diff --git "a/checkpoint-49300/trainer_state.json" "b/checkpoint-49300/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-49300/trainer_state.json"
@@ -0,0 +1,29601 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8531994981179423,
+  "eval_steps": 500,
+  "global_step": 49300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.844290657439446e-07,
+      "loss": 11.3843,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1072664359861592e-06,
+      "loss": 11.4922,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7993079584775088e-06,
+      "loss": 11.4643,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4913494809688584e-06,
+      "loss": 11.2703,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1833910034602078e-06,
+      "loss": 11.0879,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.875432525951557e-06,
+      "loss": 11.0263,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5674740484429065e-06,
+      "loss": 10.7875,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.190311418685121e-06,
+      "loss": 10.4875,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.882352941176471e-06,
+      "loss": 10.1665,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5743944636678194e-06,
+      "loss": 9.6567,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.2664359861591705e-06,
+      "loss": 9.3623,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.958477508650519e-06,
+      "loss": 9.1007,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.650519031141868e-06,
+      "loss": 8.7675,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.34256055363322e-06,
+      "loss": 8.4775,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0034602076124568e-05,
+      "loss": 8.3187,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0726643598615918e-05,
+      "loss": 8.0933,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1418685121107267e-05,
+      "loss": 7.9094,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2110726643598615e-05,
+      "loss": 7.9038,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2802768166089965e-05,
+      "loss": 7.8251,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3494809688581317e-05,
+      "loss": 7.69,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4186851211072666e-05,
+      "loss": 7.5891,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4878892733564014e-05,
+      "loss": 7.5995,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5570934256055363e-05,
+      "loss": 7.5861,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6262975778546713e-05,
+      "loss": 7.457,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6955017301038063e-05,
+      "loss": 7.4987,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7647058823529414e-05,
+      "loss": 7.3925,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8339100346020764e-05,
+      "loss": 7.3742,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.903114186851211e-05,
+      "loss": 7.2152,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.972318339100346e-05,
+      "loss": 7.1937,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.041522491349481e-05,
+      "loss": 7.2579,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1107266435986158e-05,
+      "loss": 7.1413,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.179930795847751e-05,
+      "loss": 7.1005,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.249134948096886e-05,
+      "loss": 7.0903,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.318339100346021e-05,
+      "loss": 7.0204,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.387543252595156e-05,
+      "loss": 6.9821,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4567474048442906e-05,
+      "loss": 6.9959,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5259515570934256e-05,
+      "loss": 6.9195,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5882352941176475e-05,
+      "loss": 6.8595,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6505190311418688e-05,
+      "loss": 6.8417,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7197231833910038e-05,
+      "loss": 6.8092,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7889273356401385e-05,
+      "loss": 6.6811,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8581314878892735e-05,
+      "loss": 6.6395,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.927335640138408e-05,
+      "loss": 6.664,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9965397923875432e-05,
+      "loss": 6.6459,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.065743944636678e-05,
+      "loss": 6.5786,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.134948096885813e-05,
+      "loss": 6.523,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.204152249134948e-05,
+      "loss": 6.4288,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.273356401384083e-05,
+      "loss": 6.4009,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.342560553633218e-05,
+      "loss": 6.2901,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.411764705882353e-05,
+      "loss": 6.2791,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4809688581314883e-05,
+      "loss": 6.3766,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.550173010380623e-05,
+      "loss": 6.2361,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.619377162629758e-05,
+      "loss": 6.3265,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.688581314878893e-05,
+      "loss": 6.1918,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.757785467128028e-05,
+      "loss": 6.2941,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.826989619377163e-05,
+      "loss": 6.1485,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.896193771626298e-05,
+      "loss": 6.1348,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.965397923875433e-05,
+      "loss": 6.0892,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.034602076124568e-05,
+      "loss": 6.0854,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.103806228373703e-05,
+      "loss": 5.9963,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.173010380622838e-05,
+      "loss": 5.9803,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.242214532871973e-05,
+      "loss": 5.9827,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.311418685121107e-05,
+      "loss": 5.9034,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.380622837370242e-05,
+      "loss": 5.8993,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.449826989619377e-05,
+      "loss": 5.8226,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.519031141868512e-05,
+      "loss": 5.9912,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.588235294117647e-05,
+      "loss": 5.821,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.657439446366782e-05,
+      "loss": 5.8343,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7266435986159174e-05,
+      "loss": 5.7343,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7958477508650524e-05,
+      "loss": 5.7435,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8650519031141874e-05,
+      "loss": 5.7636,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9342560553633224e-05,
+      "loss": 5.7366,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.003460207612457e-05,
+      "loss": 5.6594,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.072664359861592e-05,
+      "loss": 5.5874,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.141868512110727e-05,
+      "loss": 5.7374,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.211072664359862e-05,
+      "loss": 5.6076,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.280276816608997e-05,
+      "loss": 5.6246,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.349480968858131e-05,
+      "loss": 5.6905,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.418685121107266e-05,
+      "loss": 5.61,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.487889273356401e-05,
+      "loss": 5.5684,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.557093425605536e-05,
+      "loss": 5.5248,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.626297577854671e-05,
+      "loss": 5.5351,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.695501730103806e-05,
+      "loss": 5.5094,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.764705882352941e-05,
+      "loss": 5.3651,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.833910034602076e-05,
+      "loss": 5.4953,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.903114186851212e-05,
+      "loss": 5.4298,
+      "step": 860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.972318339100347e-05,
+      "loss": 5.4281,
+      "step": 870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0415224913494814e-05,
+      "loss": 5.4204,
+      "step": 880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.110726643598617e-05,
+      "loss": 5.4936,
+      "step": 890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.179930795847751e-05,
+      "loss": 5.3641,
+      "step": 900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.249134948096886e-05,
+      "loss": 5.3753,
+      "step": 910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.318339100346021e-05,
+      "loss": 5.3648,
+      "step": 920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.387543252595156e-05,
+      "loss": 5.3226,
+      "step": 930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.456747404844291e-05,
+      "loss": 5.3352,
+      "step": 940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.525951557093426e-05,
+      "loss": 5.3024,
+      "step": 950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.595155709342561e-05,
+      "loss": 5.3395,
+      "step": 960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.664359861591696e-05,
+      "loss": 5.3071,
+      "step": 970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.733564013840831e-05,
+      "loss": 5.2185,
+      "step": 980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.802768166089966e-05,
+      "loss": 5.22,
+      "step": 990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.871972318339101e-05,
+      "loss": 5.1606,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.941176470588236e-05,
+      "loss": 5.223,
+      "step": 1010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.01038062283737e-05,
+      "loss": 5.1348,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.079584775086505e-05,
+      "loss": 5.2111,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.14878892733564e-05,
+      "loss": 5.1559,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.217993079584775e-05,
+      "loss": 5.1723,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.28719723183391e-05,
+      "loss": 5.1858,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.356401384083045e-05,
+      "loss": 5.1041,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.425605536332181e-05,
+      "loss": 5.1212,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.494809688581316e-05,
+      "loss": 5.1205,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.564013840830451e-05,
+      "loss": 5.1145,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.633217993079585e-05,
+      "loss": 5.0884,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.70242214532872e-05,
+      "loss": 5.1211,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.771626297577855e-05,
+      "loss": 5.0219,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.84083044982699e-05,
+      "loss": 5.0685,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.910034602076125e-05,
+      "loss": 4.9674,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.97923875432526e-05,
+      "loss": 5.0026,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.048442906574395e-05,
+      "loss": 5.0192,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.11764705882353e-05,
+      "loss": 5.0607,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.186851211072665e-05,
+      "loss": 5.078,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.2560553633218e-05,
+      "loss": 4.9619,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.325259515570935e-05,
+      "loss": 4.9139,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.394463667820069e-05,
+      "loss": 4.9738,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.463667820069204e-05,
+      "loss": 4.9732,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.532871972318339e-05,
+      "loss": 4.9963,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.602076124567474e-05,
+      "loss": 4.9417,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.671280276816609e-05,
+      "loss": 4.9426,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.740484429065744e-05,
+      "loss": 4.9065,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.809688581314879e-05,
+      "loss": 4.8885,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.878892733564014e-05,
+      "loss": 4.918,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.94809688581315e-05,
+      "loss": 4.8905,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.017301038062284e-05,
+      "loss": 4.8405,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.086505190311419e-05,
+      "loss": 4.9333,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.155709342560554e-05,
+      "loss": 4.8638,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.224913494809689e-05,
+      "loss": 4.8589,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.294117647058824e-05,
+      "loss": 4.8543,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.363321799307959e-05,
+      "loss": 4.8445,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.432525951557094e-05,
+      "loss": 4.8007,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.501730103806229e-05,
+      "loss": 4.8765,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.570934256055364e-05,
+      "loss": 4.7588,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.640138408304499e-05,
+      "loss": 4.8286,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.709342560553634e-05,
+      "loss": 4.8321,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.778546712802769e-05,
+      "loss": 4.7992,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.847750865051903e-05,
+      "loss": 4.8228,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.916955017301038e-05,
+      "loss": 4.7561,
+      "step": 1440
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986159169550173e-05,
+      "loss": 4.8082,
+      "step": 1450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010055363321799309,
+      "loss": 4.7646,
+      "step": 1460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010124567474048443,
+      "loss": 4.7917,
+      "step": 1470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001019377162629758,
+      "loss": 4.7832,
+      "step": 1480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010262975778546713,
+      "loss": 4.8308,
+      "step": 1490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001033217993079585,
+      "loss": 4.7319,
+      "step": 1500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010401384083044983,
+      "loss": 4.7214,
+      "step": 1510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010470588235294118,
+      "loss": 4.7412,
+      "step": 1520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010539792387543253,
+      "loss": 4.723,
+      "step": 1530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010608996539792388,
+      "loss": 4.7602,
+      "step": 1540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010678200692041522,
+      "loss": 4.7148,
+      "step": 1550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010747404844290658,
+      "loss": 4.6936,
+      "step": 1560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010816608996539792,
+      "loss": 4.7064,
+      "step": 1570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010885813148788928,
+      "loss": 4.651,
+      "step": 1580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00010955017301038062,
+      "loss": 4.6812,
+      "step": 1590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011024221453287198,
+      "loss": 4.5911,
+      "step": 1600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011093425605536332,
+      "loss": 4.6238,
+      "step": 1610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011162629757785468,
+      "loss": 4.6017,
+      "step": 1620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011231833910034603,
+      "loss": 4.6857,
+      "step": 1630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011301038062283737,
+      "loss": 4.5994,
+      "step": 1640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011370242214532873,
+      "loss": 4.6037,
+      "step": 1650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011439446366782007,
+      "loss": 4.5733,
+      "step": 1660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011508650519031143,
+      "loss": 4.5224,
+      "step": 1670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011577854671280277,
+      "loss": 4.6287,
+      "step": 1680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011647058823529413,
+      "loss": 4.6344,
+      "step": 1690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011716262975778547,
+      "loss": 4.5715,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011764705882352942,
+      "loss": 9.2041,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011833910034602077,
+      "loss": 7.614,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011896193771626298,
+      "loss": 6.6816,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00011965397923875432,
+      "loss": 6.06,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012034602076124568,
+      "loss": 5.6681,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012103806228373702,
+      "loss": 5.4592,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012173010380622839,
+      "loss": 5.3273,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012242214532871975,
+      "loss": 5.2278,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012311418685121109,
+      "loss": 5.1185,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012380622837370245,
+      "loss": 5.0205,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012449826989619379,
+      "loss": 5.0303,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012519031141868512,
+      "loss": 4.9824,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001258823529411765,
+      "loss": 4.9583,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012657439446366782,
+      "loss": 4.8147,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012726643598615916,
+      "loss": 4.8703,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012782006920415226,
+      "loss": 7.5927,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001285121107266436,
+      "loss": 6.4347,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00012920415224913496,
+      "loss": 5.6382,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001298961937716263,
+      "loss": 5.3613,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013058823529411766,
+      "loss": 5.1758,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000131280276816609,
+      "loss": 5.0184,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013197231833910036,
+      "loss": 5.0102,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001326643598615917,
+      "loss": 4.966,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013335640138408306,
+      "loss": 4.9783,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001340484429065744,
+      "loss": 4.823,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013474048442906574,
+      "loss": 4.8571,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001354325259515571,
+      "loss": 4.8616,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013612456747404844,
+      "loss": 4.8164,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001368166089965398,
+      "loss": 4.6817,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013750865051903114,
+      "loss": 4.741,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001382006920415225,
+      "loss": 4.6959,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00013889273356401384,
+      "loss": 4.715,
+      "step": 2020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001395847750865052,
+      "loss": 4.6899,
+      "step": 2030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014027681660899654,
+      "loss": 4.6187,
+      "step": 2040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001409688581314879,
+      "loss": 4.6066,
+      "step": 2050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014166089965397924,
+      "loss": 4.6223,
+      "step": 2060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001423529411764706,
+      "loss": 4.5845,
+      "step": 2070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014304498269896194,
+      "loss": 4.6296,
+      "step": 2080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001437370242214533,
+      "loss": 4.5785,
+      "step": 2090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014442906574394464,
+      "loss": 4.5645,
+      "step": 2100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000145121107266436,
+      "loss": 4.5528,
+      "step": 2110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014581314878892734,
+      "loss": 4.5808,
+      "step": 2120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001465051903114187,
+      "loss": 4.5717,
+      "step": 2130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014719723183391004,
+      "loss": 4.4441,
+      "step": 2140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001478892733564014,
+      "loss": 4.5927,
+      "step": 2150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014858131487889274,
+      "loss": 4.5481,
+      "step": 2160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014927335640138408,
+      "loss": 4.4704,
+      "step": 2170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00014996539792387544,
+      "loss": 4.5283,
+      "step": 2180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015065743944636678,
+      "loss": 4.5318,
+      "step": 2190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015134948096885814,
+      "loss": 4.4498,
+      "step": 2200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015204152249134948,
+      "loss": 4.5004,
+      "step": 2210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015273356401384084,
+      "loss": 4.4088,
+      "step": 2220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015342560553633218,
+      "loss": 4.4286,
+      "step": 2230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015411764705882354,
+      "loss": 4.3485,
+      "step": 2240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015480968858131488,
+      "loss": 4.4573,
+      "step": 2250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015536332179930798,
+      "loss": 6.7076,
+      "step": 2260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015605536332179932,
+      "loss": 5.5923,
+      "step": 2270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015674740484429065,
+      "loss": 5.0411,
+      "step": 2280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015743944636678202,
+      "loss": 4.84,
+      "step": 2290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015813148788927336,
+      "loss": 4.7474,
+      "step": 2300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001588235294117647,
+      "loss": 4.7279,
+      "step": 2310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00015951557093425606,
+      "loss": 4.6071,
+      "step": 2320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001602076124567474,
+      "loss": 4.625,
+      "step": 2330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016089965397923876,
+      "loss": 4.5299,
+      "step": 2340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016159169550173012,
+      "loss": 4.5897,
+      "step": 2350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016228373702422146,
+      "loss": 4.5095,
+      "step": 2360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016297577854671282,
+      "loss": 4.4638,
+      "step": 2370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016366782006920416,
+      "loss": 4.4811,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016435986159169552,
+      "loss": 4.5259,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016505190311418686,
+      "loss": 4.4388,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016574394463667822,
+      "loss": 4.454,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016643598615916956,
+      "loss": 4.3943,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016712802768166092,
+      "loss": 4.5017,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016782006920415226,
+      "loss": 4.4687,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016851211072664362,
+      "loss": 4.4336,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016920415224913496,
+      "loss": 4.4298,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00016989619377162632,
+      "loss": 4.2917,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017058823529411766,
+      "loss": 4.4265,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000171280276816609,
+      "loss": 4.3893,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017197231833910036,
+      "loss": 4.3082,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001726643598615917,
+      "loss": 4.2956,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017335640138408303,
+      "loss": 4.2683,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001740484429065744,
+      "loss": 4.3174,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017474048442906573,
+      "loss": 4.2942,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001754325259515571,
+      "loss": 4.2764,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017612456747404843,
+      "loss": 4.2629,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001768166089965398,
+      "loss": 4.3498,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017750865051903116,
+      "loss": 4.2502,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001782006920415225,
+      "loss": 4.2069,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00017882352941176472,
+      "loss": 4.2616,
+      "step": 2600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00017951557093425605,
+      "loss": 4.3095,
+      "step": 2610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018020761245674742,
+      "loss": 4.211,
+      "step": 2620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018089965397923875,
+      "loss": 4.1715,
+      "step": 2630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018159169550173012,
+      "loss": 4.1979,
+      "step": 2640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018228373702422145,
+      "loss": 4.2505,
+      "step": 2650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018297577854671282,
+      "loss": 4.2162,
+      "step": 2660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018366782006920416,
+      "loss": 4.2223,
+      "step": 2670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018435986159169552,
+      "loss": 4.2917,
+      "step": 2680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018505190311418686,
+      "loss": 4.1327,
+      "step": 2690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018574394463667822,
+      "loss": 4.2182,
+      "step": 2700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018643598615916956,
+      "loss": 4.1291,
+      "step": 2710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018712802768166092,
+      "loss": 4.1336,
+      "step": 2720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018782006920415226,
+      "loss": 4.1628,
+      "step": 2730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018851211072664362,
+      "loss": 4.242,
+      "step": 2740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018920415224913496,
+      "loss": 4.0264,
+      "step": 2750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00018989619377162632,
+      "loss": 4.1376,
+      "step": 2760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019058823529411766,
+      "loss": 4.1273,
+      "step": 2770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000191280276816609,
+      "loss": 4.1551,
+      "step": 2780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019197231833910036,
+      "loss": 4.1817,
+      "step": 2790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001926643598615917,
+      "loss": 4.0894,
+      "step": 2800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019335640138408306,
+      "loss": 4.0932,
+      "step": 2810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001940484429065744,
+      "loss": 4.1004,
+      "step": 2820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019474048442906576,
+      "loss": 4.1235,
+      "step": 2830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001954325259515571,
+      "loss": 4.1434,
+      "step": 2840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019612456747404846,
+      "loss": 4.0648,
+      "step": 2850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001968166089965398,
+      "loss": 4.0025,
+      "step": 2860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019750865051903116,
+      "loss": 4.019,
+      "step": 2870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001982006920415225,
+      "loss": 4.0693,
+      "step": 2880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019889273356401386,
+      "loss": 4.1096,
+      "step": 2890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001995847750865052,
+      "loss": 4.103,
+      "step": 2900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999999737957568,
+      "loss": 3.9842,
+      "step": 2910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999996789980367,
+      "loss": 4.0666,
+      "step": 2920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999990566473897,
+      "loss": 3.9075,
+      "step": 2930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999981067440192,
+      "loss": 4.0124,
+      "step": 2940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999968292882367,
+      "loss": 3.9673,
+      "step": 2950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999952242804605,
+      "loss": 4.004,
+      "step": 2960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999932917212163,
+      "loss": 3.9915,
+      "step": 2970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999991031611137,
+      "loss": 3.9317,
+      "step": 2980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999884439509633,
+      "loss": 3.973,
+      "step": 2990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999855287415423,
+      "loss": 3.9258,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999822859838293,
+      "loss": 3.9318,
+      "step": 3010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999787156788865,
+      "loss": 3.9232,
+      "step": 3020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999748178278827,
+      "loss": 3.8838,
+      "step": 3030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999705924320958,
+      "loss": 3.8711,
+      "step": 3040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999660394929086,
+      "loss": 3.9406,
+      "step": 3050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999961159011813,
+      "loss": 3.9712,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999559509904078,
+      "loss": 3.9403,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999504154303988,
+      "loss": 3.9109,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999944552333599,
+      "loss": 3.8515,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999383617019288,
+      "loss": 3.9066,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999318435374167,
+      "loss": 3.8686,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999249978421964,
+      "loss": 3.842,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999178246185117,
+      "loss": 3.918,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001999910323868711,
+      "loss": 3.8543,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019999024955952521,
+      "loss": 3.8933,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019998943398006987,
+      "loss": 3.8672,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019998858564877222,
+      "loss": 3.869,
+      "step": 3170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998770456591017,
+      "loss": 3.896,
+      "step": 3180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999867907317723,
+      "loss": 3.9441,
+      "step": 3190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998584414665792,
