5527,
+      "step": 44250
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.600684053014109e-06,
+      "loss": 0.5062,
+      "step": 44260
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.579307396323215e-06,
+      "loss": 0.5532,
+      "step": 44270
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.5579307396323215e-06,
+      "loss": 0.5115,
+      "step": 44280
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.536554082941428e-06,
+      "loss": 0.5196,
+      "step": 44290
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.515177426250535e-06,
+      "loss": 0.5455,
+      "step": 44300
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.493800769559641e-06,
+      "loss": 0.4997,
+      "step": 44310
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.472424112868747e-06,
+      "loss": 0.5656,
+      "step": 44320
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.451047456177854e-06,
+      "loss": 0.5156,
+      "step": 44330
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.4296707994869604e-06,
+      "loss": 0.5248,
+      "step": 44340
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.4082941427960675e-06,
+      "loss": 0.549,
+      "step": 44350
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.386917486105173e-06,
+      "loss": 0.5031,
+      "step": 44360
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.36554082941428e-06,
+      "loss": 0.5632,
+      "step": 44370
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.344164172723386e-06,
+      "loss": 0.5269,
+      "step": 44380
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.322787516032493e-06,
+      "loss": 0.514,
+      "step": 44390
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.301410859341599e-06,
+      "loss": 0.5573,
+      "step": 44400
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 5.2800342026507056e-06,
+      "loss": 0.4945,
+      "step": 44410
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.258657545959813e-06,
+      "loss": 0.5596,
+      "step": 44420
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.237280889268919e-06,
+      "loss": 0.5104,
+      "step": 44430
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.215904232578025e-06,
+      "loss": 0.5142,
+      "step": 44440
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.194527575887131e-06,
+      "loss": 0.5499,
+      "step": 44450
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.173150919196238e-06,
+      "loss": 0.4981,
+      "step": 44460
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.1517742625053445e-06,
+      "loss": 0.5667,
+      "step": 44470
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.130397605814451e-06,
+      "loss": 0.5113,
+      "step": 44480
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.109020949123557e-06,
+      "loss": 0.5228,
+      "step": 44490
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.087644292432664e-06,
+      "loss": 0.5447,
+      "step": 44500
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.06626763574177e-06,
+      "loss": 0.5033,
+      "step": 44510
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.044890979050876e-06,
+      "loss": 0.557,
+      "step": 44520
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.023514322359983e-06,
+      "loss": 0.5137,
+      "step": 44530
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 5.00213766566909e-06,
+      "loss": 0.5154,
+      "step": 44540
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.980761008978196e-06,
+      "loss": 0.5396,
+      "step": 44550
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.959384352287302e-06,
+      "loss": 0.5082,
+      "step": 44560
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.938007695596409e-06,
+      "loss": 0.5665,
+      "step": 44570
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.916631038905515e-06,
+      "loss": 0.5162,
+      "step": 44580
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.8952543822146215e-06,
+      "loss": 0.5198,
+      "step": 44590
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.8738777255237285e-06,
+      "loss": 0.5526,
+      "step": 44600
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.852501068832835e-06,
+      "loss": 0.4984,
+      "step": 44610
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.831124412141942e-06,
+      "loss": 0.5647,
+      "step": 44620
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.809747755451047e-06,
+      "loss": 0.5077,
+      "step": 44630
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.788371098760154e-06,
+      "loss": 0.5248,
+      "step": 44640
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.76699444206926e-06,
+      "loss": 0.5499,
+      "step": 44650
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.7456177853783675e-06,
+      "loss": 0.5014,
+      "step": 44660
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.724241128687474e-06,
+      "loss": 0.5595,
+      "step": 44670
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.70286447199658e-06,
+      "loss": 0.5133,
+      "step": 44680
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.681487815305687e-06,
+      "loss": 0.5327,
+      "step": 44690
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.660111158614793e-06,
+      "loss": 0.541,
+      "step": 44700
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.638734501923899e-06,
+      "loss": 0.4943,
+      "step": 44710
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.6173578452330055e-06,
+      "loss": 0.5684,
+      "step": 44720
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.595981188542113e-06,
+      "loss": 0.5179,
+      "step": 44730
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.574604531851219e-06,
+      "loss": 0.5231,
+      "step": 44740
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.553227875160325e-06,
+      "loss": 0.5346,
+      "step": 44750
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.531851218469431e-06,
+      "loss": 0.5048,
+      "step": 44760