+      "loss": 3.8285,
+      "step": 3200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998486481087713,
+      "loss": 3.8605,
+      "step": 3210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998385272475066,
+      "loss": 3.8374,
+      "step": 3220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998280788861007,
+      "loss": 3.7837,
+      "step": 3230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998173030279757,
+      "loss": 3.771,
+      "step": 3240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019998061996766615,
+      "loss": 3.7959,
+      "step": 3250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997947688357948,
+      "loss": 3.7425,
+      "step": 3260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199978301050912,
+      "loss": 3.9077,
+      "step": 3270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997709247004884,
+      "loss": 3.7723,
+      "step": 3280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999758511413859,
+      "loss": 3.7547,
+      "step": 3290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997457706532977,
+      "loss": 3.756,
+      "step": 3300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997327024229774,
+      "loss": 3.7913,
+      "step": 3310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997193067271794,
+      "loss": 3.7354,
+      "step": 3320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019997055835702907,
+      "loss": 3.7399,
+      "step": 3330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019996915329568068,
+      "loss": 3.6828,
+      "step": 3340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019996771548913303,
+      "loss": 3.6727,
+      "step": 3350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199966244937857,
+      "loss": 3.6893,
+      "step": 3360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999647416423343,
+      "loss": 3.8087,
+      "step": 3370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019996320560305742,
+      "loss": 3.7271,
+      "step": 3380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019996163682052938,
+      "loss": 3.7227,
+      "step": 3390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999600352952641,
+      "loss": 3.712,
+      "step": 3400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019995840102778615,
+      "loss": 3.7924,
+      "step": 3410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019995673401863084,
+      "loss": 3.8106,
+      "step": 3420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019995503426834424,
+      "loss": 3.7323,
+      "step": 3430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019995330177748305,
+      "loss": 3.6854,
+      "step": 3440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019995153654661478,
+      "loss": 3.6432,
+      "step": 3450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999497385763176,
+      "loss": 3.7028,
+      "step": 3460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999479078671805,
+      "loss": 3.6964,
+      "step": 3470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019994604441980315,
+      "loss": 3.809,
+      "step": 3480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019994414823479583,
+      "loss": 3.7011,
+      "step": 3490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019994221931277973,
+      "loss": 3.6711,
+      "step": 3500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019994025765438664,
+      "loss": 3.6232,
+      "step": 3510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019993826326025911,
+      "loss": 3.6765,
+      "step": 3520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999362361310504,
+      "loss": 3.6333,
+      "step": 3530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999341762674245,
+      "loss": 3.6249,
+      "step": 3540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019993208367005615,
+      "loss": 3.6162,
+      "step": 3550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992995833963077,
+      "loss": 3.5603,
+      "step": 3560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992780027684452,
+      "loss": 3.6266,
+      "step": 3570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999256094824043,
+      "loss": 3.6418,
+      "step": 3580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992338595702767,
+      "loss": 3.6158,
+      "step": 3590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019992112970144298,
+      "loss": 3.6985,
+      "step": 3600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991884071638927,
+      "loss": 3.6393,
+      "step": 3610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999165190026163,
+      "loss": 3.5601,
+      "step": 3620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999141645608846,
+      "loss": 3.5953,
+      "step": 3630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019991177739196527,
+      "loss": 3.6503,
+      "step": 3640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990935749664032,
+      "loss": 3.5925,
+      "step": 3650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990690487570236,
+      "loss": 3.6135,
+      "step": 3660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001999044195299548,
+      "loss": 3.5557,
+      "step": 3670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019990190146021165,
+      "loss": 3.5932,
+      "step": 3680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989935066729774,
+      "loss": 3.6551,
+      "step": 3690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989676715204862,
+      "loss": 3.5287,
+      "step": 3700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001998941509153105,
+      "loss": 3.5807,
+      "step": 3710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019989150195794033,
+      "loss": 3.6761,
+      "step": 3720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001998888202808058,
+      "loss": 3.4438,
+      "step": 3730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019988610588478532,
+      "loss": 3.5777,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019988335877076793,
+      "loss": 3.6672,
+      "step": 3750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019988057893965356,
+      "loss": 3.6149,
+      "step": 3760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987776639235265,
+      "loss": 3.5857,
+      "step": 3770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019987492112978652,
+      "loss": 3.5003,
+      "step": 3780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998720431528871,
+      "loss": 3.6546,
+      "step": 3790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986913246259716,
+      "loss": 3.5458,
+      "step": 3800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986618905987001,
+      "loss": 3.6004,
+      "step": 3810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998632129456698,
+      "loss": 3.6526,
+      "step": 3820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019986020412097143,
+      "loss": 3.5539,
+      "step": 3830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985716258676036,
+      "loss": 3.5109,
+      "step": 3840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998540883440329,
+      "loss": 3.5406,
+      "step": 3850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019985098139379601,
+      "loss": 3.5728,
+      "step": 3860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998478417370674,
+      "loss": 3.5154,
+      "step": 3870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019984466937487545,
+      "loss": 3.6113,
+      "step": 3880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998414643082593,
+      "loss": 3.5626,
+      "step": 3890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019983822653826878,
+      "loss": 3.5237,
+      "step": 3900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998349560659644,
+      "loss": 3.4899,
+      "step": 3910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019983165289241746,
+      "loss": 3.49,
+      "step": 3920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019982831701870988,
+      "loss": 3.5472,
+      "step": 3930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019982494844593434,
+      "loss": 3.5512,
+      "step": 3940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019982154717519427,
+      "loss": 3.531,
+      "step": 3950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001998181132076037,
+      "loss": 3.4923,
+      "step": 3960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019981464654428752,
+      "loss": 3.5375,
+      "step": 3970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019981114718638118,
+      "loss": 3.4776,
+      "step": 3980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019980761513503092,
+      "loss": 3.5185,
+      "step": 3990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019980405039139368,
+      "loss": 3.5635,
+      "step": 4000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019980045295663713,
+      "loss": 3.5751,
+      "step": 4010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019979682283193954,
+      "loss": 3.5863,
+      "step": 4020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019979316001849005,
+      "loss": 3.4812,
+      "step": 4030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019978946451748838,
+      "loss": 3.457,
+      "step": 4040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019978573633014506,
+      "loss": 3.3768,
+      "step": 4050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019978197545768119,
+      "loss": 3.4967,
+      "step": 4060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001997781819013287,
+      "loss": 3.4793,
+      "step": 4070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019977435566233016,
+      "loss": 3.3565,
+      "step": 4080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019977049674193892,
+      "loss": 3.4976,
+      "step": 4090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019976660514141892,
+      "loss": 3.4365,
+      "step": 4100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001997626808620449,
+      "loss": 3.5347,
+      "step": 4110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001997587239051022,
+      "loss": 3.5448,
+      "step": 4120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199754734271887,
+      "loss": 3.4276,
+      "step": 4130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019975071196370617,
+      "loss": 3.4771,
+      "step": 4140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019974665698187713,
+      "loss": 3.5009,
+      "step": 4150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019974256932772815,
+      "loss": 3.3931,
+      "step": 4160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019973844900259813,
+      "loss": 3.3643,
+      "step": 4170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019973429600783667,
+      "loss": 3.4551,
+      "step": 4180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001997301103448042,
+      "loss": 3.3744,
+      "step": 4190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019972589201487165,
+      "loss": 3.4875,
+      "step": 4200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019972164101942074,
+      "loss": 3.4148,
+      "step": 4210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019971735735984397,
+      "loss": 3.4092,
+      "step": 4220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019971304103754442,
+      "loss": 3.4911,
+      "step": 4230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001997086920539359,
+      "loss": 3.5266,
+      "step": 4240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199704310410443,
+      "loss": 3.4856,
+      "step": 4250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019969989610850087,
+      "loss": 3.3837,
+      "step": 4260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019969544914955542,
+      "loss": 3.4171,
+      "step": 4270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019969096953506335,
+      "loss": 3.3912,
+      "step": 4280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001996864572664919,
+      "loss": 3.4548,
+      "step": 4290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019968191234531909,
+      "loss": 3.4519,
+      "step": 4300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019967733477303364,
+      "loss": 3.5497,
+      "step": 4310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001996727245511349,
+      "loss": 3.4923,
+      "step": 4320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019966808168113304,
+      "loss": 3.4534,
+      "step": 4330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001996634061645488,
+      "loss": 3.4985,
+      "step": 4340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019965869800291362,
+      "loss": 3.4667,
+      "step": 4350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019965395719776977,
+      "loss": 3.4741,
+      "step": 4360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019964918375067005,
+      "loss": 3.3865,
+      "step": 4370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019964437766317804,
+      "loss": 3.4611,
+      "step": 4380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019963953893686794,
+      "loss": 3.4155,
+      "step": 4390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019963466757332476,
+      "loss": 3.4273,
+      "step": 4400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001996297635741441,
+      "loss": 3.4075,
+      "step": 4410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019962482694093226,
+      "loss": 3.3705,
+      "step": 4420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019961985767530627,
+      "loss": 3.4634,
+      "step": 4430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019961485577889382,
+      "loss": 3.4207,
+      "step": 4440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019960982125333333,
+      "loss": 3.3596,
+      "step": 4450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019960475410027382,
+      "loss": 3.3294,
+      "step": 4460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019959965432137506,
+      "loss": 3.4237,
+      "step": 4470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019959452191830752,
+      "loss": 3.381,
+      "step": 4480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019958935689275235,
+      "loss": 3.4286,
+      "step": 4490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019958415924640135,
+      "loss": 3.3818,
+      "step": 4500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000199578928980957,
+      "loss": 3.3596,
+      "step": 4510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019957366609813252,
+      "loss": 3.3753,
+      "step": 4520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019956837059965178,
+      "loss": 3.4069,
+      "step": 4530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001995630424872493,
+      "loss": 3.3797,
+      "step": 4540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001995576817626704,
+      "loss": 3.4151,
+      "step": 4550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019955228842767096,
+      "loss": 3.3194,
+      "step": 4560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019954686248401752,
+      "loss": 3.4055,
+      "step": 4570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019954140393348745,
+      "loss": 3.4778,
+      "step": 4580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001995359127778687,
+      "loss": 3.4079,
+      "step": 4590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019953038901895988,
+      "loss": 3.3722,
+      "step": 4600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019952483265857034,
+      "loss": 3.2981,
+      "step": 4610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019951924369852008,
+      "loss": 3.4218,
+      "step": 4620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019951362214063977,
+      "loss": 3.3752,
+      "step": 4630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001995079679867708,
+      "loss": 3.3465,
+      "step": 4640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019950228123876516,
+      "loss": 3.338,
+      "step": 4650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019949656189848555,
+      "loss": 3.4215,
+      "step": 4660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019949080996780545,
+      "loss": 3.3762,
+      "step": 4670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019948502544860883,
+      "loss": 3.3842,
+      "step": 4680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019947920834279047,
+      "loss": 3.2935,
+      "step": 4690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019947335865225574,
+      "loss": 3.3565,
+      "step": 4700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001994674763789208,
+      "loss": 3.3427,
+      "step": 4710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019946156152471231,
+      "loss": 3.3501,
+      "step": 4720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001994556140915678,
+      "loss": 3.3886,
+      "step": 4730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019944963408143526,
+      "loss": 3.4035,
+      "step": 4740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019944362149627355,
+      "loss": 3.3369,
+      "step": 4750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019943757633805208,
+      "loss": 3.3126,
+      "step": 4760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000199431498608751,
+      "loss": 3.3814,
+      "step": 4770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000199425388310361,
+      "loss": 3.3848,
+      "step": 4780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019941924544488357,
+      "loss": 3.3691,
+      "step": 4790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019941307001433085,
+      "loss": 3.3897,
+      "step": 4800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001994068620207256,
+      "loss": 3.3267,
+      "step": 4810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019940062146610123,
+      "loss": 3.3391,
+      "step": 4820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019939434835250194,
+      "loss": 3.2958,
+      "step": 4830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019938804268198243,
+      "loss": 3.3767,
+      "step": 4840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019938170445660817,
+      "loss": 3.2673,
+      "step": 4850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019937533367845526,
+      "loss": 3.2909,
+      "step": 4860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019936893034961045,
+      "loss": 3.377,
+      "step": 4870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019936249447217126,
+      "loss": 3.3764,
+      "step": 4880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019935602604824565,
+      "loss": 3.3943,
+      "step": 4890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019934952507995245,
+      "loss": 3.4182,
+      "step": 4900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019934299156942106,
+      "loss": 3.2562,
+      "step": 4910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019933642551879155,
+      "loss": 3.2421,
+      "step": 4920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001993298269302146,
+      "loss": 3.2552,
+      "step": 4930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019932319580585173,
+      "loss": 3.3118,
+      "step": 4940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019931653214787484,
+      "loss": 3.3203,
+      "step": 4950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001993098359584667,
+      "loss": 3.2543,
+      "step": 4960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001993031072398207,
+      "loss": 3.3056,
+      "step": 4970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019929634599414074,
+      "loss": 3.3065,
+      "step": 4980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019928955222364158,
+      "loss": 3.3088,
+      "step": 4990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019928272593054854,
+      "loss": 3.2899,
+      "step": 5000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019927586711709756,
+      "loss": 3.2867,
+      "step": 5010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019926897578553528,
+      "loss": 3.3519,
+      "step": 5020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019926205193811894,
+      "loss": 3.3988,
+      "step": 5030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019925509557711652,
+      "loss": 3.3081,
+      "step": 5040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019924810670480657,
+      "loss": 3.3955,
+      "step": 5050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019924108532347833,
+      "loss": 3.2414,
+      "step": 5060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019923403143543166,
+      "loss": 3.3343,
+      "step": 5070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001992269450429771,
+      "loss": 3.2954,
+      "step": 5080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019921982614843579,
+      "loss": 3.3105,
+      "step": 5090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001992126747541396,
+      "loss": 3.224,
+      "step": 5100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019920549086243093,
+      "loss": 3.3824,
+      "step": 5110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019919827447566291,
+      "loss": 3.2393,
+      "step": 5120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019919102559619927,
+      "loss": 3.3379,
+      "step": 5130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019918374422641443,
+      "loss": 3.275,
+      "step": 5140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019917643036869345,
+      "loss": 3.254,
+      "step": 5150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019916908402543192,
+      "loss": 3.3366,
+      "step": 5160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019916170519903626,
+      "loss": 3.2938,
+      "step": 5170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019915429389192335,
+      "loss": 3.2917,
+      "step": 5180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019914685010652083,
+      "loss": 3.2354,
+      "step": 5190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001991393738452669,
+      "loss": 3.2558,
+      "step": 5200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019913186511061047,
+      "loss": 3.3324,
+      "step": 5210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019912432390501103,
+      "loss": 3.3408,
+      "step": 5220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001991167502309387,
+      "loss": 3.229,
+      "step": 5230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001991091440908743,
+      "loss": 3.2444,
+      "step": 5240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001991015054873092,
+      "loss": 3.2503,
+      "step": 5250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001990938344227455,
+      "loss": 3.2946,
+      "step": 5260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019908613089969585,
+      "loss": 3.2723,
+      "step": 5270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019907839492068358,
+      "loss": 3.3381,
+      "step": 5280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001990706264882426,
+      "loss": 3.3007,
+      "step": 5290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019906282560491753,
+      "loss": 3.2017,
+      "step": 5300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019905499227326355,
+      "loss": 3.2486,
+      "step": 5310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019904712649584645,
+      "loss": 3.2933,
+      "step": 5320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019903922827524276,
+      "loss": 3.1922,
+      "step": 5330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019903129761403955,
+      "loss": 3.2654,
+      "step": 5340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019902333451483447,
+      "loss": 3.256,
+      "step": 5350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019901533898023595,
+      "loss": 3.3318,
+      "step": 5360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019900731101286288,
+      "loss": 3.2453,
+      "step": 5370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019899925061534485,
+      "loss": 3.2682,
+      "step": 5380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019899115779032213,
+      "loss": 3.2109,
+      "step": 5390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001989830325404455,
+      "loss": 3.2141,
+      "step": 5400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001989748748683764,
+      "loss": 3.2428,
+      "step": 5410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019896668477678694,
+      "loss": 3.1986,
+      "step": 5420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001989584622683598,
+      "loss": 3.3107,
+      "step": 5430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019895020734578826,
+      "loss": 3.1527,
+      "step": 5440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019894192001177624,
+      "loss": 3.1769,
+      "step": 5450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001989336002690383,
+      "loss": 3.2245,
+      "step": 5460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001989252481202996,
+      "loss": 3.2321,
+      "step": 5470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019891686356829594,
+      "loss": 3.1875,
+      "step": 5480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019890844661577366,
+      "loss": 3.2488,
+      "step": 5490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019889999726548975,
+      "loss": 3.1772,
+      "step": 5500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001988915155202119,
+      "loss": 3.2611,
+      "step": 5510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019888300138271823,
+      "loss": 3.2415,
+      "step": 5520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019887445485579762,
+      "loss": 3.3145,
+      "step": 5530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019886587594224953,
+      "loss": 3.2168,
+      "step": 5540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019885726464488398,
+      "loss": 3.2943,
+      "step": 5550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019884862096652162,
+      "loss": 3.2841,
+      "step": 5560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019883994490999373,
+      "loss": 3.2148,
+      "step": 5570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019883123647814217,
+      "loss": 3.1774,
+      "step": 5580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019882249567381943,
+      "loss": 3.0887,
+      "step": 5590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019881372249988858,
+      "loss": 3.2588,
+      "step": 5600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001988049169592233,
+      "loss": 3.2019,
+      "step": 5610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019879607905470785,
+      "loss": 3.1482,
+      "step": 5620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019878720878923717,
+      "loss": 3.1703,
+      "step": 5630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019877830616571665,
+      "loss": 3.2301,
+      "step": 5640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019876937118706246,
+      "loss": 3.2534,
+      "step": 5650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001987604038562012,
+      "loss": 3.2172,
+      "step": 5660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001987514041760702,
+      "loss": 3.2074,
+      "step": 5670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019874237214961735,
+      "loss": 3.3045,
+      "step": 5680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019873330777980109,
+      "loss": 3.1463,
+      "step": 5690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019872421106959044,
+      "loss": 3.1664,
+      "step": 5700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019871508202196512,
+      "loss": 3.2653,
+      "step": 5710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019870592063991535,
+      "loss": 3.2599,
+      "step": 5720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019869672692644197,
+      "loss": 3.304,
+      "step": 5730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019868750088455642,
+      "loss": 3.2631,
+      "step": 5740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001986782425172807,
+      "loss": 3.252,
+      "step": 5750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019866895182764745,
+      "loss": 3.223,
+      "step": 5760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019865962881869978,
+      "loss": 3.2239,
+      "step": 5770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019865027349349158,
+      "loss": 3.1763,
+      "step": 5780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019864088585508716,
+      "loss": 3.2352,
+      "step": 5790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019863146590656146,
+      "loss": 3.1935,
+      "step": 5800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019862296033028464,
+      "loss": 3.1874,
+      "step": 5810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019861347900103788,
+      "loss": 3.1787,
+      "step": 5820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001986039653706471,
+      "loss": 3.2122,
+      "step": 5830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019859441944222844,
+      "loss": 3.2559,
+      "step": 5840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019858484121890876,
+      "loss": 3.23,
+      "step": 5850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001985752307038254,
+      "loss": 3.1969,
+      "step": 5860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019856558790012628,
+      "loss": 3.2084,
+      "step": 5870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019855591281097,
+      "loss": 3.2568,
+      "step": 5880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019854620543952565,
+      "loss": 3.1049,
+      "step": 5890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019853646578897285,
+      "loss": 3.2152,
+      "step": 5900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019852669386250188,
+      "loss": 3.1283,
+      "step": 5910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019851688966331362,
+      "loss": 3.2262,
+      "step": 5920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001985070531946194,
+      "loss": 3.1105,
+      "step": 5930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001984971844596412,
+      "loss": 3.1419,
+      "step": 5940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019848728346161156,
+      "loss": 3.2079,
+      "step": 5950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001984773502037736,
+      "loss": 3.1263,
+      "step": 5960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019846738468938099,
+      "loss": 3.1069,
+      "step": 5970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001984573869216979,
+      "loss": 3.1432,
+      "step": 5980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019844735690399917,
+      "loss": 3.1582,
+      "step": 5990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019843729463957019,
+      "loss": 3.1746,
+      "step": 6000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019842720013170685,
+      "loss": 3.2542,
+      "step": 6010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019841707338371566,
+      "loss": 3.2188,
+      "step": 6020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019840691439891367,
+      "loss": 3.2271,
+      "step": 6030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019839672318062846,
+      "loss": 3.1342,
+      "step": 6040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019838649973219822,
+      "loss": 3.2106,
+      "step": 6050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019837624405697163,
+      "loss": 3.1548,
+      "step": 6060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000198365956158308,
+      "loss": 3.1651,
+      "step": 6070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019835563603957717,
+      "loss": 3.2291,
+      "step": 6080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001983452837041595,
+      "loss": 3.1544,
+      "step": 6090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019833489915544597,
+      "loss": 3.1993,
+      "step": 6100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019832448239683802,
+      "loss": 3.1555,
+      "step": 6110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001983140334317477,
+      "loss": 3.1418,
+      "step": 6120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019830355226359766,
+      "loss": 3.1266,
+      "step": 6130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000198293038895821,
+      "loss": 3.2538,
+      "step": 6140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019828249333186135,
+      "loss": 3.1589,
+      "step": 6150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019827191557517302,
+      "loss": 3.1727,
+      "step": 6160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019826130562922072,
+      "loss": 3.2665,
+      "step": 6170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019825066349747985,
+      "loss": 3.1745,
+      "step": 6180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001982399891834362,
+      "loss": 3.1506,
+      "step": 6190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019822928269058623,
+      "loss": 3.0515,
+      "step": 6200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019821854402243686,
+      "loss": 3.2564,
+      "step": 6210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001982077731825055,
+      "loss": 3.1443,
+      "step": 6220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001981969701743203,
+      "loss": 3.183,
+      "step": 6230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019818613500141978,
+      "loss": 3.1295,
+      "step": 6240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000198175267667353,
+      "loss": 3.1562,
+      "step": 6250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019816436817567955,
+      "loss": 3.1277,
+      "step": 6260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001981534365299697,
+      "loss": 3.2041,
+      "step": 6270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019814247273380406,
+      "loss": 3.1994,
+      "step": 6280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001981314767907739,
+      "loss": 3.1519,
+      "step": 6290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019812044870448094,
+      "loss": 3.1597,
+      "step": 6300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001981093884785375,
+      "loss": 3.2095,
+      "step": 6310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019809829611656637,
+      "loss": 3.1634,
+      "step": 6320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019808717162220086,
+      "loss": 3.1504,
+      "step": 6330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001980760149990849,
+      "loss": 3.1078,
+      "step": 6340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019806482625087278,
+      "loss": 3.1072,
+      "step": 6350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001980536053812295,
+      "loss": 3.0916,
+      "step": 6360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019804235239383045,
+      "loss": 3.1548,
+      "step": 6370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019803106729236153,
+      "loss": 3.0599,
+      "step": 6380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019801975008051933,
+      "loss": 3.1896,
+      "step": 6390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019800840076201074,
+      "loss": 3.161,
+      "step": 6400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019799701934055327,
+      "loss": 3.1419,
+      "step": 6410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019798560581987503,
+      "loss": 3.0797,
+      "step": 6420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019797416020371446,
+      "loss": 3.183,
+      "step": 6430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001979626824958206,
+      "loss": 3.179,
+      "step": 6440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019795117269995306,
+      "loss": 3.1881,
+      "step": 6450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019793963081988185,
+      "loss": 3.144,
+      "step": 6460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019792805685938763,
+      "loss": 3.0872,
+      "step": 6470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019791645082226143,
+      "loss": 3.0941,
+      "step": 6480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019790481271230484,
+      "loss": 3.1313,
+      "step": 6490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019789314253333,
+      "loss": 3.11,
+      "step": 6500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019788144028915947,
+      "loss": 3.1541,
+      "step": 6510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019786970598362636,
+      "loss": 3.1038,
+      "step": 6520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001978579396205743,
+      "loss": 3.1532,
+      "step": 6530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001978461412038574,
+      "loss": 3.0607,
+      "step": 6540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019783431073734021,
+      "loss": 3.1238,
+      "step": 6550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001978224482248979,
+      "loss": 3.2191,
+      "step": 6560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019781055367041608,
+      "loss": 3.1481,
+      "step": 6570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019779862707779077,
+      "loss": 3.1215,
+      "step": 6580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019778666845092865,
+      "loss": 3.1182,
+      "step": 6590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019777467779374675,
+      "loss": 3.1303,
+      "step": 6600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019776265511017263,
+      "loss": 3.1505,
+      "step": 6610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001977506004041444,
+      "loss": 3.127,
+      "step": 6620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019773851367961066,
+      "loss": 3.1541,
+      "step": 6630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019772639494053032,
+      "loss": 3.1242,
+      "step": 6640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019771424419087305,
+      "loss": 3.2074,
+      "step": 6650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019770206143461874,
+      "loss": 3.0326,
+      "step": 6660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019768984667575796,
+      "loss": 3.0881,
+      "step": 6670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001976775999182917,
+      "loss": 3.1403,
+      "step": 6680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019766532116623139,
+      "loss": 3.1631,
+      "step": 6690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019765301042359897,
+      "loss": 3.1143,
+      "step": 6700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001976406676944269,
+      "loss": 3.0861,
+      "step": 6710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019762829298275802,
+      "loss": 3.0551,
+      "step": 6720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019761588629264577,
+      "loss": 3.1081,
+      "step": 6730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019760344762815397,
+      "loss": 3.1582,
+      "step": 6740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001975909769933569,
+      "loss": 3.1374,
+      "step": 6750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019757847439233944,
+      "loss": 3.1343,
+      "step": 6760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001975659398291968,
+      "loss": 3.1131,
+      "step": 6770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001975533733080347,
+      "loss": 3.0765,
+      "step": 6780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001975407748329694,
+      "loss": 3.0863,
+      "step": 6790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001975281444081275,
+      "loss": 3.0742,
+      "step": 6800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019751548203764618,
+      "loss": 3.0474,
+      "step": 6810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019750278772567303,
+      "loss": 3.1571,
+      "step": 6820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001974900614763661,
+      "loss": 3.1361,
+      "step": 6830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001974773032938939,
+      "loss": 3.225,
+      "step": 6840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001974645131824355,
+      "loss": 3.1431,
+      "step": 6850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019745169114618018,
+      "loss": 3.1303,
+      "step": 6860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019743883718932796,
+      "loss": 3.1156,
+      "step": 6870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019742595131608915,
+      "loss": 3.0963,
+      "step": 6880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001974130335306846,
+      "loss": 3.0889,
+      "step": 6890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001974000838373455,
+      "loss": 3.1806,
+      "step": 6900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001973871022403136,
+      "loss": 3.1749,
+      "step": 6910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019737408874384108,
+      "loss": 3.1413,
+      "step": 6920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001973610433521905,
+      "loss": 3.1434,
+      "step": 6930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019734796606963494,
+      "loss": 3.1044,
+      "step": 6940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019733485690045792,
+      "loss": 3.1476,
+      "step": 6950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019732171584895341,
+      "loss": 3.0641,
+      "step": 6960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019730854291942574,
+      "loss": 3.1433,
+      "step": 6970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019729533811618975,
+      "loss": 3.1025,
+      "step": 6980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019728210144357076,
+      "loss": 3.0796,
+      "step": 6990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019726883290590444,
+      "loss": 3.1143,
+      "step": 7000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019725553250753697,
+      "loss": 3.1092,
+      "step": 7010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001972422002528249,
+      "loss": 3.152,
+      "step": 7020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001972288361461353,
+      "loss": 3.0712,
+      "step": 7030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019721544019184558,
+      "loss": 3.1106,
+      "step": 7040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019720201239434366,
+      "loss": 3.0513,
+      "step": 7050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019718855275802782,
+      "loss": 3.14,
+      "step": 7060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019717506128730686,
+      "loss": 3.1225,
+      "step": 7070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019716153798659985,
+      "loss": 3.0405,
+      "step": 7080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019714798286033652,
+      "loss": 3.0569,
+      "step": 7090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019713439591295678,
+      "loss": 3.1378,
+      "step": 7100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001971207771489111,
+      "loss": 3.1517,
+      "step": 7110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019710712657266042,
+      "loss": 3.1643,
+      "step": 7120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019709344418867596,
+      "loss": 3.0738,
+      "step": 7130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019707973000143944,
+      "loss": 3.0676,
+      "step": 7140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019706598401544298,
+      "loss": 3.1118,
+      "step": 7150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019705220623518913,
+      "loss": 3.233,
+      "step": 7160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019703839666519084,
+      "loss": 3.0361,
+      "step": 7170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019702455530997147,
+      "loss": 3.1511,
+      "step": 7180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001970106821740648,
+      "loss": 3.0971,
+      "step": 7190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019699677726201507,
+      "loss": 3.0902,
+      "step": 7200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019698284057837677,
+      "loss": 3.0847,
+      "step": 7210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019696887212771496,
+      "loss": 3.0655,
+      "step": 7220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019695487191460508,
+      "loss": 3.1484,
+      "step": 7230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001969408399436329,
+      "loss": 3.1138,
+      "step": 7240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019692677621939466,
+      "loss": 3.1026,
+      "step": 7250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019691268074649695,
+      "loss": 3.0968,
+      "step": 7260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019689855352955678,
+      "loss": 3.0486,
+      "step": 7270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001968843945732016,
+      "loss": 3.1301,
+      "step": 7280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019687020388206922,
+      "loss": 3.0379,
+      "step": 7290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019685598146080774,
+      "loss": 3.1262,
+      "step": 7300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019684172731407592,
+      "loss": 2.9428,
+      "step": 7310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019682744144654263,
+      "loss": 3.1398,
+      "step": 7320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001968131238628873,
+      "loss": 3.0266,
+      "step": 7330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019679877456779967,
+      "loss": 3.1056,
+      "step": 7340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019678439356597994,
+      "loss": 3.1434,
+      "step": 7350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001967699808621386,
+      "loss": 3.0334,
+      "step": 7360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001967555364609966,
+      "loss": 3.1146,
+      "step": 7370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019674106036728523,
+      "loss": 3.0151,
+      "step": 7380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019672655258574622,
+      "loss": 3.0994,
+      "step": 7390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019671201312113158,
+      "loss": 3.0733,
+      "step": 7400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001966974419782038,
+      "loss": 2.9758,
+      "step": 7410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019668283916173568,
+      "loss": 2.9764,
+      "step": 7420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019666820467651044,
+      "loss": 3.1575,
+      "step": 7430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019665353852732162,
+      "loss": 3.0078,
+      "step": 7440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001966388407189732,
+      "loss": 3.1733,
+      "step": 7450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019662411125627945,
+      "loss": 3.0998,
+      "step": 7460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019660935014406504,
+      "loss": 3.0089,
+      "step": 7470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001965945573871651,
+      "loss": 3.0505,
+      "step": 7480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019657973299042495,
+      "loss": 3.0794,
+      "step": 7490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001965648769587004,
+      "loss": 3.0451,
+      "step": 7500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001965499892968576,
+      "loss": 3.0551,
+      "step": 7510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019653507000977306,
+      "loss": 3.0634,
+      "step": 7520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001965201191023336,
+      "loss": 3.0605,
+      "step": 7530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019650513657943646,
+      "loss": 3.0078,
+      "step": 7540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019649012244598918,
+      "loss": 3.0178,
+      "step": 7550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019647507670690972,
+      "loss": 3.0934,
+      "step": 7560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001964599993671264,
+      "loss": 3.0159,
+      "step": 7570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001964448904315777,
+      "loss": 3.1005,
+      "step": 7580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019642974990521274,
+      "loss": 3.0167,
+      "step": 7590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001964145777929908,
+      "loss": 3.0454,
+      "step": 7600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019639937409988157,
+      "loss": 3.0773,
+      "step": 7610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000196384138830865,
+      "loss": 3.1258,
+      "step": 7620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019636887199093152,
+      "loss": 3.0474,
+      "step": 7630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001963535735850818,
+      "loss": 3.1178,
+      "step": 7640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019633824361832688,
+      "loss": 3.0835,
+      "step": 7650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019632288209568815,
+      "loss": 3.0319,
+      "step": 7660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019630748902219728,
+      "loss": 3.0051,
+      "step": 7670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001962920644028964,
+      "loss": 3.0526,
+      "step": 7680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001962766082428378,
+      "loss": 3.1007,
+      "step": 7690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019626112054708426,
+      "loss": 2.9811,
+      "step": 7700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019624560132070878,
+      "loss": 3.0545,
+      "step": 7710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019623005056879477,
+      "loss": 3.0022,
+      "step": 7720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019621446829643584,
+      "loss": 3.032,
+      "step": 7730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019619885450873612,
+      "loss": 2.9852,
+      "step": 7740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001961832092108099,
+      "loss": 3.0601,
+      "step": 7750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019616753240778186,
+      "loss": 3.0654,
+      "step": 7760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019615182410478695,
+      "loss": 3.0509,
+      "step": 7770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019613608430697047,
+      "loss": 3.059,
+      "step": 7780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001961203130194881,
+      "loss": 3.143,
+      "step": 7790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019610451024750576,
+      "loss": 3.0043,
+      "step": 7800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019608867599619963,
+      "loss": 3.0687,
+      "step": 7810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019607281027075635,
+      "loss": 3.0853,
+      "step": 7820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019605691307637274,
+      "loss": 3.0864,
+      "step": 7830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019604098441825595,
+      "loss": 3.0374,
+      "step": 7840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001960250243016235,
+      "loss": 3.1481,
+      "step": 7850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001960090327317032,
+      "loss": 3.0954,
+      "step": 7860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019599300971373306,
+      "loss": 3.1255,
+      "step": 7870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019597695525296155,
+      "loss": 3.044,
+      "step": 7880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001959608693546473,
+      "loss": 3.0297,
+      "step": 7890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019594475202405932,
+      "loss": 3.122,
+      "step": 7900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001959286032664769,
+      "loss": 3.0691,
+      "step": 7910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019591242308718957,
+      "loss": 3.1496,
+      "step": 7920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019589621149149723,
+      "loss": 2.9686,
+      "step": 7930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019587996848471003,
+      "loss": 3.0222,
+      "step": 7940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019586369407214842,
+      "loss": 3.0509,
+      "step": 7950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019584738825914315,
+      "loss": 3.0222,
+      "step": 7960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019583105105103521,
+      "loss": 3.0021,
+      "step": 7970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001958146824531759,
+      "loss": 3.0103,
+      "step": 7980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019579828247092684,
+      "loss": 3.046,
+      "step": 7990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019578185110965984,
+      "loss": 3.0718,
+      "step": 8000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001957653883747571,
+      "loss": 3.0021,
+      "step": 8010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019574889427161104,
+      "loss": 3.095,
+      "step": 8020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019573236880562425,
+      "loss": 3.0722,
+      "step": 8030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019571581198220983,
+      "loss": 3.0323,
+      "step": 8040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019569922380679094,
+      "loss": 2.9698,
+      "step": 8050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001956826042848011,
+      "loss": 2.9884,
+      "step": 8060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019566595342168408,
+      "loss": 2.9934,
+      "step": 8070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019564927122289397,
+      "loss": 3.0421,
+      "step": 8080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019563255769389503,
+      "loss": 2.9711,
+      "step": 8090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001956158128401618,
+      "loss": 3.1048,
+      "step": 8100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019559903666717916,
+      "loss": 3.0151,
+      "step": 8110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019558222918044215,
+      "loss": 3.0414,
+      "step": 8120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019556539038545617,
+      "loss": 3.0675,
+      "step": 8130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001955485202877368,
+      "loss": 2.9681,
+      "step": 8140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001955316188928099,
+      "loss": 3.075,
+      "step": 8150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001955146862062115,
+      "loss": 3.0497,
+      "step": 8160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019549772223348804,
+      "loss": 3.0303,
+      "step": 8170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019548072698019606,
+      "loss": 2.9958,
+      "step": 8180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019546370045190245,
+      "loss": 3.0126,
+      "step": 8190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001954466426541843,
+      "loss": 3.0382,
+      "step": 8200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001954295535926289,
+      "loss": 3.061,
+      "step": 8210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001954124332728339,
+      "loss": 3.0589,
+      "step": 8220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000195395281700407,
+      "loss": 3.0847,
+      "step": 8230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019537809888096636,
+      "loss": 2.9358,
+      "step": 8240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001953608848201402,
+      "loss": 3.1177,
+      "step": 8250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019534363952356708,
+      "loss": 3.0948,
+      "step": 8260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001953263629968957,
+      "loss": 2.9618,
+      "step": 8270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001953090552457851,
+      "loss": 3.0322,
+      "step": 8280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019529171627590442,
+      "loss": 2.9583,
+      "step": 8290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019527434609293312,
+      "loss": 2.9391,
+      "step": 8300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019525694470256085,
+      "loss": 3.0146,
+      "step": 8310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019523951211048755,
+      "loss": 3.0339,
+      "step": 8320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019522204832242323,
+      "loss": 2.9596,
+      "step": 8330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019520455334408825,
+      "loss": 3.057,
+      "step": 8340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019518702718121316,
+      "loss": 2.9783,
+      "step": 8350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019516946983953864,
+      "loss": 3.0599,
+      "step": 8360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001951518813248157,
+      "loss": 2.9684,
+      "step": 8370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019513426164280553,
+      "loss": 2.9378,
+      "step": 8380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001951166107992795,
+      "loss": 2.9944,
+      "step": 8390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019509892880001915,
+      "loss": 3.0276,
+      "step": 8400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001950812156508163,
+      "loss": 3.0099,
+      "step": 8410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019506347135747298,
+      "loss": 3.0482,
+      "step": 8420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019504569592580136,
+      "loss": 3.0347,
+      "step": 8430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001950278893616238,
+      "loss": 3.0826,
+      "step": 8440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019501005167077296,
+      "loss": 3.003,
+      "step": 8450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001949921828590916,
+      "loss": 3.0255,
+      "step": 8460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019497428293243268,
+      "loss": 3.0936,
+      "step": 8470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019495635189665942,
+      "loss": 3.0192,
+      "step": 8480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019493838975764515,
+      "loss": 3.0006,
+      "step": 8490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019492039652127345,
+      "loss": 3.0142,
+      "step": 8500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019490237219343804,
+      "loss": 2.9863,
+      "step": 8510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019488431678004283,
+      "loss": 3.1561,
+      "step": 8520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019486623028700195,
+      "loss": 3.0416,
+      "step": 8530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001948481127202397,
+      "loss": 2.9737,
+      "step": 8540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001948299640856905,
+      "loss": 3.0237,
+      "step": 8550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019481178438929904,
+      "loss": 2.9978,
+      "step": 8560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001947935736370201,
+      "loss": 3.0166,
+      "step": 8570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019477533183481866,
+      "loss": 3.0747,
+      "step": 8580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001947570589886699,
+      "loss": 3.0458,
+      "step": 8590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019473875510455914,
+      "loss": 3.0557,
+      "step": 8600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001947204201884819,
+      "loss": 2.9801,
+      "step": 8610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019470205424644376,
+      "loss": 2.9823,
+      "step": 8620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019468365728446063,
+      "loss": 3.0997,
+      "step": 8630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019466522930855842,
+      "loss": 3.0054,
+      "step": 8640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019464677032477332,
+      "loss": 3.0284,
+      "step": 8650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001946282803391516,
+      "loss": 3.02,
+      "step": 8660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001946097593577497,
+      "loss": 2.989,
+      "step": 8670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019459120738663427,
+      "loss": 2.9651,
+      "step": 8680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000194572624431882,
+      "loss": 3.0457,
+      "step": 8690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019455401049957985,
+      "loss": 2.9985,
+      "step": 8700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019453536559582485,
+      "loss": 3.0568,
+      "step": 8710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019451668972672419,
+      "loss": 3.0005,
+      "step": 8720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019449798289839523,
+      "loss": 2.9818,
+      "step": 8730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001944792451169654,
+      "loss": 2.987,
+      "step": 8740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019446047638857234,
+      "loss": 3.0396,
+      "step": 8750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019444167671936382,
+      "loss": 3.0405,
+      "step": 8760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019442284611549774,
+      "loss": 3.0238,
+      "step": 8770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019440398458314207,
+      "loss": 3.0507,
+      "step": 8780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019438509212847498,
+      "loss": 2.8835,
+      "step": 8790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001943661687576848,
+      "loss": 2.947,
+      "step": 8800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001943472144769699,
+      "loss": 2.9981,
+      "step": 8810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019432822929253875,
+      "loss": 2.9992,
+      "step": 8820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019430921321061012,
+      "loss": 2.9892,
+      "step": 8830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001942901662374127,
+      "loss": 2.9797,
+      "step": 8840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019427108837918546,
+      "loss": 2.9886,
+      "step": 8850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019425197964217735,
+      "loss": 3.0563,
+      "step": 8860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001942328400326475,
+      "loss": 2.9298,
+      "step": 8870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019421366955686518,
+      "loss": 2.9856,
+      "step": 8880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001941944682211097,
+      "loss": 3.034,
+      "step": 8890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019417523603167052,
+      "loss": 3.0259,
+      "step": 8900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019415597299484724,
+      "loss": 3.0395,
+      "step": 8910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019413667911694947,
+      "loss": 3.0012,
+      "step": 8920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000194117354404297,
+      "loss": 2.9629,
+      "step": 8930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019409799886321977,
+      "loss": 3.0074,
+      "step": 8940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001940786125000576,
+      "loss": 3.0157,
+      "step": 8950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019405919532116067,
+      "loss": 2.9597,
+      "step": 8960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001940397473328891,
+      "loss": 3.0023,
+      "step": 8970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019402026854161314,
+      "loss": 2.9688,
+      "step": 8980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019400075895371313,
+      "loss": 2.9743,
+      "step": 8990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019398121857557948,
+      "loss": 2.9644,