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.510474561778538e-06,
+      "loss": 0.5601,
+      "step": 44770
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.4890979050876445e-06,
+      "loss": 0.5118,
+      "step": 44780
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.467721248396751e-06,
+      "loss": 0.5176,
+      "step": 44790
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.446344591705858e-06,
+      "loss": 0.5437,
+      "step": 44800
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.424967935014964e-06,
+      "loss": 0.5068,
+      "step": 44810
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.40359127832407e-06,
+      "loss": 0.5567,
+      "step": 44820
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.382214621633176e-06,
+      "loss": 0.5122,
+      "step": 44830
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.360837964942283e-06,
+      "loss": 0.5219,
+      "step": 44840
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.33946130825139e-06,
+      "loss": 0.5458,
+      "step": 44850
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.318084651560496e-06,
+      "loss": 0.5036,
+      "step": 44860
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.296707994869603e-06,
+      "loss": 0.5615,
+      "step": 44870
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.275331338178709e-06,
+      "loss": 0.5088,
+      "step": 44880
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.253954681487816e-06,
+      "loss": 0.5225,
+      "step": 44890
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.2325780247969215e-06,
+      "loss": 0.5501,
+      "step": 44900
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.2112013681060285e-06,
+      "loss": 0.4966,
+      "step": 44910
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.189824711415135e-06,
+      "loss": 0.5606,
+      "step": 44920
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.168448054724242e-06,
+      "loss": 0.5131,
+      "step": 44930
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.147071398033347e-06,
+      "loss": 0.5167,
+      "step": 44940
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.125694741342454e-06,
+      "loss": 0.5414,
+      "step": 44950
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.104318084651561e-06,
+      "loss": 0.5018,
+      "step": 44960
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.0829414279606674e-06,
+      "loss": 0.5527,
+      "step": 44970
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.061564771269774e-06,
+      "loss": 0.5169,
+      "step": 44980
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.04018811457888e-06,
+      "loss": 0.5262,
+      "step": 44990
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.018811457887987e-06,
+      "loss": 0.551,
+      "step": 45000
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.997434801197093e-06,
+      "loss": 0.5051,
+      "step": 45010
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.976058144506199e-06,
+      "loss": 0.5705,
+      "step": 45020
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.9546814878153055e-06,
+      "loss": 0.5099,
+      "step": 45030
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.9333048311244126e-06,
+      "loss": 0.5173,
+      "step": 45040
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.911928174433519e-06,
+      "loss": 0.5525,
+      "step": 45050
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.890551517742625e-06,
+      "loss": 0.5022,
+      "step": 45060
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.869174861051732e-06,
+      "loss": 0.5544,
+      "step": 45070
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.847798204360838e-06,
+      "loss": 0.5133,
+      "step": 45080
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.8264215476699444e-06,
+      "loss": 0.5148,
+      "step": 45090
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.805044890979051e-06,
+      "loss": 0.5559,
+      "step": 45100
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.7836682342881577e-06,
+      "loss": 0.4956,
+      "step": 45110
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.7622915775972643e-06,
+      "loss": 0.5639,
+      "step": 45120
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.74091492090637e-06,
+      "loss": 0.4992,
+      "step": 45130
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.7195382642154767e-06,
+      "loss": 0.5274,
+      "step": 45140
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.6981616075245834e-06,
+      "loss": 0.5568,
+      "step": 45150
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.67678495083369e-06,
+      "loss": 0.496,
+      "step": 45160
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.6554082941427958e-06,
+      "loss": 0.5591,
+      "step": 45170
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.634031637451903e-06,
+      "loss": 0.5132,
+      "step": 45180
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.6126549807610095e-06,
+      "loss": 0.5102,
+      "step": 45190
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.591278324070116e-06,
+      "loss": 0.5412,
+      "step": 45200
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.569901667379222e-06,
+      "loss": 0.4961,
+      "step": 45210
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.5485250106883285e-06,
+      "loss": 0.556,
+      "step": 45220
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.527148353997435e-06,
+      "loss": 0.5214,
+      "step": 45230
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.5057716973065417e-06,
+      "loss": 0.52,
+      "step": 45240
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.4843950406156475e-06,
+      "loss": 0.546,
+      "step": 45250
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.463018383924754e-06,
+      "loss": 0.4999,
+      "step": 45260
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.4416417272338608e-06,
+      "loss": 0.5628,
+      "step": 45270
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.4202650705429674e-06,