+      "step": 9000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019396164741361272,
+      "loss": 3.0216,
+      "step": 9010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019394204547422343,
+      "loss": 3.0599,
+      "step": 9020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019392241276383227,
+      "loss": 2.9999,
+      "step": 9030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019390274928887002,
+      "loss": 2.9275,
+      "step": 9040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001938830550557775,
+      "loss": 2.9654,
+      "step": 9050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019386333007100564,
+      "loss": 2.9792,
+      "step": 9060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019384357434101536,
+      "loss": 2.9648,
+      "step": 9070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019382378787227779,
+      "loss": 2.9937,
+      "step": 9080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019380397067127395,
+      "loss": 2.979,
+      "step": 9090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001937841227444951,
+      "loss": 2.9957,
+      "step": 9100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001937642440984425,
+      "loss": 2.9197,
+      "step": 9110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019374433473962738,
+      "loss": 2.9366,
+      "step": 9120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019372439467457117,
+      "loss": 2.9739,
+      "step": 9130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019370442390980532,
+      "loss": 2.9835,
+      "step": 9140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019368442245187124,
+      "loss": 2.9313,
+      "step": 9150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019366439030732053,
+      "loss": 2.9542,
+      "step": 9160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019364432748271475,
+      "loss": 2.9474,
+      "step": 9170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019362423398462558,
+      "loss": 2.9174,
+      "step": 9180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001936041098196346,
+      "loss": 2.9661,
+      "step": 9190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001935839549943337,
+      "loss": 2.9808,
+      "step": 9200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001935637695153245,
+      "loss": 3.0645,
+      "step": 9210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019354355338921893,
+      "loss": 2.9555,
+      "step": 9220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001935233066226388,
+      "loss": 2.9731,
+      "step": 9230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019350302922221596,
+      "loss": 3.0524,
+      "step": 9240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001934827211945924,
+      "loss": 2.9097,
+      "step": 9250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019346238254642004,
+      "loss": 3.0397,
+      "step": 9260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019344201328436088,
+      "loss": 3.028,
+      "step": 9270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019342161341508694,
+      "loss": 2.8918,
+      "step": 9280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019340118294528024,
+      "loss": 2.9814,
+      "step": 9290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001933807218816328,
+      "loss": 2.9865,
+      "step": 9300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019336023023084682,
+      "loss": 3.0644,
+      "step": 9310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001933397079996343,
+      "loss": 3.0276,
+      "step": 9320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001933191551947174,
+      "loss": 2.9845,
+      "step": 9330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019329857182282824,
+      "loss": 3.0101,
+      "step": 9340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000193277957890709,
+      "loss": 3.0122,
+      "step": 9350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019325731340511177,
+      "loss": 2.921,
+      "step": 9360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019323663837279877,
+      "loss": 2.9458,
+      "step": 9370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019321593280054216,
+      "loss": 2.9696,
+      "step": 9380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019319519669512412,
+      "loss": 2.9398,
+      "step": 9390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019317443006333678,
+      "loss": 3.0334,
+      "step": 9400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019315363291198234,
+      "loss": 2.9426,
+      "step": 9410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000193132805247873,
+      "loss": 3.0177,
+      "step": 9420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019311194707783086,
+      "loss": 2.9086,
+      "step": 9430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019309105840868815,
+      "loss": 2.9704,
+      "step": 9440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019307013924728698,
+      "loss": 3.0218,
+      "step": 9450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019304918960047946,
+      "loss": 2.9821,
+      "step": 9460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019302820947512776,
+      "loss": 2.8981,
+      "step": 9470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019300719887810396,
+      "loss": 2.9528,
+      "step": 9480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001929861578162901,
+      "loss": 2.9179,
+      "step": 9490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019296508629657836,
+      "loss": 2.9857,
+      "step": 9500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019294398432587065,
+      "loss": 2.9373,
+      "step": 9510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019292285191107905,
+      "loss": 2.9522,
+      "step": 9520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019290168905912558,
+      "loss": 2.9873,
+      "step": 9530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001928804957769421,
+      "loss": 2.9096,
+      "step": 9540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019285927207147064,
+      "loss": 2.8875,
+      "step": 9550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019283801794966302,
+      "loss": 3.0336,
+      "step": 9560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001928167334184811,
+      "loss": 3.034,
+      "step": 9570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019279541848489672,
+      "loss": 3.0551,
+      "step": 9580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019277407315589163,
+      "loss": 2.9556,
+      "step": 9590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019275269743845757,
+      "loss": 3.007,
+      "step": 9600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001927312913395962,
+      "loss": 3.0204,
+      "step": 9610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019270985486631915,
+      "loss": 2.9308,
+      "step": 9620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019268838802564804,
+      "loss": 2.9708,
+      "step": 9630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019266689082461435,
+      "loss": 3.0383,
+      "step": 9640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019264536327025963,
+      "loss": 2.9349,
+      "step": 9650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019262380536963521,
+      "loss": 2.9566,
+      "step": 9660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001926022171298025,
+      "loss": 2.9425,
+      "step": 9670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019258059855783272,
+      "loss": 2.8805,
+      "step": 9680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001925589496608072,
+      "loss": 2.9582,
+      "step": 9690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019253727044581703,
+      "loss": 3.004,
+      "step": 9700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019251556091996337,
+      "loss": 2.96,
+      "step": 9710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019249382109035715,
+      "loss": 3.0309,
+      "step": 9720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001924720509641194,
+      "loss": 2.8948,
+      "step": 9730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019245025054838094,
+      "loss": 2.9517,
+      "step": 9740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001924284198502826,
+      "loss": 2.9324,
+      "step": 9750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019240655887697505,
+      "loss": 3.0396,
+      "step": 9760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019238466763561894,
+      "loss": 2.9349,
+      "step": 9770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019236274613338483,
+      "loss": 2.9799,
+      "step": 9780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019234079437745313,
+      "loss": 2.9472,
+      "step": 9790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019231881237501423,
+      "loss": 2.9399,
+      "step": 9800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019229680013326843,
+      "loss": 2.9866,
+      "step": 9810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019227475765942585,
+      "loss": 2.9636,
+      "step": 9820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001922526849607066,
+      "loss": 2.9818,
+      "step": 9830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019223058204434067,
+      "loss": 2.9283,
+      "step": 9840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019220844891756791,
+      "loss": 2.9251,
+      "step": 9850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019218628558763813,
+      "loss": 2.9721,
+      "step": 9860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019216409206181093,
+      "loss": 2.9624,
+      "step": 9870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019214186834735594,
+      "loss": 3.0221,
+      "step": 9880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019211961445155255,
+      "loss": 2.8671,
+      "step": 9890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019209733038169015,
+      "loss": 2.8799,
+      "step": 9900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019207501614506791,
+      "loss": 2.9431,
+      "step": 9910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001920526717489949,
+      "loss": 2.9562,
+      "step": 9920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019203029720079017,
+      "loss": 3.0197,
+      "step": 9930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019200789250778253,
+      "loss": 2.9525,
+      "step": 9940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001919854576773107,
+      "loss": 2.9427,
+      "step": 9950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019196299271672325,
+      "loss": 3.0564,
+      "step": 9960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001919404976333787,
+      "loss": 2.9436,
+      "step": 9970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019191797243464537,
+      "loss": 3.0733,
+      "step": 9980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019189541712790145,
+      "loss": 2.9817,
+      "step": 9990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000191872831720535,
+      "loss": 2.9418,
+      "step": 10000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019185021621994394,
+      "loss": 2.9505,
+      "step": 10010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019182757063353602,
+      "loss": 2.9282,
+      "step": 10020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001918048949687289,
+      "loss": 2.9131,
+      "step": 10030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019178218923295007,
+      "loss": 2.8732,
+      "step": 10040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019175945343363683,
+      "loss": 3.0232,
+      "step": 10050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001917366875782364,
+      "loss": 2.9362,
+      "step": 10060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019171389167420578,
+      "loss": 2.9533,
+      "step": 10070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019169106572901185,
+      "loss": 2.9948,
+      "step": 10080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019166820975013125,
+      "loss": 2.9238,
+      "step": 10090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019164532374505068,
+      "loss": 2.9195,
+      "step": 10100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019162240772126635,
+      "loss": 2.8758,
+      "step": 10110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001915994616862846,
+      "loss": 2.983,
+      "step": 10120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019157648564762134,
+      "loss": 2.8939,
+      "step": 10130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019155347961280258,
+      "loss": 2.9098,
+      "step": 10140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019153044358936395,
+      "loss": 2.9207,
+      "step": 10150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019150737758485097,
+      "loss": 2.9154,
+      "step": 10160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000191484281606819,
+      "loss": 2.9161,
+      "step": 10170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019146115566283318,
+      "loss": 2.987,
+      "step": 10180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001914379997604685,
+      "loss": 2.9092,
+      "step": 10190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001914148139073097,
+      "loss": 2.9463,
+      "step": 10200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019139159811095143,
+      "loss": 2.9294,
+      "step": 10210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019136835237899805,
+      "loss": 2.876,
+      "step": 10220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019134507671906382,
+      "loss": 2.9768,
+      "step": 10230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019132177113877273,
+      "loss": 2.9916,
+      "step": 10240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001912984356457586,
+      "loss": 2.9477,
+      "step": 10250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019127507024766503,
+      "loss": 2.926,
+      "step": 10260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019125167495214542,
+      "loss": 2.8116,
+      "step": 10270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019122824976686297,
+      "loss": 2.9138,
+      "step": 10280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001912047946994907,
+      "loss": 3.0209,
+      "step": 10290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019118130975771135,
+      "loss": 3.0251,
+      "step": 10300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019115779494921752,
+      "loss": 2.9224,
+      "step": 10310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019113425028171154,
+      "loss": 2.9678,
+      "step": 10320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019111067576290554,
+      "loss": 3.0192,
+      "step": 10330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019108707140052137,
+      "loss": 2.9724,
+      "step": 10340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019106343720229083,
+      "loss": 2.8939,
+      "step": 10350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019103977317595528,
+      "loss": 3.0247,
+      "step": 10360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019101607932926596,
+      "loss": 2.8776,
+      "step": 10370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019099235566998388,
+      "loss": 2.9,
+      "step": 10380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019096860220587977,
+      "loss": 3.0157,
+      "step": 10390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001909448189447342,
+      "loss": 2.9109,
+      "step": 10400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019092100589433738,
+      "loss": 2.9372,
+      "step": 10410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019089716306248939,
+      "loss": 2.9234,
+      "step": 10420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019087329045700005,
+      "loss": 2.8724,
+      "step": 10430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019084938808568884,
+      "loss": 2.8899,
+      "step": 10440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019082545595638508,
+      "loss": 2.9045,
+      "step": 10450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019080149407692783,
+      "loss": 3.0254,
+      "step": 10460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019077750245516585,
+      "loss": 3.0135,
+      "step": 10470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019075348109895765,
+      "loss": 2.9534,
+      "step": 10480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019072943001617157,
+      "loss": 2.9523,
+      "step": 10490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019070534921468557,
+      "loss": 2.8766,
+      "step": 10500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019068123870238736,
+      "loss": 2.9016,
+      "step": 10510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019065709848717444,
+      "loss": 3.0137,
+      "step": 10520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019063292857695406,
+      "loss": 2.9743,
+      "step": 10530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019060872897964308,
+      "loss": 2.9845,
+      "step": 10540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019058449970316816,
+      "loss": 2.963,
+      "step": 10550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001905602407554657,
+      "loss": 2.9458,
+      "step": 10560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019053595214448178,
+      "loss": 2.9201,
+      "step": 10570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019051163387817218,
+      "loss": 2.9116,
+      "step": 10580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019048728596450247,
+      "loss": 2.9384,
+      "step": 10590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019046290841144786,
+      "loss": 2.9408,
+      "step": 10600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019043850122699328,
+      "loss": 2.9321,
+      "step": 10610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001904140644191334,
+      "loss": 2.8799,
+      "step": 10620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019038959799587258,
+      "loss": 2.9364,
+      "step": 10630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019036510196522482,
+      "loss": 2.9343,
+      "step": 10640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019034057633521393,
+      "loss": 2.9246,
+      "step": 10650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001903160211138733,
+      "loss": 2.9373,
+      "step": 10660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019029143630924607,
+      "loss": 2.9617,
+      "step": 10670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019026682192938512,
+      "loss": 2.8688,
+      "step": 10680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019024217798235292,
+      "loss": 2.9576,
+      "step": 10690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019021750447622169,
+      "loss": 2.985,
+      "step": 10700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001901928014190733,
+      "loss": 2.9384,
+      "step": 10710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019016806881899933,
+      "loss": 2.9793,
+      "step": 10720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000190143306684101,
+      "loss": 2.8323,
+      "step": 10730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019011851502248923,
+      "loss": 2.9148,
+      "step": 10740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019009369384228458,
+      "loss": 2.9439,
+      "step": 10750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019006884315161734,
+      "loss": 2.9247,
+      "step": 10760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019004396295862743,
+      "loss": 2.937,
+      "step": 10770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001900190532714644,
+      "loss": 2.8896,
+      "step": 10780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018999411409828748,
+      "loss": 2.9185,
+      "step": 10790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018996914544726562,
+      "loss": 2.8663,
+      "step": 10800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018994414732657737,
+      "loss": 2.9704,
+      "step": 10810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018991911974441093,
+      "loss": 2.9734,
+      "step": 10820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018989406270896416,
+      "loss": 2.9308,
+      "step": 10830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018986897622844454,
+      "loss": 3.0134,
+      "step": 10840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018984386031106927,
+      "loss": 2.9338,
+      "step": 10850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018981871496506512,
+      "loss": 2.9158,
+      "step": 10860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018979354019866852,
+      "loss": 2.8965,
+      "step": 10870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018976833602012559,
+      "loss": 2.9444,
+      "step": 10880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018974310243769196,
+      "loss": 2.883,
+      "step": 10890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000189717839459633,
+      "loss": 2.9124,
+      "step": 10900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018969254709422367,
+      "loss": 2.8878,
+      "step": 10910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001896672253497486,
+      "loss": 2.8779,
+      "step": 10920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018964187423450195,
+      "loss": 2.8612,
+      "step": 10930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018961649375678755,
+      "loss": 2.8846,
+      "step": 10940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018959108392491893,
+      "loss": 2.9795,
+      "step": 10950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018956564474721907,
+      "loss": 2.897,
+      "step": 10960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001895401762320207,
+      "loss": 2.8896,
+      "step": 10970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001895146783876661,
+      "loss": 2.9596,
+      "step": 10980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018948915122250716,
+      "loss": 2.9487,
+      "step": 10990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018946359474490535,
+      "loss": 2.7977,
+      "step": 11000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018943800896323185,
+      "loss": 2.9845,
+      "step": 11010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001894123938858673,
+      "loss": 2.9513,
+      "step": 11020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000189386749521202,
+      "loss": 2.9211,
+      "step": 11030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018936107587763586,
+      "loss": 3.0171,
+      "step": 11040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018933537296357834,
+      "loss": 2.9208,
+      "step": 11050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018930964078744852,
+      "loss": 2.8953,
+      "step": 11060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018928387935767505,
+      "loss": 2.9189,
+      "step": 11070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018925808868269618,
+      "loss": 2.9682,
+      "step": 11080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018923226877095968,
+      "loss": 2.8637,
+      "step": 11090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018920641963092298,
+      "loss": 2.971,
+      "step": 11100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018918054127105304,
+      "loss": 2.8539,
+      "step": 11110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018915463369982636,
+      "loss": 2.8878,
+      "step": 11120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001891286969257291,
+      "loss": 2.8327,
+      "step": 11130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018910273095725687,
+      "loss": 2.9209,
+      "step": 11140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018907673580291498,
+      "loss": 2.9701,
+      "step": 11150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018905071147121814,
+      "loss": 2.8829,
+      "step": 11160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018902465797069072,
+      "loss": 2.8963,
+      "step": 11170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018899857530986665,
+      "loss": 2.8913,
+      "step": 11180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018897246349728936,
+      "loss": 2.9454,
+      "step": 11190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018894893794828968,
+      "loss": 2.9662,
+      "step": 11200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018892277077095274,
+      "loss": 2.779,
+      "step": 11210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018889657446669256,
+      "loss": 2.9785,
+      "step": 11220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018887034904408985,
+      "loss": 2.8737,
+      "step": 11230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018884409451173486,
+      "loss": 2.9645,
+      "step": 11240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001888178108782273,
+      "loss": 3.018,
+      "step": 11250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018879149815217647,
+      "loss": 2.9226,
+      "step": 11260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018876515634220117,
+      "loss": 2.9008,
+      "step": 11270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018873878545692976,
+      "loss": 2.876,
+      "step": 11280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001887123855050001,
+      "loss": 2.9043,
+      "step": 11290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018868595649505955,
+      "loss": 3.0338,
+      "step": 11300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018865949843576506,
+      "loss": 2.9622,
+      "step": 11310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000188633011335783,
+      "loss": 2.9477,
+      "step": 11320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018860649520378931,
+      "loss": 2.8959,
+      "step": 11330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018857995004846944,
+      "loss": 2.8437,
+      "step": 11340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018855337587851833,
+      "loss": 2.9983,
+      "step": 11350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018852677270264048,
+      "loss": 3.0235,
+      "step": 11360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018850014052954972,
+      "loss": 2.8903,
+      "step": 11370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018847347936796962,
+      "loss": 2.958,
+      "step": 11380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018844678922663307,
+      "loss": 2.9808,
+      "step": 11390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001884200701142825,
+      "loss": 3.0096,
+      "step": 11400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018839332203966986,
+      "loss": 2.8775,
+      "step": 11410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018836654501155656,
+      "loss": 2.9041,
+      "step": 11420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001883397390387135,
+      "loss": 2.9459,
+      "step": 11430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018831290412992102,
+      "loss": 2.9058,
+      "step": 11440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000188286040293969,
+      "loss": 2.948,
+      "step": 11450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018825914753965682,
+      "loss": 2.8676,
+      "step": 11460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018823222587579318,
+      "loss": 2.9027,
+      "step": 11470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018820527531119647,
+      "loss": 2.9032,
+      "step": 11480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018817829585469434,
+      "loss": 2.9231,
+      "step": 11490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018815128751512402,
+      "loss": 2.9603,
+      "step": 11500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001881242503013322,
+      "loss": 2.9196,
+      "step": 11510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018809718422217492,
+      "loss": 2.8995,
+      "step": 11520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018807008928651786,
+      "loss": 2.8717,
+      "step": 11530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018804296550323598,
+      "loss": 2.8334,
+      "step": 11540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018801581288121382,
+      "loss": 2.8974,
+      "step": 11550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018798863142934521,
+      "loss": 2.8754,
+      "step": 11560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001879614211565336,
+      "loss": 2.8607,
+      "step": 11570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018793418207169177,
+      "loss": 2.9684,
+      "step": 11580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018790691418374194,
+      "loss": 2.8542,
+      "step": 11590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018787961750161585,
+      "loss": 2.8301,
+      "step": 11600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018785229203425454,
+      "loss": 2.8551,
+      "step": 11610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018782493779060858,
+      "loss": 2.8584,
+      "step": 11620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018779755477963796,
+      "loss": 2.8726,
+      "step": 11630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018777014301031203,
+      "loss": 2.9886,
+      "step": 11640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018774270249160963,
+      "loss": 2.989,
+      "step": 11650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018771523323251894,
+      "loss": 2.9735,
+      "step": 11660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018768773524203766,
+      "loss": 2.8517,
+      "step": 11670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001876602085291728,
+      "loss": 2.8776,
+      "step": 11680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018763265310294084,
+      "loss": 2.8806,
+      "step": 11690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001876050689723676,
+      "loss": 2.913,
+      "step": 11700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001875774561464884,
+      "loss": 2.9279,
+      "step": 11710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018754981463434787,
+      "loss": 2.8531,
+      "step": 11720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018752214444500007,
+      "loss": 2.8862,
+      "step": 11730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018749444558750845,
+      "loss": 2.9839,
+      "step": 11740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018746671807094592,
+      "loss": 2.851,
+      "step": 11750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018743896190439463,
+      "loss": 2.9503,
+      "step": 11760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018741117709694628,
+      "loss": 2.8775,
+      "step": 11770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018738336365770175,
+      "loss": 2.9188,
+      "step": 11780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001873555215957715,
+      "loss": 2.9352,
+      "step": 11790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001873276509202753,
+      "loss": 2.8976,
+      "step": 11800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001872997516403422,
+      "loss": 2.8732,
+      "step": 11810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018727182376511076,
+      "loss": 2.9441,
+      "step": 11820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018724386730372883,
+      "loss": 2.9046,
+      "step": 11830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001872158822653536,
+      "loss": 2.8626,
+      "step": 11840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018718786865915164,
+      "loss": 2.9396,
+      "step": 11850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018715982649429895,
+      "loss": 2.9242,
+      "step": 11860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001871317557799808,
+      "loss": 2.8892,
+      "step": 11870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018710365652539185,
+      "loss": 2.8815,
+      "step": 11880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018707552873973606,
+      "loss": 2.9481,
+      "step": 11890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018704737243222682,
+      "loss": 2.9255,
+      "step": 11900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018701918761208677,
+      "loss": 2.9414,
+      "step": 11910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018699097428854798,
+      "loss": 2.9348,
+      "step": 11920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018696273247085173,
+      "loss": 2.954,
+      "step": 11930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001869344621682488,
+      "loss": 2.896,
+      "step": 11940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001869061633899992,
+      "loss": 2.9492,
+      "step": 11950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001868778361453722,
+      "loss": 2.8333,
+      "step": 11960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001868494804436466,
+      "loss": 2.843,
+      "step": 11970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001868210962941103,
+      "loss": 2.9017,
+      "step": 11980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001867926837060606,
+      "loss": 2.9657,
+      "step": 11990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018676424268880424,
+      "loss": 2.9743,
+      "step": 12000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018673577325165706,
+      "loss": 2.8867,
+      "step": 12010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018670727540394432,
+      "loss": 2.9656,
+      "step": 12020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018667874915500067,
+      "loss": 2.966,
+      "step": 12030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018665019451416986,
+      "loss": 2.8191,
+      "step": 12040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018662161149080508,
+      "loss": 2.9215,
+      "step": 12050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018659300009426884,
+      "loss": 2.7995,
+      "step": 12060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001865643603339328,
+      "loss": 2.9132,
+      "step": 12070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018653569221917807,
+      "loss": 2.896,
+      "step": 12080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018650699575939494,
+      "loss": 2.8174,
+      "step": 12090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018647827096398306,
+      "loss": 2.8002,
+      "step": 12100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018644951784235126,
+      "loss": 2.9853,
+      "step": 12110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018642073640391776,
+      "loss": 2.8966,
+      "step": 12120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018639192665811005,
+      "loss": 2.9778,
+      "step": 12130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018636308861436475,
+      "loss": 2.8704,
+      "step": 12140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018633422228212792,
+      "loss": 2.9054,
+      "step": 12150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018630532767085478,
+      "loss": 2.7541,
+      "step": 12160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018627640479000986,
+      "loss": 2.9026,
+      "step": 12170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018624745364906694,
+      "loss": 2.8469,
+      "step": 12180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001862184742575091,
+      "loss": 2.8999,
+      "step": 12190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001861894666248285,
+      "loss": 2.8714,
+      "step": 12200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001861604307605268,
+      "loss": 2.9258,
+      "step": 12210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018613136667411473,
+      "loss": 2.9277,
+      "step": 12220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018610227437511237,
+      "loss": 2.7877,
+      "step": 12230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001860731538730489,
+      "loss": 2.9708,
+      "step": 12240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018604400517746293,
+      "loss": 2.9528,
+      "step": 12250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018601482829790208,
+      "loss": 2.8791,
+      "step": 12260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018598562324392349,
+      "loss": 2.8895,
+      "step": 12270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018595639002509323,
+      "loss": 2.8381,
+      "step": 12280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001859271286509868,
+      "loss": 2.9248,
+      "step": 12290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001858978391311888,
+      "loss": 2.8708,
+      "step": 12300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001858685214752931,
+      "loss": 2.9205,
+      "step": 12310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018583917569290286,
+      "loss": 2.9129,
+      "step": 12320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018580980179363032,
+      "loss": 2.8233,
+      "step": 12330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000185780399787097,
+      "loss": 2.8203,
+      "step": 12340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018575096968293363,
+      "loss": 2.9456,
+      "step": 12350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018572151149078012,
+      "loss": 2.9582,
+      "step": 12360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001856920252202856,
+      "loss": 2.8487,
+      "step": 12370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018566251088110834,
+      "loss": 2.84,
+      "step": 12380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018563296848291594,
+      "loss": 2.8533,
+      "step": 12390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000185603398035385,
+      "loss": 2.835,
+      "step": 12400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001855737995482015,
+      "loss": 2.8809,
+      "step": 12410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018554417303106045,
+      "loss": 2.9712,
+      "step": 12420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018551451849366615,
+      "loss": 2.9968,
+      "step": 12430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018548483594573202,
+      "loss": 2.9386,
+      "step": 12440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018545512539698065,
+      "loss": 2.9477,
+      "step": 12450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018542538685714383,
+      "loss": 2.8778,
+      "step": 12460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001853956203359625,
+      "loss": 2.8282,
+      "step": 12470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018536582584318683,
+      "loss": 2.8913,
+      "step": 12480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000185336003388576,
+      "loss": 2.8826,
+      "step": 12490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018530615298189852,
+      "loss": 2.8559,
+      "step": 12500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000185276274632932,
+      "loss": 2.8964,
+      "step": 12510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001852463683514631,
+      "loss": 2.8693,
+      "step": 12520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018521643414728774,
+      "loss": 2.8168,
+      "step": 12530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000185186472030211,
+      "loss": 2.8633,
+      "step": 12540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018515648201004704,
+      "loss": 2.8486,
+      "step": 12550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018512646409661918,
+      "loss": 2.9089,
+      "step": 12560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018509641829975987,
+      "loss": 2.8959,
+      "step": 12570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018506634462931072,
+      "loss": 2.9367,
+      "step": 12580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018503624309512245,
+      "loss": 2.9012,
+      "step": 12590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001850061137070549,
+      "loss": 2.9078,
+      "step": 12600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018497595647497703,
+      "loss": 2.9025,
+      "step": 12610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018494577140876691,
+      "loss": 2.8727,
+      "step": 12620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018491555851831185,
+      "loss": 2.8372,
+      "step": 12630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001848853178135081,
+      "loss": 2.934,
+      "step": 12640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018485504930426108,
+      "loss": 2.8812,
+      "step": 12650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018482475300048536,
+      "loss": 2.8345,
+      "step": 12660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018479442891210462,
+      "loss": 2.8295,
+      "step": 12670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018476407704905151,
+      "loss": 2.896,
+      "step": 12680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018473369742126798,
+      "loss": 2.8275,
+      "step": 12690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018470329003870492,
+      "loss": 2.834,
+      "step": 12700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018467285491132235,
+      "loss": 2.8596,
+      "step": 12710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018464239204908945,
+      "loss": 2.8922,
+      "step": 12720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001846119014619843,
+      "loss": 2.8825,
+      "step": 12730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001845813831599943,
+      "loss": 2.8464,
+      "step": 12740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018455083715311578,
+      "loss": 2.8591,
+      "step": 12750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018452026345135416,
+      "loss": 2.9373,
+      "step": 12760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018448966206472396,
+      "loss": 2.7819,
+      "step": 12770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018445903300324875,
+      "loss": 2.8864,
+      "step": 12780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018442837627696119,
+      "loss": 2.9488,
+      "step": 12790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018439769189590295,
+      "loss": 2.8584,
+      "step": 12800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001843669798701248,
+      "loss": 2.8946,
+      "step": 12810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018433624020968662,
+      "loss": 2.7667,
+      "step": 12820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001843054729246572,
+      "loss": 2.8346,
+      "step": 12830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018427467802511452,
+      "loss": 2.8882,
+      "step": 12840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018424385552114544,
+      "loss": 2.8984,
+      "step": 12850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018421300542284608,
+      "loss": 2.9422,
+      "step": 12860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018418212774032143,
+      "loss": 2.8852,
+      "step": 12870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018415122248368557,
+      "loss": 2.9235,
+      "step": 12880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018412028966306162,
+      "loss": 2.8973,
+      "step": 12890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018408932928858175,
+      "loss": 2.8216,
+      "step": 12900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018405834137038706,
+      "loss": 2.7905,
+      "step": 12910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018402732591862774,
+      "loss": 2.9434,
+      "step": 12920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018399628294346308,
+      "loss": 2.8369,
+      "step": 12930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018396521245506118,
+      "loss": 2.8772,
+      "step": 12940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018393411446359937,
+      "loss": 2.8226,
+      "step": 12950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018390298897926383,
+      "loss": 2.9417,
+      "step": 12960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018387183601224985,
+      "loss": 2.8922,
+      "step": 12970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001838406555727617,
+      "loss": 2.889,
+      "step": 12980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018380944767101257,
+      "loss": 2.834,
+      "step": 12990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001837782123172247,
+      "loss": 2.9279,
+      "step": 13000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018374694952162933,
+      "loss": 2.8931,
+      "step": 13010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018371565929446674,
+      "loss": 2.8168,
+      "step": 13020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001836843416459861,
+      "loss": 2.8405,
+      "step": 13030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018365299658644558,
+      "loss": 2.9036,
+      "step": 13040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018362162412611236,
+      "loss": 2.853,
+      "step": 13050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018359022427526264,
+      "loss": 2.8501,
+      "step": 13060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018355879704418145,
+      "loss": 2.8945,
+      "step": 13070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018352734244316294,
+      "loss": 2.8679,
+      "step": 13080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001834958604825101,
+      "loss": 2.8951,
+      "step": 13090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000183464351172535,
+      "loss": 2.8413,
+      "step": 13100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001834328145235586,
+      "loss": 2.89,
+      "step": 13110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001834012505459108,
+      "loss": 2.898,
+      "step": 13120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018336965924993048,
+      "loss": 2.9128,
+      "step": 13130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001833380406459655,
+      "loss": 2.8674,
+      "step": 13140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018330639474437258,
+      "loss": 2.9914,
+      "step": 13150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018327472155551746,
+      "loss": 2.8589,
+      "step": 13160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001832430210897748,
+      "loss": 2.8687,
+      "step": 13170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018321129335752812,
+      "loss": 2.8548,
+      "step": 13180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018317953836917,
+      "loss": 2.8668,
+      "step": 13190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001831477561351019,
+      "loss": 2.8973,
+      "step": 13200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001831159466657341,
+      "loss": 2.9491,
+      "step": 13210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018308410997148596,
+      "loss": 2.877,
+      "step": 13220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001830522460627857,
+      "loss": 2.9205,
+      "step": 13230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018302035495007036,
+      "loss": 2.8641,
+      "step": 13240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018298843664378605,
+      "loss": 2.8727,
+      "step": 13250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018295649115438766,
+      "loss": 2.8695,
+      "step": 13260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018292451849233901,
+      "loss": 2.928,
+      "step": 13270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018289251866811292,
+      "loss": 2.8496,
+      "step": 13280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000182860491692191,
+      "loss": 2.7952,
+      "step": 13290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018282843757506373,
+      "loss": 2.8437,
+      "step": 13300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001827963563272306,
+      "loss": 2.8567,
+      "step": 13310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018276424795919989,
+      "loss": 2.8917,
+      "step": 13320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001827321124814888,
+      "loss": 2.903,
+      "step": 13330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001826999499046234,
+      "loss": 3.0026,
+      "step": 13340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018266776023913864,
+      "loss": 2.8257,
+      "step": 13350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018263554349557832,
+      "loss": 2.7549,
+      "step": 13360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001826032996844952,
+      "loss": 2.8128,
+      "step": 13370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018257102881645074,
+      "loss": 2.816,
+      "step": 13380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018253873090201544,
+      "loss": 2.8572,
+      "step": 13390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018250640595176857,
+      "loss": 2.8478,
+      "step": 13400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018247405397629822,
+      "loss": 2.8681,
+      "step": 13410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001824416749862014,
+      "loss": 2.8826,
+      "step": 13420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000182409268992084,
+      "loss": 2.8709,
+      "step": 13430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018237683600456063,
+      "loss": 2.8938,
+      "step": 13440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018234437603425481,
+      "loss": 2.842,
+      "step": 13450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018231188909179896,
+      "loss": 2.9086,
+      "step": 13460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018227937518783426,
+      "loss": 2.9323,
+      "step": 13470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001822468343330107,
+      "loss": 2.867,
+      "step": 13480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018221426653798717,
+      "loss": 2.9124,
+      "step": 13490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018218167181343137,
+      "loss": 2.8527,
+      "step": 13500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001821490501700197,
+      "loss": 2.8697,
+      "step": 13510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018211640161843758,
+      "loss": 2.8534,
+      "step": 13520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018208372616937912,
+      "loss": 2.9427,
+      "step": 13530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018205102383354725,
+      "loss": 2.8513,
+      "step": 13540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018201829462165372,
+      "loss": 2.9051,
+      "step": 13550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001819855385444191,
+      "loss": 2.8396,
+      "step": 13560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001819527556125727,
+      "loss": 2.8508,
+      "step": 13570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001819199458368527,
+      "loss": 2.8423,
+      "step": 13580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018188710922800608,
+      "loss": 2.8918,
+      "step": 13590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018185424579678848,
+      "loss": 2.8684,
+      "step": 13600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018182135555396448,
+      "loss": 2.8604,
+      "step": 13610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018178843851030739,
+      "loss": 2.8701,
+      "step": 13620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018175549467659921,
+      "loss": 2.7862,
+      "step": 13630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001817225240636309,
+      "loss": 2.922,
+      "step": 13640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018168952668220196,
+      "loss": 2.8412,
+      "step": 13650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001816565025431209,
+      "loss": 2.7602,
+      "step": 13660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001816234516572048,
+      "loss": 2.8685,
+      "step": 13670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018159037403527967,
+      "loss": 2.9078,
+      "step": 13680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018155726968818004,
+      "loss": 2.9449,
+      "step": 13690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018152413862674948,
+      "loss": 2.8123,
+      "step": 13700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018149098086184007,
+      "loss": 2.8536,
+      "step": 13710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018145779640431277,
+      "loss": 2.7804,
+      "step": 13720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018142458526503727,
+      "loss": 2.9019,
+      "step": 13730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018139134745489196,
+      "loss": 2.7999,
+      "step": 13740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018135808298476403,
+      "loss": 2.8085,
+      "step": 13750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018132479186554927,
+      "loss": 2.9006,
+      "step": 13760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018129147410815238,
+      "loss": 2.9661,
+      "step": 13770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018125812972348662,
+      "loss": 2.8471,
+      "step": 13780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018122475872247407,
+      "loss": 2.8751,
+      "step": 13790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018119136111604552,
+      "loss": 2.8315,
+      "step": 13800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018115793691514043,
+      "loss": 2.8093,
+      "step": 13810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018112448613070701,
+      "loss": 2.872,
+      "step": 13820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018109100877370217,
+      "loss": 2.924,
+      "step": 13830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018105750485509153,
+      "loss": 2.9473,
+      "step": 13840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018102397438584937,
+      "loss": 2.8886,
+      "step": 13850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018099041737695872,
+      "loss": 2.8165,
+      "step": 13860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018095683383941128,
+      "loss": 2.7789,
+      "step": 13870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001809232237842074,
+      "loss": 2.763,
+      "step": 13880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001808895872223562,
+      "loss": 2.8323,
+      "step": 13890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018085592416487546,
+      "loss": 2.9105,
+      "step": 13900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018082223462279154,
+      "loss": 2.8475,
+      "step": 13910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018078851860713963,
+      "loss": 2.8808,
+      "step": 13920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018075477612896346,
+      "loss": 2.8329,
+      "step": 13930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001807210071993155,
+      "loss": 2.9317,
+      "step": 13940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001806872118292569,
+      "loss": 2.858,
+      "step": 13950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018065339002985736,
+      "loss": 2.9045,
+      "step": 13960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018061954181219534,
+      "loss": 2.8583,
+      "step": 13970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018058566718735797,
+      "loss": 2.8354,
+      "step": 13980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018055176616644093,
+      "loss": 2.8926,
+      "step": 13990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018051783876054867,
+      "loss": 2.8634,
+      "step": 14000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018048388498079413,
+      "loss": 2.8381,
+      "step": 14010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018044990483829905,
+      "loss": 2.8273,
+      "step": 14020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018041589834419364,
+      "loss": 2.807,
+      "step": 14030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018038186550961693,
+      "loss": 2.8978,
+      "step": 14040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001803478063457164,
+      "loss": 2.8621,
+      "step": 14050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018031372086364832,
+      "loss": 2.8211,
+      "step": 14060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018027960907457738,
+      "loss": 2.7936,
+      "step": 14070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001802454709896771,
+      "loss": 2.8054,
+      "step": 14080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018021130662012946,
+      "loss": 2.8877,
+      "step": 14090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001801771159771251,
+      "loss": 2.8549,
+      "step": 14100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018014289907186332,
+      "loss": 2.7887,
+      "step": 14110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001801086559155519,
+      "loss": 2.8175,
+      "step": 14120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001800743865194073,
+      "loss": 2.9059,
+      "step": 14130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018004009089465464,
+      "loss": 2.7316,
+      "step": 14140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00018000576905252748,
+      "loss": 2.7509,
+      "step": 14150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017997142100426808,
+      "loss": 2.8573,
+      "step": 14160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017993704676112722,
+      "loss": 2.845,
+      "step": 14170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001799026463343643,
+      "loss": 2.8722,
+      "step": 14180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001798682197352473,
+      "loss": 2.7838,
+      "step": 14190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017983376697505276,
+      "loss": 2.8039,
+      "step": 14200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017979928806506577,
+      "loss": 2.8431,
+      "step": 14210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017976478301657994,
+      "loss": 2.8357,
+      "step": 14220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017973025184089763,
+      "loss": 2.7971,
+      "step": 14230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017969569454932954,
+      "loss": 2.8599,
+      "step": 14240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017966111115319503,
+      "loss": 2.8887,
+      "step": 14250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017962650166382203,
+      "loss": 2.8399,
+      "step": 14260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017959186609254693,
+      "loss": 2.7691,
+      "step": 14270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017955720445071477,
+      "loss": 2.9113,
+      "step": 14280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017952251674967906,
+      "loss": 2.7674,
+      "step": 14290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017948780300080185,
+      "loss": 2.8812,
+      "step": 14300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017945306321545378,
+      "loss": 2.8575,
+      "step": 14310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017941829740501385,
+      "loss": 2.7275,
+      "step": 14320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001793835055808698,
+      "loss": 2.9002,
+      "step": 14330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017934868775441783,
+      "loss": 2.8513,
+      "step": 14340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017931384393706258,
+      "loss": 2.7923,
+      "step": 14350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017927897414021722,
+      "loss": 2.8384,
+      "step": 14360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017924407837530351,
+      "loss": 2.8243,
+      "step": 14370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017920915665375163,
+      "loss": 2.8286,
+      "step": 14380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017917420898700027,
+      "loss": 2.8883,
+      "step": 14390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001791392353864967,
+      "loss": 2.9265,
+      "step": 14400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017910423586369663,
+      "loss": 2.7002,
+      "step": 14410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001790692104300642,
+      "loss": 2.8627,
+      "step": 14420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017903415909707214,
+      "loss": 2.914,
+      "step": 14430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017899908187620164,
+      "loss": 2.8426,
+      "step": 14440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017896397877894234,
+      "loss": 2.7995,
+      "step": 14450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017892884981679229,
+      "loss": 2.932,
+      "step": 14460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017889369500125821,
+      "loss": 2.8078,
+      "step": 14470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017885851434385508,
+      "loss": 2.8462,
+      "step": 14480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017882330785610643,
+      "loss": 2.7802,
+      "step": 14490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001787880755495443,
+      "loss": 2.8518,
+      "step": 14500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017875281743570914,
+      "loss": 2.8423,
+      "step": 14510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001787175335261498,
+      "loss": 2.808,
+      "step": 14520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017868222383242367,
+      "loss": 2.8658,
+      "step": 14530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017864688836609656,
+      "loss": 2.8834,
+      "step": 14540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017861152713874266,
+      "loss": 2.8919,
+      "step": 14550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001785761401619447,
+      "loss": 2.8937,
+      "step": 14560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017854072744729374,
+      "loss": 2.7923,
+      "step": 14570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017850528900638938,
+      "loss": 2.8302,
+      "step": 14580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001784698248508395,
+      "loss": 2.8647,
+      "step": 14590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001784343349922606,
+      "loss": 2.8692,
+      "step": 14600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017839881944227737,
+      "loss": 2.7796,
+      "step": 14610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017836327821252318,
+      "loss": 2.7829,
+      "step": 14620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017832771131463958,
+      "loss": 2.796,
+      "step": 14630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001782921187602766,
+      "loss": 2.8144,
+      "step": 14640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017825650056109272,
+      "loss": 2.79,
+      "step": 14650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017822085672875484,
+      "loss": 2.7903,
+      "step": 14660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001781851872749381,
+      "loss": 2.7257,
+      "step": 14670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001781494922113262,
+      "loss": 2.8658,
+      "step": 14680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017811377154961115,
+      "loss": 2.7379,
+      "step": 14690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017807802530149337,
+      "loss": 2.8914,
+      "step": 14700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017804225347868168,
+      "loss": 2.8099,
+      "step": 14710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017800645609289317,
+      "loss": 2.8294,
+      "step": 14720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017797063315585346,
+      "loss": 2.9011,
+      "step": 14730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017793478467929643,
+      "loss": 2.8232,
+      "step": 14740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017789891067496435,
+      "loss": 2.8373,
+      "step": 14750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017786301115460786,