+      "loss": 0.5185,
+      "step": 45280
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.3988884138520736e-06,
+      "loss": 0.5123,
+      "step": 45290
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.3775117571611802e-06,
+      "loss": 0.5503,
+      "step": 45300
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.356135100470287e-06,
+      "loss": 0.5033,
+      "step": 45310
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.3347584437793927e-06,
+      "loss": 0.5605,
+      "step": 45320
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.3133817870884993e-06,
+      "loss": 0.5105,
+      "step": 45330
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.292005130397606e-06,
+      "loss": 0.5252,
+      "step": 45340
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.2706284737067125e-06,
+      "loss": 0.5512,
+      "step": 45350
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.2492518170158187e-06,
+      "loss": 0.4962,
+      "step": 45360
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.2278751603249254e-06,
+      "loss": 0.5559,
+      "step": 45370
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.206498503634032e-06,
+      "loss": 0.5138,
+      "step": 45380
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.1851218469431386e-06,
+      "loss": 0.5135,
+      "step": 45390
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.1637451902522444e-06,
+      "loss": 0.5462,
+      "step": 45400
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.142368533561351e-06,
+      "loss": 0.5033,
+      "step": 45410
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.1209918768704577e-06,
+      "loss": 0.5678,
+      "step": 45420
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.099615220179564e-06,
+      "loss": 0.5046,
+      "step": 45430
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.0782385634886705e-06,
+      "loss": 0.5194,
+      "step": 45440
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.0568619067977767e-06,
+      "loss": 0.5476,
+      "step": 45450
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.0354852501068833e-06,
+      "loss": 0.5,
+      "step": 45460
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.01410859341599e-06,
+      "loss": 0.5694,
+      "step": 45470
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.9927319367250966e-06,
+      "loss": 0.5191,
+      "step": 45480
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.971355280034203e-06,
+      "loss": 0.5117,
+      "step": 45490
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.9499786233433094e-06,
+      "loss": 0.5498,
+      "step": 45500
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.9286019666524156e-06,
+      "loss": 0.4985,
+      "step": 45510
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.9072253099615223e-06,
+      "loss": 0.5649,
+      "step": 45520
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8858486532706285e-06,
+      "loss": 0.5135,
+      "step": 45530
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.864471996579735e-06,
+      "loss": 0.5233,
+      "step": 45540
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8430953398888417e-06,
+      "loss": 0.5406,
+      "step": 45550
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8217186831979483e-06,
+      "loss": 0.5069,
+      "step": 45560
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.8003420265070546e-06,
+      "loss": 0.5656,
+      "step": 45570
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.7789653698161608e-06,
+      "loss": 0.5166,
+      "step": 45580
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.7575887131252674e-06,
+      "loss": 0.5159,
+      "step": 45590
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7362120564343736e-06,
+      "loss": 0.5441,
+      "step": 45600
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.7148353997434802e-06,
+      "loss": 0.499,
+      "step": 45610
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6934587430525864e-06,
+      "loss": 0.5624,
+      "step": 45620
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.672082086361693e-06,
+      "loss": 0.5165,
+      "step": 45630
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6507054296707997e-06,
+      "loss": 0.5171,
+      "step": 45640
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6293287729799063e-06,
+      "loss": 0.5401,
+      "step": 45650
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.6079521162890125e-06,
+      "loss": 0.5031,
+      "step": 45660
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.586575459598119e-06,
+      "loss": 0.5583,
+      "step": 45670
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.5651988029072253e-06,
+      "loss": 0.5144,
+      "step": 45680
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.543822146216332e-06,
+      "loss": 0.5251,
+      "step": 45690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.522445489525438e-06,
+      "loss": 0.5468,
+      "step": 45700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.501068832834545e-06,
+      "loss": 0.502,
+      "step": 45710
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.479692176143651e-06,
+      "loss": 0.5657,
+      "step": 45720
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4583155194527576e-06,
+      "loss": 0.5112,
+      "step": 45730
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4369388627618643e-06,
+      "loss": 0.5223,
+      "step": 45740
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.415562206070971e-06,
+      "loss": 0.559,
+      "step": 45750
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.394185549380077e-06,
+      "loss": 0.4934,
+      "step": 45760
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.3728088926891837e-06,
+      "loss": 0.5538,
+      "step": 45770
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.35143223599829e-06,
+      "loss": 0.5193,
+      "step": 45780
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.3300555793073966e-06,