+      "loss": 2.8774,
+      "step": 14760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000177827086129986,
+      "loss": 2.8459,
+      "step": 14770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017779113561286605,
+      "loss": 2.8621,
+      "step": 14780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017775515961502376,
+      "loss": 2.8535,
+      "step": 14790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017771915814824313,
+      "loss": 2.868,
+      "step": 14800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017768313122431664,
+      "loss": 2.8009,
+      "step": 14810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017764707885504493,
+      "loss": 2.7355,
+      "step": 14820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001776110010522371,
+      "loss": 2.8461,
+      "step": 14830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017757489782771054,
+      "loss": 2.8092,
+      "step": 14840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017753876919329096,
+      "loss": 2.8095,
+      "step": 14850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001775026151608124,
+      "loss": 2.861,
+      "step": 14860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017746643574211729,
+      "loss": 2.8374,
+      "step": 14870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017743023094905617,
+      "loss": 2.7997,
+      "step": 14880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017739400079348818,
+      "loss": 2.9162,
+      "step": 14890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017735774528728054,
+      "loss": 2.888,
+      "step": 14900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017732146444230886,
+      "loss": 2.7305,
+      "step": 14910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017728515827045702,
+      "loss": 2.8512,
+      "step": 14920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017724882678361726,
+      "loss": 2.8966,
+      "step": 14930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017721246999369004,
+      "loss": 2.8744,
+      "step": 14940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017717608791258417,
+      "loss": 2.8306,
+      "step": 14950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017713968055221664,
+      "loss": 2.7961,
+      "step": 14960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017710324792451286,
+      "loss": 2.8248,
+      "step": 14970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017706679004140643,
+      "loss": 2.8738,
+      "step": 14980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017703030691483922,
+      "loss": 2.8515,
+      "step": 14990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001769937985567614,
+      "loss": 2.8115,
+      "step": 15000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001769572649791314,
+      "loss": 2.845,
+      "step": 15010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017692070619391593,
+      "loss": 2.8685,
+      "step": 15020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017688412221308986,
+      "loss": 2.7558,
+      "step": 15030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017684751304863645,
+      "loss": 2.8388,
+      "step": 15040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017681087871254713,
+      "loss": 2.897,
+      "step": 15050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017677421921682157,
+      "loss": 2.8361,
+      "step": 15060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017673753457346768,
+      "loss": 2.8457,
+      "step": 15070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017670082479450167,
+      "loss": 2.8193,
+      "step": 15080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001766640898919479,
+      "loss": 2.8846,
+      "step": 15090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017662732987783905,
+      "loss": 2.8781,
+      "step": 15100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017659054476421596,
+      "loss": 2.869,
+      "step": 15110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017655373456312767,
+      "loss": 2.7992,
+      "step": 15120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017651689928663148,
+      "loss": 2.8153,
+      "step": 15130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001764800389467929,
+      "loss": 2.8294,
+      "step": 15140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017644315355568567,
+      "loss": 2.8415,
+      "step": 15150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017640624312539172,
+      "loss": 2.8016,
+      "step": 15160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001763693076680011,
+      "loss": 2.8653,
+      "step": 15170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001763323471956122,
+      "loss": 2.8849,
+      "step": 15180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001762953617203315,
+      "loss": 2.8165,
+      "step": 15190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001762583512542738,
+      "loss": 2.8689,
+      "step": 15200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017622131580956183,
+      "loss": 2.8253,
+      "step": 15210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017618425539832677,
+      "loss": 2.8532,
+      "step": 15220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017614717003270784,
+      "loss": 2.7452,
+      "step": 15230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017611005972485247,
+      "loss": 2.8525,
+      "step": 15240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017607292448691624,
+      "loss": 2.7575,
+      "step": 15250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017603576433106294,
+      "loss": 2.8183,
+      "step": 15260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017599857926946445,
+      "loss": 2.9309,
+      "step": 15270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001759613693143009,
+      "loss": 2.6982,
+      "step": 15280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017592413447776046,
+      "loss": 2.9148,
+      "step": 15290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017588687477203957,
+      "loss": 2.8193,
+      "step": 15300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017584959020934275,
+      "loss": 2.809,
+      "step": 15310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017581228080188263,
+      "loss": 2.7372,
+      "step": 15320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017577494656188008,
+      "loss": 2.863,
+      "step": 15330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017573758750156403,
+      "loss": 2.8275,
+      "step": 15340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001757002036331715,
+      "loss": 2.8025,
+      "step": 15350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017566279496894773,
+      "loss": 2.7685,
+      "step": 15360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017562536152114603,
+      "loss": 2.8752,
+      "step": 15370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017558790330202787,
+      "loss": 2.8779,
+      "step": 15380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017555042032386275,
+      "loss": 2.8491,
+      "step": 15390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001755129125989284,
+      "loss": 2.8259,
+      "step": 15400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017547538013951055,
+      "loss": 2.9001,
+      "step": 15410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017543782295790304,
+      "loss": 2.8691,
+      "step": 15420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017540024106640792,
+      "loss": 2.8202,
+      "step": 15430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017536263447733518,
+      "loss": 2.8396,
+      "step": 15440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000175325003203003,
+      "loss": 2.7954,
+      "step": 15450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001752873472557376,
+      "loss": 2.8284,
+      "step": 15460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001752496666478733,
+      "loss": 2.7724,
+      "step": 15470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017521196139175257,
+      "loss": 2.8546,
+      "step": 15480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017517423149972577,
+      "loss": 2.7889,
+      "step": 15490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017513647698415153,
+      "loss": 2.8595,
+      "step": 15500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017509869785739638,
+      "loss": 2.8139,
+      "step": 15510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017506089413183505,
+      "loss": 2.7785,
+      "step": 15520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017502306581985026,
+      "loss": 2.867,
+      "step": 15530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017498521293383276,
+      "loss": 2.8092,
+      "step": 15540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017494733548618137,
+      "loss": 2.7434,
+      "step": 15550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017490943348930302,
+      "loss": 2.7996,
+      "step": 15560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017487150695561255,
+      "loss": 2.8005,
+      "step": 15570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017483355589753297,
+      "loss": 2.8484,
+      "step": 15580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001747955803274952,
+      "loss": 2.79,
+      "step": 15590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017475758025793834,
+      "loss": 2.7952,
+      "step": 15600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017471955570130937,
+      "loss": 2.8099,
+      "step": 15610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017468150667006334,
+      "loss": 2.7668,
+      "step": 15620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017464343317666336,
+      "loss": 2.7918,
+      "step": 15630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001746053352335805,
+      "loss": 2.8613,
+      "step": 15640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017456721285329384,
+      "loss": 2.7719,
+      "step": 15650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017452906604829054,
+      "loss": 2.8371,
+      "step": 15660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017449089483106565,
+      "loss": 2.7978,
+      "step": 15670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001744526992141223,
+      "loss": 2.7318,
+      "step": 15680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017441447920997152,
+      "loss": 2.8229,
+      "step": 15690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017437623483113244,
+      "loss": 2.7449,
+      "step": 15700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017433796609013212,
+      "loss": 2.8586,
+      "step": 15710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017429967299950562,
+      "loss": 2.8102,
+      "step": 15720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001742613555717959,
+      "loss": 2.7687,
+      "step": 15730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017422301381955402,
+      "loss": 2.6833,
+      "step": 15740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017418464775533887,
+      "loss": 2.8388,
+      "step": 15750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017414625739171736,
+      "loss": 2.7716,
+      "step": 15760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017410784274126447,
+      "loss": 2.8253,
+      "step": 15770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017406940381656296,
+      "loss": 2.6807,
+      "step": 15780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017403094063020365,
+      "loss": 2.7787,
+      "step": 15790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017399245319478523,
+      "loss": 2.7793,
+      "step": 15800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001739539415229144,
+      "loss": 2.7521,
+      "step": 15810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017391540562720578,
+      "loss": 2.8091,
+      "step": 15820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017387684552028188,
+      "loss": 2.7906,
+      "step": 15830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017383826121477328,
+      "loss": 2.7664,
+      "step": 15840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017379965272331827,
+      "loss": 2.8499,
+      "step": 15850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017376102005856327,
+      "loss": 2.8617,
+      "step": 15860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017372236323316248,
+      "loss": 2.7716,
+      "step": 15870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017368368225977805,
+      "loss": 2.801,
+      "step": 15880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017364497715108008,
+      "loss": 2.7974,
+      "step": 15890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017360624791974653,
+      "loss": 2.7365,
+      "step": 15900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001735674945784633,
+      "loss": 2.8028,
+      "step": 15910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001735287171399241,
+      "loss": 2.8272,
+      "step": 15920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017348991561683065,
+      "loss": 2.7809,
+      "step": 15930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001734510900218925,
+      "loss": 2.8632,
+      "step": 15940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001734122403678271,
+      "loss": 2.8332,
+      "step": 15950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017337336666735977,
+      "loss": 2.805,
+      "step": 15960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001733344689332237,
+      "loss": 2.8409,
+      "step": 15970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017329554717815997,
+      "loss": 2.8026,
+      "step": 15980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017325660141491748,
+      "loss": 2.7206,
+      "step": 15990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017321763165625308,
+      "loss": 2.8222,
+      "step": 16000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001731786379149314,
+      "loss": 2.836,
+      "step": 16010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000173139620203725,
+      "loss": 2.7957,
+      "step": 16020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001731005785354142,
+      "loss": 2.737,
+      "step": 16030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017306151292278724,
+      "loss": 2.7849,
+      "step": 16040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001730224233786402,
+      "loss": 2.8386,
+      "step": 16050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017298330991577694,
+      "loss": 2.9137,
+      "step": 16060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017294417254700922,
+      "loss": 2.8477,
+      "step": 16070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017290501128515662,
+      "loss": 2.7599,
+      "step": 16080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017286582614304646,
+      "loss": 2.7999,
+      "step": 16090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017282661713351403,
+      "loss": 2.8544,
+      "step": 16100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017278738426940233,
+      "loss": 2.7988,
+      "step": 16110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001727481275635622,
+      "loss": 2.8953,
+      "step": 16120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017270884702885226,
+      "loss": 2.7554,
+      "step": 16130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017266954267813902,
+      "loss": 2.7036,
+      "step": 16140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017263021452429674,
+      "loss": 2.7541,
+      "step": 16150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017259086258020746,
+      "loss": 2.831,
+      "step": 16160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001725554255005192,
+      "loss": 2.8176,
+      "step": 16170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017251602839047858,
+      "loss": 2.7917,
+      "step": 16180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017247660752759295,
+      "loss": 2.8281,
+      "step": 16190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017243716292477478,
+      "loss": 2.7732,
+      "step": 16200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017239769459494425,
+      "loss": 2.8012,
+      "step": 16210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001723582025510293,
+      "loss": 2.7527,
+      "step": 16220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017231868680596572,
+      "loss": 2.7483,
+      "step": 16230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017227914737269698,
+      "loss": 2.7863,
+      "step": 16240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017223958426417432,
+      "loss": 2.8457,
+      "step": 16250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017219999749335682,
+      "loss": 2.7804,
+      "step": 16260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017216038707321122,
+      "loss": 2.9018,
+      "step": 16270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017212075301671197,
+      "loss": 2.7788,
+      "step": 16280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017208109533684143,
+      "loss": 2.805,
+      "step": 16290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017204141404658952,
+      "loss": 2.7992,
+      "step": 16300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000172001709158954,
+      "loss": 2.7677,
+      "step": 16310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001719619806869403,
+      "loss": 2.82,
+      "step": 16320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017192222864356166,
+      "loss": 2.8026,
+      "step": 16330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001718824530418389,
+      "loss": 2.7741,
+      "step": 16340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001718426538948007,
+      "loss": 2.7807,
+      "step": 16350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017180283121548336,
+      "loss": 2.754,
+      "step": 16360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017176298501693092,
+      "loss": 2.7757,
+      "step": 16370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017172311531219516,
+      "loss": 2.8049,
+      "step": 16380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017168322211433547,
+      "loss": 2.8367,
+      "step": 16390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017164330543641905,
+      "loss": 2.754,
+      "step": 16400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017160336529152063,
+      "loss": 2.8085,
+      "step": 16410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017156340169272282,
+      "loss": 2.805,
+      "step": 16420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001715234146531158,
+      "loss": 2.8988,
+      "step": 16430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017148340418579743,
+      "loss": 2.8403,
+      "step": 16440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017144337030387325,
+      "loss": 2.8547,
+      "step": 16450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017140331302045652,
+      "loss": 2.8366,
+      "step": 16460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017136323234866803,
+      "loss": 2.7689,
+      "step": 16470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017132312830163646,
+      "loss": 2.8008,
+      "step": 16480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001712830008924979,
+      "loss": 2.8867,
+      "step": 16490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017124285013439628,
+      "loss": 2.7388,
+      "step": 16500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017120267604048305,
+      "loss": 2.7848,
+      "step": 16510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001711624786239174,
+      "loss": 2.6922,
+      "step": 16520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001711222578978661,
+      "loss": 2.7744,
+      "step": 16530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001710820138755035,
+      "loss": 2.8596,
+      "step": 16540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017104174657001176,
+      "loss": 2.8308,
+      "step": 16550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017100145599458053,
+      "loss": 2.7423,
+      "step": 16560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017096114216240706,
+      "loss": 2.9013,
+      "step": 16570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017092080508669633,
+      "loss": 2.8298,
+      "step": 16580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001708804447806608,
+      "loss": 2.8333,
+      "step": 16590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001708400612575207,
+      "loss": 2.8003,
+      "step": 16600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017079965453050373,
+      "loss": 2.8075,
+      "step": 16610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001707592246128452,
+      "loss": 2.7465,
+      "step": 16620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001707187715177881,
+      "loss": 2.7346,
+      "step": 16630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017067829525858294,
+      "loss": 2.8067,
+      "step": 16640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001706377958484879,
+      "loss": 2.7989,
+      "step": 16650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017059727330076858,
+      "loss": 2.8021,
+      "step": 16660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017055672762869835,
+      "loss": 2.7519,
+      "step": 16670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017051615884555802,
+      "loss": 2.826,
+      "step": 16680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017047556696463606,
+      "loss": 2.8159,
+      "step": 16690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001704349519992284,
+      "loss": 2.7786,
+      "step": 16700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017039431396263867,
+      "loss": 2.8358,
+      "step": 16710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017035365286817794,
+      "loss": 2.8248,
+      "step": 16720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001703129687291649,
+      "loss": 2.8596,
+      "step": 16730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017027226155892573,
+      "loss": 2.8337,
+      "step": 16740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001702315313707942,
+      "loss": 2.813,
+      "step": 16750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001701907781781116,
+      "loss": 2.817,
+      "step": 16760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017015000199422677,
+      "loss": 2.8936,
+      "step": 16770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017010920283249607,
+      "loss": 2.8002,
+      "step": 16780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00017006838070628342,
+      "loss": 2.778,
+      "step": 16790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001700275356289602,
+      "loss": 2.8323,
+      "step": 16800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001699866676139053,
+      "loss": 2.8026,
+      "step": 16810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016994577667450524,
+      "loss": 2.8119,
+      "step": 16820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016990486282415394,
+      "loss": 2.771,
+      "step": 16830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016986392607625284,
+      "loss": 2.7806,
+      "step": 16840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001698229664442109,
+      "loss": 2.7357,
+      "step": 16850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001697819839414446,
+      "loss": 2.7574,
+      "step": 16860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016974097858137786,
+      "loss": 2.8391,
+      "step": 16870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001696999503774421,
+      "loss": 2.7713,
+      "step": 16880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016965889934307624,
+      "loss": 2.7782,
+      "step": 16890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016961782549172668,
+      "loss": 2.7961,
+      "step": 16900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016957672883684727,
+      "loss": 2.7787,
+      "step": 16910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001695356093918994,
+      "loss": 2.7742,
+      "step": 16920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001694944671703518,
+      "loss": 2.7579,
+      "step": 16930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016945330218568069,
+      "loss": 2.7735,
+      "step": 16940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001694121144513699,
+      "loss": 2.7466,
+      "step": 16950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016937090398091057,
+      "loss": 2.8751,
+      "step": 16960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016932967078780125,
+      "loss": 2.837,
+      "step": 16970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016928841488554809,
+      "loss": 2.756,
+      "step": 16980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016924713628766448,
+      "loss": 2.7481,
+      "step": 16990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001692058350076714,
+      "loss": 2.8744,
+      "step": 17000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016916451105909722,
+      "loss": 2.7689,
+      "step": 17010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016912316445547775,
+      "loss": 2.8006,
+      "step": 17020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016908179521035613,
+      "loss": 2.7571,
+      "step": 17030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016904040333728303,
+      "loss": 2.7424,
+      "step": 17040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016899898884981647,
+      "loss": 2.8445,
+      "step": 17050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016895755176152188,
+      "loss": 2.8398,
+      "step": 17060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016891609208597211,
+      "loss": 2.8671,
+      "step": 17070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001688746098367474,
+      "loss": 2.8487,
+      "step": 17080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001688331050274354,
+      "loss": 2.8736,
+      "step": 17090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016879157767163113,
+      "loss": 2.7988,
+      "step": 17100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000168750027782937,
+      "loss": 2.8284,
+      "step": 17110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001687084553749628,
+      "loss": 2.8394,
+      "step": 17120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001686668604613257,
+      "loss": 2.7947,
+      "step": 17130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016862524305565025,
+      "loss": 2.8229,
+      "step": 17140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001685836031715683,
+      "loss": 2.8209,
+      "step": 17150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016854194082271924,
+      "loss": 2.7725,
+      "step": 17160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016850025602274958,
+      "loss": 2.7916,
+      "step": 17170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016845854878531338,
+      "loss": 2.8027,
+      "step": 17180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001684168191240719,
+      "loss": 2.783,
+      "step": 17190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016837506705269387,
+      "loss": 2.7411,
+      "step": 17200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016833329258485527,
+      "loss": 2.8206,
+      "step": 17210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016829149573423954,
+      "loss": 2.7904,
+      "step": 17220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016824967651453727,
+      "loss": 2.777,
+      "step": 17230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016820783493944652,
+      "loss": 2.747,
+      "step": 17240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016816597102267257,
+      "loss": 2.7883,
+      "step": 17250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016812408477792816,
+      "loss": 2.7897,
+      "step": 17260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001680821762189332,
+      "loss": 2.7149,
+      "step": 17270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000168040245359415,
+      "loss": 2.7443,
+      "step": 17280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016799829221310808,
+      "loss": 2.7742,
+      "step": 17290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016795631679375434,
+      "loss": 2.7295,
+      "step": 17300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016791431911510301,
+      "loss": 2.804,
+      "step": 17310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016787229919091047,
+      "loss": 2.7104,
+      "step": 17320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016783025703494058,
+      "loss": 2.8249,
+      "step": 17330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016778819266096432,
+      "loss": 2.7639,
+      "step": 17340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016774610608275996,
+      "loss": 2.8273,
+      "step": 17350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016770399731411316,
+      "loss": 2.787,
+      "step": 17360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016766186636881671,
+      "loss": 2.8083,
+      "step": 17370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001676197132606708,
+      "loss": 2.8092,
+      "step": 17380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016757753800348275,
+      "loss": 2.7616,
+      "step": 17390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016753534061106722,
+      "loss": 2.7979,
+      "step": 17400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016749312109724606,
+      "loss": 2.8149,
+      "step": 17410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001674508794758485,
+      "loss": 2.7651,
+      "step": 17420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016740861576071075,
+      "loss": 2.8121,
+      "step": 17430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016736632996567656,
+      "loss": 2.8179,
+      "step": 17440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001673240221045967,
+      "loss": 2.7966,
+      "step": 17450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016728169219132927,
+      "loss": 2.7597,
+      "step": 17460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001672393402397395,
+      "loss": 2.7775,
+      "step": 17470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001671969662637,
+      "loss": 2.7973,
+      "step": 17480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016715457027709042,
+      "loss": 2.8243,
+      "step": 17490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001671121522937977,
+      "loss": 2.7233,
+      "step": 17500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016706971232771601,
+      "loss": 2.7947,
+      "step": 17510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001670272503927467,
+      "loss": 2.7361,
+      "step": 17520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016698476650279823,
+      "loss": 2.7608,
+      "step": 17530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016694226067178643,
+      "loss": 2.8324,
+      "step": 17540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001668997329136341,
+      "loss": 2.746,
+      "step": 17550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016685718324227145,
+      "loss": 2.8628,
+      "step": 17560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016681461167163567,
+      "loss": 2.7929,
+      "step": 17570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016677201821567125,
+      "loss": 2.8012,
+      "step": 17580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016672940288832978,
+      "loss": 2.75,
+      "step": 17590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016668676570357006,
+      "loss": 2.746,
+      "step": 17600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016664410667535803,
+      "loss": 2.7613,
+      "step": 17610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016660142581766674,
+      "loss": 2.7674,
+      "step": 17620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001665587231444765,
+      "loss": 2.8047,