+      "loss": 0.5197,
+      "step": 45790
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.3086789226165028e-06,
+      "loss": 0.5506,
+      "step": 45800
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.2873022659256094e-06,
+      "loss": 0.5065,
+      "step": 45810
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.2659256092347156e-06,
+      "loss": 0.5607,
+      "step": 45820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.2445489525438222e-06,
+      "loss": 0.5093,
+      "step": 45830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.223172295852929e-06,
+      "loss": 0.5257,
+      "step": 45840
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.201795639162035e-06,
+      "loss": 0.5583,
+      "step": 45850
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1804189824711417e-06,
+      "loss": 0.499,
+      "step": 45860
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.159042325780248e-06,
+      "loss": 0.5638,
+      "step": 45870
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1376656690893545e-06,
+      "loss": 0.519,
+      "step": 45880
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.1162890123984607e-06,
+      "loss": 0.5166,
+      "step": 45890
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0949123557075674e-06,
+      "loss": 0.552,
+      "step": 45900
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0735356990166736e-06,
+      "loss": 0.5035,
+      "step": 45910
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0521590423257806e-06,
+      "loss": 0.5601,
+      "step": 45920
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.030782385634887e-06,
+      "loss": 0.5117,
+      "step": 45930
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.0094057289439935e-06,
+      "loss": 0.5137,
+      "step": 45940
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9880290722530997e-06,
+      "loss": 0.5462,
+      "step": 45950
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9666524155622063e-06,
+      "loss": 0.4989,
+      "step": 45960
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9452757588713125e-06,
+      "loss": 0.5567,
+      "step": 45970
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.923899102180419e-06,
+      "loss": 0.5115,
+      "step": 45980
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.9025224454895255e-06,
+      "loss": 0.511,
+      "step": 45990
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8811457887986322e-06,
+      "loss": 0.5463,
+      "step": 46000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8597691321077384e-06,
+      "loss": 0.5069,
+      "step": 46010
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.838392475416845e-06,
+      "loss": 0.5675,
+      "step": 46020
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8170158187259514e-06,
+      "loss": 0.519,
+      "step": 46030
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.795639162035058e-06,
+      "loss": 0.5113,
+      "step": 46040
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7742625053441642e-06,
+      "loss": 0.5314,
+      "step": 46050
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.7528858486532709e-06,
+      "loss": 0.5022,
+      "step": 46060
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.731509191962377e-06,
+      "loss": 0.5652,
+      "step": 46070
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.7101325352714837e-06,
+      "loss": 0.5118,
+      "step": 46080
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.6887558785805901e-06,
+      "loss": 0.5104,
+      "step": 46090
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.6673792218896963e-06,
+      "loss": 0.5496,
+      "step": 46100
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.646002565198803e-06,
+      "loss": 0.5052,
+      "step": 46110
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.6246259085079094e-06,
+      "loss": 0.5558,
+      "step": 46120
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.603249251817016e-06,
+      "loss": 0.5149,
+      "step": 46130
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5818725951261222e-06,
+      "loss": 0.5134,
+      "step": 46140
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5604959384352288e-06,
+      "loss": 0.5534,
+      "step": 46150
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5391192817443353e-06,
+      "loss": 0.4978,
+      "step": 46160
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.5177426250534417e-06,
+      "loss": 0.553,
+      "step": 46170
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4963659683625483e-06,
+      "loss": 0.5211,
+      "step": 46180
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4749893116716547e-06,
+      "loss": 0.5135,
+      "step": 46190
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4536126549807611e-06,
+      "loss": 0.5457,
+      "step": 46200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4322359982898675e-06,
+      "loss": 0.4989,
+      "step": 46210
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.4108593415989742e-06,
+      "loss": 0.5647,
+      "step": 46220
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3894826849080804e-06,
+      "loss": 0.5102,
+      "step": 46230
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3681060282171868e-06,
+      "loss": 0.5151,
+      "step": 46240
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3467293715262932e-06,
+      "loss": 0.5475,
+      "step": 46250
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3253527148353998e-06,
+      "loss": 0.502,
+      "step": 46260
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.3039760581445063e-06,
+      "loss": 0.5578,
+      "step": 46270
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.2825994014536127e-06,
+      "loss": 0.5078,
+      "step": 46280
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.261222744762719e-06,
+      "loss": 0.5188,
+      "step": 46290
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.2398460880718255e-06,
+      "loss": 0.5539,
+      "step": 46300