+      "step": 17630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016651599866977467,
+      "loss": 2.7876,
+      "step": 17640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001664732524075558,
+      "loss": 2.7934,
+      "step": 17650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001664304843718215,
+      "loss": 2.7955,
+      "step": 17660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016638769457658062,
+      "loss": 2.7478,
+      "step": 17670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001663448830358491,
+      "loss": 2.8141,
+      "step": 17680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016630204976364994,
+      "loss": 2.8292,
+      "step": 17690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016625919477401336,
+      "loss": 2.7946,
+      "step": 17700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001662163180809766,
+      "loss": 2.7482,
+      "step": 17710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016617341969858406,
+      "loss": 2.7981,
+      "step": 17720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016613049964088724,
+      "loss": 2.8179,
+      "step": 17730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016608755792194474,
+      "loss": 2.7227,
+      "step": 17740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016604459455582226,
+      "loss": 2.7816,
+      "step": 17750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016600160955659255,
+      "loss": 2.672,
+      "step": 17760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001659586029383355,
+      "loss": 2.7563,
+      "step": 17770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016591557471513807,
+      "loss": 2.6879,
+      "step": 17780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016587252490109423,
+      "loss": 2.8775,
+      "step": 17790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016582945351030517,
+      "loss": 2.7716,
+      "step": 17800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016578636055687896,
+      "loss": 2.8443,
+      "step": 17810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001657432460549309,
+      "loss": 2.7336,
+      "step": 17820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001657001100185832,
+      "loss": 2.7399,
+      "step": 17830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016565695246196528,
+      "loss": 2.7403,
+      "step": 17840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001656137733992135,
+      "loss": 2.7536,
+      "step": 17850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016557057284447124,
+      "loss": 2.7578,
+      "step": 17860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016552735081188904,
+      "loss": 2.7914,
+      "step": 17870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016548410731562438,
+      "loss": 2.7923,
+      "step": 17880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001654408423698418,
+      "loss": 2.7951,
+      "step": 17890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016539755598871288,
+      "loss": 2.7983,
+      "step": 17900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016535424818641618,
+      "loss": 2.8124,
+      "step": 17910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016531091897713736,
+      "loss": 2.7834,
+      "step": 17920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016526756837506894,
+      "loss": 2.7905,
+      "step": 17930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001652241963944106,
+      "loss": 2.7818,
+      "step": 17940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016518080304936893,
+      "loss": 2.693,
+      "step": 17950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001651373883541576,
+      "loss": 2.7837,
+      "step": 17960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001650939523229972,
+      "loss": 2.7306,
+      "step": 17970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016505049497011535,
+      "loss": 2.7726,
+      "step": 17980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016500701630974663,
+      "loss": 2.7473,
+      "step": 17990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016496351635613256,
+      "loss": 2.7496,
+      "step": 18000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016491999512352175,
+      "loss": 2.8323,
+      "step": 18010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016487645262616967,
+      "loss": 2.7375,
+      "step": 18020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001648328888783388,
+      "loss": 2.8327,
+      "step": 18030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001647893038942986,
+      "loss": 2.7846,
+      "step": 18040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016474569768832543,
+      "loss": 2.8807,
+      "step": 18050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001647020702747027,
+      "loss": 2.7446,
+      "step": 18060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001646584216677206,
+      "loss": 2.8671,
+      "step": 18070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001646147518816765,
+      "loss": 2.7568,
+      "step": 18080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016457106093087443,
+      "loss": 2.8826,
+      "step": 18090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016452734882962554,
+      "loss": 2.6848,
+      "step": 18100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016448361559224793,
+      "loss": 2.7694,
+      "step": 18110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016443986123306647,
+      "loss": 2.7647,
+      "step": 18120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016439608576641304,
+      "loss": 2.8478,
+      "step": 18130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001643522892066265,
+      "loss": 2.6932,
+      "step": 18140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016430847156805246,
+      "loss": 2.8351,
+      "step": 18150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016426463286504359,
+      "loss": 2.7583,
+      "step": 18160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016422077311195935,
+      "loss": 2.742,
+      "step": 18170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016417689232316613,
+      "loss": 2.7747,
+      "step": 18180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016413299051303723,
+      "loss": 2.8436,
+      "step": 18190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016408906769595281,
+      "loss": 2.7222,
+      "step": 18200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001640451238863,
+      "loss": 2.6685,
+      "step": 18210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016400115909847256,
+      "loss": 2.7496,
+      "step": 18220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016395717334687145,
+      "loss": 2.8089,
+      "step": 18230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016391316664590423,
+      "loss": 2.8253,
+      "step": 18240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016386913900998552,
+      "loss": 2.7025,
+      "step": 18250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016382509045353665,
+      "loss": 2.7637,
+      "step": 18260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016378102099098586,
+      "loss": 2.8046,
+      "step": 18270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016373693063676822,
+      "loss": 2.7284,
+      "step": 18280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016369281940532571,
+      "loss": 2.8653,
+      "step": 18290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016364868731110705,
+      "loss": 2.7966,
+      "step": 18300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016360453436856787,
+      "loss": 2.752,
+      "step": 18310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016356036059217058,
+      "loss": 2.7609,
+      "step": 18320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016351616599638447,
+      "loss": 2.7556,
+      "step": 18330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016347195059568558,
+      "loss": 2.7536,
+      "step": 18340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016342771440455679,
+      "loss": 2.7187,
+      "step": 18350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016338345743748786,
+      "loss": 2.7796,
+      "step": 18360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016333917970897522,
+      "loss": 2.8558,
+      "step": 18370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016329488123352222,
+      "loss": 2.8219,
+      "step": 18380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016325056202563893,
+      "loss": 2.7469,
+      "step": 18390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016320622209984226,
+      "loss": 2.7916,
+      "step": 18400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016316186147065587,
+      "loss": 2.8748,
+      "step": 18410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016311748015261025,
+      "loss": 2.717,
+      "step": 18420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016307307816024262,
+      "loss": 2.6727,
+      "step": 18430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016302865550809695,
+      "loss": 2.7932,
+      "step": 18440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001629842122107241,
+      "loss": 2.7418,
+      "step": 18450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016293974828268152,
+      "loss": 2.8404,
+      "step": 18460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016289526373853357,
+      "loss": 2.7298,
+      "step": 18470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016285075859285124,
+      "loss": 2.7675,
+      "step": 18480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016280623286021238,
+      "loss": 2.7687,
+      "step": 18490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016276168655520145,
+      "loss": 2.718,
+      "step": 18500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001627171196924098,
+      "loss": 2.7862,
+      "step": 18510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016267253228643543,
+      "loss": 2.8187,
+      "step": 18520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016262792435188308,
+      "loss": 2.7027,
+      "step": 18530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016258329590336417,
+      "loss": 2.796,
+      "step": 18540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016253864695549695,
+      "loss": 2.8032,
+      "step": 18550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001624939775229063,
+      "loss": 2.7271,
+      "step": 18560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001624492876202238,
+      "loss": 2.749,
+      "step": 18570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016240457726208774,
+      "loss": 2.7137,
+      "step": 18580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001623598464631432,
+      "loss": 2.7466,
+      "step": 18590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016231509523804185,
+      "loss": 2.7104,
+      "step": 18600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001622703236014421,
+      "loss": 2.7345,
+      "step": 18610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016222553156800906,
+      "loss": 2.7403,
+      "step": 18620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016218071915241443,
+      "loss": 2.8228,
+      "step": 18630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016213588636933672,
+      "loss": 2.7333,
+      "step": 18640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000162091033233461,
+      "loss": 2.7432,
+      "step": 18650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016204615975947908,
+      "loss": 2.7037,
+      "step": 18660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001620012659620894,
+      "loss": 2.7618,
+      "step": 18670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000161956351855997,
+      "loss": 2.7488,
+      "step": 18680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016191141745591374,
+      "loss": 2.7264,
+      "step": 18690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016186646277655794,
+      "loss": 2.7886,
+      "step": 18700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016182148783265465,
+      "loss": 2.7069,
+      "step": 18710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016177649263893556,
+      "loss": 2.7515,
+      "step": 18720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016173147721013903,
+      "loss": 2.8095,
+      "step": 18730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001616864415610099,
+      "loss": 2.7623,
+      "step": 18740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016164138570629982,
+      "loss": 2.7688,
+      "step": 18750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001615963096607669,
+      "loss": 2.7323,
+      "step": 18760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000161551213439176,
+      "loss": 2.8057,
+      "step": 18770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001615060970562985,
+      "loss": 2.8108,
+      "step": 18780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001614609605269124,
+      "loss": 2.7498,
+      "step": 18790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001614158038658023,
+      "loss": 2.73,
+      "step": 18800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016137062708775944,
+      "loss": 2.8167,
+      "step": 18810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016132543020758159,
+      "loss": 2.74,
+      "step": 18820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001612802132400731,
+      "loss": 2.7977,
+      "step": 18830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016123497620004496,
+      "loss": 2.771,
+      "step": 18840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016118971910231467,
+      "loss": 2.7887,
+      "step": 18850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001611444419617063,
+      "loss": 2.7498,
+      "step": 18860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016109914479305065,
+      "loss": 2.686,
+      "step": 18870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001610538276111848,
+      "loss": 2.6906,
+      "step": 18880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016100849043095258,
+      "loss": 2.8141,
+      "step": 18890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016096313326720437,
+      "loss": 2.695,
+      "step": 18900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016091775613479695,
+      "loss": 2.7656,
+      "step": 18910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016087235904859384,
+      "loss": 2.7393,
+      "step": 18920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016082694202346488,
+      "loss": 2.7764,
+      "step": 18930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016078150507428668,
+      "loss": 2.7309,
+      "step": 18940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016073604821594212,
+      "loss": 2.6834,
+      "step": 18950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016069057146332085,
+      "loss": 2.7424,
+      "step": 18960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016064507483131885,
+      "loss": 2.7592,
+      "step": 18970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016059955833483872,
+      "loss": 2.7035,
+      "step": 18980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016055402198878946,
+      "loss": 2.6945,
+      "step": 18990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016050846580808673,
+      "loss": 2.8646,
+      "step": 19000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016046288980765252,
+      "loss": 2.8323,
+      "step": 19010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001604172940024154,
+      "loss": 2.6545,
+      "step": 19020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016037167840731047,
+      "loss": 2.6773,
+      "step": 19030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001603260430372792,
+      "loss": 2.7939,
+      "step": 19040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016028038790726957,
+      "loss": 2.7909,
+      "step": 19050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001602347130322361,
+      "loss": 2.7868,
+      "step": 19060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016018901842713976,
+      "loss": 2.8165,
+      "step": 19070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016014330410694792,
+      "loss": 2.7664,
+      "step": 19080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016009757008663444,
+      "loss": 2.7246,
+      "step": 19090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016005181638117965,
+      "loss": 2.7647,
+      "step": 19100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00016000604300557032,
+      "loss": 2.7532,
+      "step": 19110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015996024997479963,
+      "loss": 2.7743,
+      "step": 19120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015991443730386726,
+      "loss": 2.6578,
+      "step": 19130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015986860500777925,
+      "loss": 2.7618,
+      "step": 19140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015982275310154814,
+      "loss": 2.7006,
+      "step": 19150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001597768816001929,
+      "loss": 2.8215,
+      "step": 19160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001597309905187388,
+      "loss": 2.7073,
+      "step": 19170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001596850798722176,
+      "loss": 2.8101,
+      "step": 19180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015963914967566755,
+      "loss": 2.787,
+      "step": 19190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015959319994413316,
+      "loss": 2.7274,
+      "step": 19200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015954723069266545,
+      "loss": 2.8307,
+      "step": 19210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015950124193632174,
+      "loss": 2.6689,
+      "step": 19220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015945523369016582,
+      "loss": 2.6937,
+      "step": 19230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015940920596926782,
+      "loss": 2.773,
+      "step": 19240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015936315878870426,
+      "loss": 2.7769,
+      "step": 19250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015931709216355803,
+      "loss": 2.7217,
+      "step": 19260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001592710061089184,
+      "loss": 2.7107,
+      "step": 19270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000159224900639881,
+      "loss": 2.7202,
+      "step": 19280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001591787757715478,
+      "loss": 2.7545,
+      "step": 19290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015913263151902718,
+      "loss": 2.6773,
+      "step": 19300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015908646789743376,
+      "loss": 2.7541,
+      "step": 19310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001590402849218886,
+      "loss": 2.7656,
+      "step": 19320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015899408260751914,
+      "loss": 2.8368,
+      "step": 19330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000158947860969459,
+      "loss": 2.7538,
+      "step": 19340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015890162002284825,
+      "loss": 2.7658,
+      "step": 19350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015885535978283328,
+      "loss": 2.7537,
+      "step": 19360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015880908026456677,
+      "loss": 2.7521,
+      "step": 19370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015876278148320764,
+      "loss": 2.856,
+      "step": 19380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015871646345392128,
+      "loss": 2.7736,
+      "step": 19390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015867012619187926,
+      "loss": 2.6862,
+      "step": 19400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001586237697122595,
+      "loss": 2.7895,
+      "step": 19410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001585773940302462,
+      "loss": 2.7902,
+      "step": 19420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001585309991610299,
+      "loss": 2.7456,
+      "step": 19430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001584845851198073,
+      "loss": 2.8468,
+      "step": 19440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015843815192178153,
+      "loss": 2.6724,
+      "step": 19450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001583916995821619,
+      "loss": 2.6813,
+      "step": 19460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000158345228116164,
+      "loss": 2.7364,
+      "step": 19470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015829873753900973,
+      "loss": 2.7707,
+      "step": 19480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015825222786592717,
+      "loss": 2.8153,
+      "step": 19490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015820569911215076,
+      "loss": 2.782,
+      "step": 19500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001581591512929211,
+      "loss": 2.7578,
+      "step": 19510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015811258442348508,
+      "loss": 2.7634,
+      "step": 19520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015806599851909582,
+      "loss": 2.7278,
+      "step": 19530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015801939359501267,
+      "loss": 2.766,
+      "step": 19540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015797276966650125,
+      "loss": 2.7659,
+      "step": 19550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015792612674883332,
+      "loss": 2.7484,
+      "step": 19560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001578794648572869,
+      "loss": 2.7803,
+      "step": 19570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001578327840071463,
+      "loss": 2.7983,
+      "step": 19580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015778608421370189,
+      "loss": 2.7684,
+      "step": 19590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015773936549225037,
+      "loss": 2.7221,
+      "step": 19600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001576926278580946,
+      "loss": 2.7229,
+      "step": 19610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015764587132654373,
+      "loss": 2.7506,
+      "step": 19620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015759909591291283,
+      "loss": 2.8404,
+      "step": 19630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001575523016325234,
+      "loss": 2.7427,
+      "step": 19640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015750548850070308,
+      "loss": 2.7382,
+      "step": 19650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015745865653278562,
+      "loss": 2.7698,
+      "step": 19660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000157411805744111,
+      "loss": 2.6701,
+      "step": 19670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001573649361500253,
+      "loss": 2.6558,
+      "step": 19680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015731804776588078,
+      "loss": 2.7024,
+      "step": 19690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015727114060703598,
+      "loss": 2.7297,
+      "step": 19700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001572242146888554,
+      "loss": 2.6867,
+      "step": 19710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015717727002670975,
+      "loss": 2.7126,
+      "step": 19720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015713030663597595,
+      "loss": 2.7263,
+      "step": 19730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015708332453203699,
+      "loss": 2.7076,
+      "step": 19740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015703632373028197,
+      "loss": 2.7176,
+      "step": 19750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015698930424610613,
+      "loss": 2.6239,
+      "step": 19760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015694226609491092,
+      "loss": 2.7948,
+      "step": 19770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015689520929210377,
+      "loss": 2.6523,
+      "step": 19780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001568481338530983,
+      "loss": 2.7988,
+      "step": 19790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001568010397933142,
+      "loss": 2.7948,
+      "step": 19800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001567539271281773,
+      "loss": 2.7339,
+      "step": 19810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015670679587311947,
+      "loss": 2.7298,
+      "step": 19820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015665964604357868,
+      "loss": 2.8211,
+      "step": 19830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015661247765499903,
+      "loss": 2.7849,
+      "step": 19840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015656529072283069,
+      "loss": 2.7536,
+      "step": 19850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015651808526252983,
+      "loss": 2.7694,
+      "step": 19860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015647086128955871,
+      "loss": 2.7349,
+      "step": 19870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001564236188193858,
+      "loss": 2.7772,
+      "step": 19880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015637635786748546,
+      "loss": 2.6842,
+      "step": 19890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015632907844933812,
+      "loss": 2.665,
+      "step": 19900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015628178058043037,
+      "loss": 2.662,
+      "step": 19910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001562344642762547,
+      "loss": 2.8948,
+      "step": 19920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015618712955230976,
+      "loss": 2.7319,
+      "step": 19930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015613977642410013,
+      "loss": 2.7757,
+      "step": 19940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015609240490713653,
+      "loss": 2.8242,
+      "step": 19950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015604501501693559,
+      "loss": 2.6964,
+      "step": 19960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015599760676902004,
+      "loss": 2.8491,
+      "step": 19970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001559501801789186,
+      "loss": 2.6541,
+      "step": 19980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015590273526216598,
+      "loss": 2.8059,
+      "step": 19990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015585527203430295,
+      "loss": 2.7107,
+      "step": 20000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015580779051087615,
+      "loss": 2.7319,
+      "step": 20010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001557602907074384,
+      "loss": 2.7049,
+      "step": 20020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015571277263954828,
+      "loss": 2.69,
+      "step": 20030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015566523632277058,
+      "loss": 2.7461,
+      "step": 20040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001556176817726759,
+      "loss": 2.704,
+      "step": 20050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015557010900484096,
+      "loss": 2.7858,
+      "step": 20060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001555225180348483,
+      "loss": 2.7577,
+      "step": 20070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001554749088782865,
+      "loss": 2.7188,
+      "step": 20080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015542728155075003,
+      "loss": 2.7508,
+      "step": 20090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015537963606783946,
+      "loss": 2.7614,
+      "step": 20100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001553319724451612,
+      "loss": 2.7492,
+      "step": 20110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015528429069832752,
+      "loss": 2.7263,
+      "step": 20120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001552365908429568,
+      "loss": 2.8144,
+      "step": 20130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015518887289467327,
+      "loss": 2.687,
+      "step": 20140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015514113686910708,
+      "loss": 2.7554,
+      "step": 20150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001550933827818943,
+      "loss": 2.7284,
+      "step": 20160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015504561064867692,
+      "loss": 2.6662,
+      "step": 20170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001549978204851029,
+      "loss": 2.7311,
+      "step": 20180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015495001230682597,
+      "loss": 2.7394,
+      "step": 20190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001549021861295059,
+      "loss": 2.743,
+      "step": 20200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001548543419688083,
+      "loss": 2.7293,
+      "step": 20210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015480647984040467,
+      "loss": 2.7846,
+      "step": 20220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001547585997599723,
+      "loss": 2.6246,
+      "step": 20230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001547107017431946,
+      "loss": 2.6712,
+      "step": 20240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015466278580576065,
+      "loss": 2.7177,
+      "step": 20250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015461485196336544,
+      "loss": 2.6898,
+      "step": 20260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015456690023170987,
+      "loss": 2.7407,
+      "step": 20270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015451893062650064,
+      "loss": 2.7841,
+      "step": 20280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001544709431634504,
+      "loss": 2.8458,
+      "step": 20290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015442293785827756,
+      "loss": 2.7177,
+      "step": 20300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015437491472670636,
+      "loss": 2.7203,
+      "step": 20310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015432687378446697,
+      "loss": 2.6715,
+      "step": 20320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015427881504729535,
+      "loss": 2.7098,
+      "step": 20330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001542307385309333,
+      "loss": 2.7528,
+      "step": 20340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015418264425112835,
+      "loss": 2.721,
+      "step": 20350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015413453222363403,
+      "loss": 2.7958,
+      "step": 20360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001540864024642095,
+      "loss": 2.7251,
+      "step": 20370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015403825498861985,
+      "loss": 2.6831,
+      "step": 20380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015399008981263593,
+      "loss": 2.7772,
+      "step": 20390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015394190695203436,
+      "loss": 2.784,
+      "step": 20400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001538937064225976,
+      "loss": 2.6635,
+      "step": 20410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015384548824011392,
+      "loss": 2.6868,
+      "step": 20420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015379725242037725,
+      "loss": 2.7667,
+      "step": 20430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015374899897918742,
+      "loss": 2.7311,
+      "step": 20440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015370072793235,
+      "loss": 2.6391,
+      "step": 20450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015365243929567633,
+      "loss": 2.7,
+      "step": 20460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015360413308498347,
+      "loss": 2.6935,
+      "step": 20470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015355580931609425,
+      "loss": 2.7163,
+      "step": 20480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015350746800483735,
+      "loss": 2.8145,
+      "step": 20490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000153459109167047,
+      "loss": 2.7342,
+      "step": 20500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015341073281856336,