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.2184694313809321e-06,
+      "loss": 0.498,
+      "step": 46310
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.1970927746900386e-06,
+      "loss": 0.5576,
+      "step": 46320
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.175716117999145e-06,
+      "loss": 0.5165,
+      "step": 46330
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.1543394613082514e-06,
+      "loss": 0.524,
+      "step": 46340
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.1329628046173578e-06,
+      "loss": 0.5566,
+      "step": 46350
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.1115861479264644e-06,
+      "loss": 0.5068,
+      "step": 46360
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0902094912355708e-06,
+      "loss": 0.5629,
+      "step": 46370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0688328345446773e-06,
+      "loss": 0.5063,
+      "step": 46380
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0474561778537837e-06,
+      "loss": 0.5274,
+      "step": 46390
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0260795211628903e-06,
+      "loss": 0.5525,
+      "step": 46400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.0047028644719967e-06,
+      "loss": 0.5004,
+      "step": 46410
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.833262077811031e-07,
+      "loss": 0.5528,
+      "step": 46420
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.619495510902096e-07,
+      "loss": 0.5156,
+      "step": 46430
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.405728943993161e-07,
+      "loss": 0.5121,
+      "step": 46440
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.191962377084225e-07,
+      "loss": 0.5511,
+      "step": 46450
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.97819581017529e-07,
+      "loss": 0.5045,
+      "step": 46460
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.764429243266354e-07,
+      "loss": 0.5557,
+      "step": 46470
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.550662676357419e-07,
+      "loss": 0.5069,
+      "step": 46480
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.336896109448482e-07,
+      "loss": 0.5123,
+      "step": 46490
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.123129542539547e-07,
+      "loss": 0.5495,
+      "step": 46500
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.909362975630611e-07,
+      "loss": 0.5034,
+      "step": 46510
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 7.695596408721676e-07,
+      "loss": 0.5571,
+      "step": 46520
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.481829841812741e-07,
+      "loss": 0.5152,
+      "step": 46530
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.268063274903806e-07,
+      "loss": 0.5227,
+      "step": 46540
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 7.054296707994871e-07,
+      "loss": 0.5537,
+      "step": 46550
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.840530141085934e-07,
+      "loss": 0.4961,
+      "step": 46560
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.626763574176999e-07,
+      "loss": 0.5548,
+      "step": 46570
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.412997007268063e-07,
+      "loss": 0.5136,
+      "step": 46580
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.199230440359128e-07,
+      "loss": 0.5169,
+      "step": 46590
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.985463873450193e-07,
+      "loss": 0.5556,
+      "step": 46600
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.771697306541257e-07,
+      "loss": 0.4982,
+      "step": 46610
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.557930739632322e-07,
+      "loss": 0.5576,
+      "step": 46620
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.344164172723386e-07,
+      "loss": 0.5068,
+      "step": 46630
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 5.130397605814452e-07,
+      "loss": 0.5199,
+      "step": 46640
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.916631038905516e-07,
+      "loss": 0.5374,
+      "step": 46650
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.7028644719965804e-07,
+      "loss": 0.4979,
+      "step": 46660
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.489097905087645e-07,
+      "loss": 0.5569,
+      "step": 46670
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.2753313381787093e-07,
+      "loss": 0.5127,
+      "step": 46680
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.0615647712697734e-07,
+      "loss": 0.5196,
+      "step": 46690
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.847798204360838e-07,
+      "loss": 0.5442,
+      "step": 46700
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.634031637451903e-07,
+      "loss": 0.4944,
+      "step": 46710
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.420265070542967e-07,
+      "loss": 0.5522,
+      "step": 46720
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.2064985036340317e-07,
+      "loss": 0.511,
+      "step": 46730
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.9927319367250964e-07,
+      "loss": 0.5214,
+      "step": 46740
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.778965369816161e-07,
+      "loss": 0.5511,
+      "step": 46750
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.565198802907226e-07,
+      "loss": 0.4981,
+      "step": 46760
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.3514322359982902e-07,
+      "loss": 0.566,
+      "step": 46770
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.1376656690893546e-07,
+      "loss": 0.5076,
+      "step": 46780
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.923899102180419e-07,
+      "loss": 0.5226,
+      "step": 46790
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.7101325352714835e-07,
+      "loss": 0.5459,
+      "step": 46800
+    }
+  ],
+  "max_steps": 46880,
+  "num_train_epochs": 2,
+  "total_flos": 2.979121974892036e+19,
+  "trial_name": null,
+  "trial_params": null
+}