+      "loss": 2.6504,
+      "step": 20510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015336233897523226,
+      "loss": 2.8228,
+      "step": 20520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001533187695712408,
+      "loss": 2.6675,
+      "step": 20530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001532703425313752,
+      "loss": 2.7795,
+      "step": 20540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015322189804264734,
+      "loss": 2.8105,
+      "step": 20550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015317343612092542,
+      "loss": 2.718,
+      "step": 20560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015312980549928539,
+      "loss": 2.7152,
+      "step": 20570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015308131049861188,
+      "loss": 2.7477,
+      "step": 20580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015303279811099414,
+      "loss": 2.7541,
+      "step": 20590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015298426835232257,
+      "loss": 2.7632,
+      "step": 20600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015293572123849326,
+      "loss": 2.7326,
+      "step": 20610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015288715678540794,
+      "loss": 2.7914,
+      "step": 20620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015283857500897406,
+      "loss": 2.7749,
+      "step": 20630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015278997592510466,
+      "loss": 2.7378,
+      "step": 20640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015274135954971863,
+      "loss": 2.7256,
+      "step": 20650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015269272589874032,
+      "loss": 2.7346,
+      "step": 20660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015264407498809986,
+      "loss": 2.7561,
+      "step": 20670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015259540683373304,
+      "loss": 2.7157,
+      "step": 20680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015254672145158121,
+      "loss": 2.7591,
+      "step": 20690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015249801885759142,
+      "loss": 2.7036,
+      "step": 20700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001524492990677164,
+      "loss": 2.6785,
+      "step": 20710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001524005620979144,
+      "loss": 2.6594,
+      "step": 20720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001523518079641494,
+      "loss": 2.7488,
+      "step": 20730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015230303668239097,
+      "loss": 2.7353,
+      "step": 20740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015225424826861424,
+      "loss": 2.6718,
+      "step": 20750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001522054427388001,
+      "loss": 2.6583,
+      "step": 20760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015215662010893482,
+      "loss": 2.7337,
+      "step": 20770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015210778039501053,
+      "loss": 2.7845,
+      "step": 20780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001520589236130247,
+      "loss": 2.6942,
+      "step": 20790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015201004977898066,
+      "loss": 2.8251,
+      "step": 20800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001519611589088871,
+      "loss": 2.7954,
+      "step": 20810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015191225101875833,
+      "loss": 2.6657,
+      "step": 20820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015186332612461434,
+      "loss": 2.6709,
+      "step": 20830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015181438424248061,
+      "loss": 2.7979,
+      "step": 20840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015176542538838825,
+      "loss": 2.7096,
+      "step": 20850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015171644957837378,
+      "loss": 2.6988,
+      "step": 20860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015166745682847945,
+      "loss": 2.695,
+      "step": 20870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000151618447154753,
+      "loss": 2.6938,
+      "step": 20880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015156942057324764,
+      "loss": 2.7298,
+      "step": 20890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015152037710002217,
+      "loss": 2.744,
+      "step": 20900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015147131675114102,
+      "loss": 2.7394,
+      "step": 20910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015142223954267396,
+      "loss": 2.7949,
+      "step": 20920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001513731454906964,
+      "loss": 2.7543,
+      "step": 20930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015132403461128927,
+      "loss": 2.6816,
+      "step": 20940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015127490692053893,
+      "loss": 2.7787,
+      "step": 20950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001512257624345374,
+      "loss": 2.7848,
+      "step": 20960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015117660116938206,
+      "loss": 2.8004,
+      "step": 20970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001511274231411758,
+      "loss": 2.7383,
+      "step": 20980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015107822836602706,
+      "loss": 2.7191,
+      "step": 20990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001510290168600498,
+      "loss": 2.813,
+      "step": 21000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015097978863936327,
+      "loss": 2.6981,
+      "step": 21010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015093054372009238,
+      "loss": 2.8092,
+      "step": 21020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015088128211836747,
+      "loss": 2.7569,
+      "step": 21030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015083200385032432,
+      "loss": 2.7536,
+      "step": 21040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015078270893210415,
+      "loss": 2.8101,
+      "step": 21050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015073339737985372,
+      "loss": 2.6394,
+      "step": 21060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015068406920972516,
+      "loss": 2.7755,
+      "step": 21070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015063472443787597,
+      "loss": 2.7336,
+      "step": 21080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00015058536308046932,
+      "loss": 2.74,
+      "step": 21090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015053598515367357,
+      "loss": 2.716,
+      "step": 21100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015048659067366265,
+      "loss": 2.7168,
+      "step": 21110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015043717965661587,
+      "loss": 2.7088,
+      "step": 21120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015038775211871793,
+      "loss": 2.7301,
+      "step": 21130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015033830807615903,
+      "loss": 2.6422,
+      "step": 21140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015028884754513463,
+      "loss": 2.7346,
+      "step": 21150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001502393705418458,
+      "loss": 2.732,
+      "step": 21160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015018987708249876,
+      "loss": 2.7609,
+      "step": 21170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001501403671833053,
+      "loss": 2.714,
+      "step": 21180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015009084086048256,
+      "loss": 2.6448,
+      "step": 21190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015004129813025297,
+      "loss": 2.8385,
+      "step": 21200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014999173900884447,
+      "loss": 2.7753,
+      "step": 21210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014994216351249025,
+      "loss": 2.6628,
+      "step": 21220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001498925716574289,
+      "loss": 2.7229,
+      "step": 21230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014984296345990446,
+      "loss": 2.7417,
+      "step": 21240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001497933389361662,
+      "loss": 2.6896,
+      "step": 21250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014974369810246877,
+      "loss": 2.7438,
+      "step": 21260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001496940409750722,
+      "loss": 2.7229,
+      "step": 21270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014964436757024184,
+      "loss": 2.684,
+      "step": 21280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001495946779042483,
+      "loss": 2.7428,
+      "step": 21290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014954497199336765,
+      "loss": 2.747,
+      "step": 21300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014949524985388121,
+      "loss": 2.6723,
+      "step": 21310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014944551150207556,
+      "loss": 2.7017,
+      "step": 21320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014939575695424272,
+      "loss": 2.6889,
+      "step": 21330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001493459862266799,
+      "loss": 2.746,
+      "step": 21340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014929619933568965,
+      "loss": 2.6791,
+      "step": 21350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014924639629757984,
+      "loss": 2.7519,
+      "step": 21360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014919657712866358,
+      "loss": 2.6973,
+      "step": 21370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001491467418452593,
+      "loss": 2.7623,
+      "step": 21380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014909689046369075,
+      "loss": 2.7513,
+      "step": 21390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001490470230002868,
+      "loss": 2.6854,
+      "step": 21400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014899713947138173,
+      "loss": 2.8011,
+      "step": 21410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014894723989331511,
+      "loss": 2.7108,
+      "step": 21420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014889732428243163,
+      "loss": 2.7174,
+      "step": 21430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001488473926550813,
+      "loss": 2.7685,
+      "step": 21440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014879744502761938,
+      "loss": 2.7387,
+      "step": 21450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001487474814164064,
+      "loss": 2.6919,
+      "step": 21460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014869750183780803,
+      "loss": 2.6923,
+      "step": 21470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014864750630819528,
+      "loss": 2.7323,
+      "step": 21480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014859749484394435,
+      "loss": 2.6721,
+      "step": 21490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001485474674614366,
+      "loss": 2.6804,
+      "step": 21500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014849742417705867,
+      "loss": 2.8032,
+      "step": 21510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001484473650072024,
+      "loss": 2.6757,
+      "step": 21520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014839728996826482,
+      "loss": 2.7323,
+      "step": 21530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014834719907664819,
+      "loss": 2.6579,
+      "step": 21540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014829709234875986,
+      "loss": 2.6809,
+      "step": 21550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014824696980101246,
+      "loss": 2.7687,
+      "step": 21560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014819683144982386,
+      "loss": 2.6952,
+      "step": 21570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014814667731161696,
+      "loss": 2.8125,
+      "step": 21580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001480965074028199,
+      "loss": 2.7814,
+      "step": 21590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014804632173986596,
+      "loss": 2.7141,
+      "step": 21600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001479961203391937,
+      "loss": 2.6923,
+      "step": 21610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014794590321724665,
+      "loss": 2.6761,
+      "step": 21620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014789567039047365,
+      "loss": 2.7466,
+      "step": 21630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014784542187532855,
+      "loss": 2.6914,
+      "step": 21640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014779515768827047,
+      "loss": 2.7441,
+      "step": 21650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014774487784576353,
+      "loss": 2.6901,
+      "step": 21660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014769458236427708,
+      "loss": 2.6345,
+      "step": 21670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014764427126028558,
+      "loss": 2.7547,
+      "step": 21680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014759394455026855,
+      "loss": 2.6468,
+      "step": 21690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014754360225071063,
+      "loss": 2.8358,
+      "step": 21700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014749324437810166,
+      "loss": 2.784,
+      "step": 21710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014744287094893647,
+      "loss": 2.7218,
+      "step": 21720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014739248197971505,
+      "loss": 2.6992,
+      "step": 21730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014734207748694244,
+      "loss": 2.6977,
+      "step": 21740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001472916574871288,
+      "loss": 2.6616,
+      "step": 21750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014724122199678935,
+      "loss": 2.755,
+      "step": 21760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001471907710324444,
+      "loss": 2.8135,
+      "step": 21770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014714030461061928,
+      "loss": 2.8107,
+      "step": 21780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014708982274784448,
+      "loss": 2.7428,
+      "step": 21790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014703932546065543,
+      "loss": 2.6953,
+      "step": 21800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014698881276559264,
+      "loss": 2.6936,
+      "step": 21810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001469382846792018,
+      "loss": 2.7564,
+      "step": 21820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001468877412180335,
+      "loss": 2.7895,
+      "step": 21830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014683718239864334,
+      "loss": 2.7478,
+      "step": 21840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014678660823759206,
+      "loss": 2.7409,
+      "step": 21850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014673601875144545,
+      "loss": 2.7318,
+      "step": 21860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001466854139567741,
+      "loss": 2.76,
+      "step": 21870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014663479387015392,
+      "loss": 2.7723,
+      "step": 21880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014658415850816553,
+      "loss": 2.7567,
+      "step": 21890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001465335078873948,
+      "loss": 2.643,
+      "step": 21900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014648284202443246,
+      "loss": 2.7356,
+      "step": 21910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014643216093587427,
+      "loss": 2.635,
+      "step": 21920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014638146463832093,
+      "loss": 2.7221,
+      "step": 21930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014633075314837826,
+      "loss": 2.6714,
+      "step": 21940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001462800264826569,
+      "loss": 2.6765,
+      "step": 21950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001462292846577725,
+      "loss": 2.6695,
+      "step": 21960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014617852769034575,
+      "loss": 2.6916,
+      "step": 21970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014612775559700225,
+      "loss": 2.7364,
+      "step": 21980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001460769683943725,
+      "loss": 2.7165,
+      "step": 21990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014602616609909203,
+      "loss": 2.6742,
+      "step": 22000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001459753487278013,
+      "loss": 2.7318,
+      "step": 22010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014592451629714574,
+      "loss": 2.7778,
+      "step": 22020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014587366882377553,
+      "loss": 2.7038,
+      "step": 22030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014582280632434604,
+      "loss": 2.7361,
+      "step": 22040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001457719288155174,
+      "loss": 2.6822,
+      "step": 22050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014572103631395463,
+      "loss": 2.7903,
+      "step": 22060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014567012883632782,
+      "loss": 2.7905,
+      "step": 22070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001456192063993118,
+      "loss": 2.6545,
+      "step": 22080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014556826901958645,
+      "loss": 2.8058,
+      "step": 22090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014551731671383635,
+      "loss": 2.6541,
+      "step": 22100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014546634949875114,
+      "loss": 2.7656,
+      "step": 22110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014541536739102532,
+      "loss": 2.7469,
+      "step": 22120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001453643704073582,
+      "loss": 2.7594,
+      "step": 22130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014531335856445398,
+      "loss": 2.7056,
+      "step": 22140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014526233187902175,
+      "loss": 2.7079,
+      "step": 22150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014521129036777547,
+      "loss": 2.7001,
+      "step": 22160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014516023404743394,
+      "loss": 2.7898,
+      "step": 22170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014510916293472082,
+      "loss": 2.6482,
+      "step": 22180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001450580770463646,
+      "loss": 2.6631,
+      "step": 22190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001450069763990986,
+      "loss": 2.7013,
+      "step": 22200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000144955861009661,
+      "loss": 2.7463,
+      "step": 22210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001449047308947948,
+      "loss": 2.7401,
+      "step": 22220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001448535860712479,
+      "loss": 2.7186,
+      "step": 22230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001448024265557728,
+      "loss": 2.719,
+      "step": 22240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014475125236512704,
+      "loss": 2.7679,
+      "step": 22250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014470006351607287,
+      "loss": 2.7483,
+      "step": 22260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001446488600253774,
+      "loss": 2.6986,
+      "step": 22270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001445976419098124,
+      "loss": 2.7859,
+      "step": 22280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014454640918615456,
+      "loss": 2.6637,
+      "step": 22290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014449516187118534,
+      "loss": 2.7067,
+      "step": 22300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001444438999816909,
+      "loss": 2.7429,
+      "step": 22310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014439262353446229,
+      "loss": 2.6646,
+      "step": 22320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001443413325462952,
+      "loss": 2.7351,
+      "step": 22330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014429002703399016,
+      "loss": 2.6265,
+      "step": 22340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001442387070143525,
+      "loss": 2.6105,
+      "step": 22350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001441873725041922,
+      "loss": 2.6801,
+      "step": 22360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014413602352032407,
+      "loss": 2.7197,
+      "step": 22370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001440846600795676,
+      "loss": 2.7517,
+      "step": 22380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014403328219874704,
+      "loss": 2.6817,
+      "step": 22390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014398188989469136,
+      "loss": 2.7716,
+      "step": 22400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014393048318423428,
+      "loss": 2.7196,
+      "step": 22410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014387906208421423,
+      "loss": 2.731,
+      "step": 22420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014382762661147434,
+      "loss": 2.6808,
+      "step": 22430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014377617678286244,
+      "loss": 2.6679,
+      "step": 22440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014372471261523111,
+      "loss": 2.6433,
+      "step": 22450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001436732341254376,
+      "loss": 2.7429,
+      "step": 22460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014362174133034382,
+      "loss": 2.6631,
+      "step": 22470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014357023424681637,
+      "loss": 2.8399,
+      "step": 22480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014351871289172658,
+      "loss": 2.7728,
+      "step": 22490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014346717728195045,
+      "loss": 2.7668,
+      "step": 22500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014341562743436855,
+      "loss": 2.6185,
+      "step": 22510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014336406336586626,
+      "loss": 2.7796,
+      "step": 22520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014331248509333352,
+      "loss": 2.7239,
+      "step": 22530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014326089263366496,
+      "loss": 2.7723,
+      "step": 22540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014320928600375984,
+      "loss": 2.6278,
+      "step": 22550
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014315766522052204,
+      "loss": 2.6796,
+      "step": 22560
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014310603030086018,
+      "loss": 2.7273,
+      "step": 22570
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014305438126168733,
+      "loss": 2.7127,
+      "step": 22580
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001430027181199214,
+      "loss": 2.6422,
+      "step": 22590
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001429510408924847,
+      "loss": 2.7338,
+      "step": 22600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014289934959630437,
+      "loss": 2.7889,
+      "step": 22610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014284764424831193,
+      "loss": 2.6941,
+      "step": 22620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014279592486544376,
+      "loss": 2.7124,
+      "step": 22630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014274419146464062,
+      "loss": 2.7101,
+      "step": 22640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014269244406284792,
+      "loss": 2.7017,
+      "step": 22650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014264068267701572,
+      "loss": 2.6391,
+      "step": 22660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014258890732409864,
+      "loss": 2.7083,
+      "step": 22670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001425371180210558,
+      "loss": 2.7225,
+      "step": 22680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014248531478485097,
+      "loss": 2.7597,
+      "step": 22690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014243349763245244,
+      "loss": 2.742,
+      "step": 22700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001423816665808331,
+      "loss": 2.6301,
+      "step": 22710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001423298216469704,
+      "loss": 2.6993,
+      "step": 22720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014227796284784622,
+      "loss": 2.6375,
+      "step": 22730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014222609020044712,
+      "loss": 2.6783,
+      "step": 22740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014217420372176416,
+      "loss": 2.723,
+      "step": 22750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014212230342879284,
+      "loss": 2.6615,
+      "step": 22760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014207038933853336,
+      "loss": 2.6848,
+      "step": 22770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014201846146799025,
+      "loss": 2.781,
+      "step": 22780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014196651983417269,
+      "loss": 2.7071,
+      "step": 22790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001419145644540943,
+      "loss": 2.7238,
+      "step": 22800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014186259534477319,
+      "loss": 2.6999,
+      "step": 22810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014181061252323209,
+      "loss": 2.6859,
+      "step": 22820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014175861600649806,
+      "loss": 2.7344,
+      "step": 22830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014170660581160272,
+      "loss": 2.6097,
+      "step": 22840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014165458195558218,
+      "loss": 2.6826,
+      "step": 22850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014160254445547702,
+      "loss": 2.6388,
+      "step": 22860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014155049332833228,
+      "loss": 2.6891,
+      "step": 22870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014149842859119744,
+      "loss": 2.8171,
+      "step": 22880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001414463502611265,
+      "loss": 2.7277,
+      "step": 22890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014139425835517786,
+      "loss": 2.7167,
+      "step": 22900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014134215289041433,
+      "loss": 2.7274,
+      "step": 22910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014129003388390333,
+      "loss": 2.8325,
+      "step": 22920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001412379013527165,
+      "loss": 2.7535,
+      "step": 22930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014118575531393002,
+      "loss": 2.6545,
+      "step": 22940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014113359578462454,
+      "loss": 2.681,
+      "step": 22950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014108142278188498,
+      "loss": 2.6005,
+      "step": 22960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014102923632280085,
+      "loss": 2.6567,
+      "step": 22970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014097703642446596,
+      "loss": 2.74,
+      "step": 22980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001409248231039785,
+      "loss": 2.6758,
+      "step": 22990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014087259637844118,
+      "loss": 2.6205,
+      "step": 23000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014082035626496094,
+      "loss": 2.699,
+      "step": 23010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014076810278064924,
+      "loss": 2.6557,
+      "step": 23020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014071583594262184,
+      "loss": 2.6958,
+      "step": 23030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014066355576799897,
+      "loss": 2.7001,
+      "step": 23040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014061126227390508,
+      "loss": 2.6947,
+      "step": 23050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014055895547746905,
+      "loss": 2.5979,
+      "step": 23060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001405066353958242,
+      "loss": 2.7026,
+      "step": 23070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001404543020461081,
+      "loss": 2.6958,
+      "step": 23080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001404019554454627,
+      "loss": 2.653,
+      "step": 23090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014034959561103424,
+      "loss": 2.6806,
+      "step": 23100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014029722255997344,
+      "loss": 2.6239,
+      "step": 23110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014024483630943518,
+      "loss": 2.7089,
+      "step": 23120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014019243687657875,
+      "loss": 2.6802,
+      "step": 23130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014014002427856776,
+      "loss": 2.6758,
+      "step": 23140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014008759853257008,
+      "loss": 2.6941,
+      "step": 23150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00014003515965575798,
+      "loss": 2.7214,
+      "step": 23160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001399827076653079,
+      "loss": 2.7635,
+      "step": 23170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001399302425784007,
+      "loss": 2.6466,
+      "step": 23180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013987776441222146,
+      "loss": 2.7407,
+      "step": 23190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013982527318395958,
+      "loss": 2.7196,
+      "step": 23200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001397727689108087,
+      "loss": 2.7343,
+      "step": 23210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013972025160996675,
+      "loss": 2.6659,
+      "step": 23220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013966772129863598,
+      "loss": 2.7176,
+      "step": 23230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013961517799402278,
+      "loss": 2.6925,
+      "step": 23240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001395626217133379,
+      "loss": 2.7605,
+      "step": 23250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013951005247379633,
+      "loss": 2.721,
+      "step": 23260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013945747029261727,
+      "loss": 2.6122,
+      "step": 23270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013940487518702415,
+      "loss": 2.641,
+      "step": 23280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001393522671742447,
+      "loss": 2.6135,
+      "step": 23290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013929964627151082,
+      "loss": 2.6247,
+      "step": 23300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013924701249605858,
+      "loss": 2.7593,
+      "step": 23310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013919436586512843,
+      "loss": 2.7614,
+      "step": 23320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001391417063959649,
+      "loss": 2.6264,
+      "step": 23330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013908903410581673,
+      "loss": 2.7066,
+      "step": 23340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001390363490119369,
+      "loss": 2.7264,
+      "step": 23350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013898365113158262,
+      "loss": 2.6689,
+      "step": 23360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001389309404820152,
+      "loss": 2.7244,
+      "step": 23370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013887821708050016,
+      "loss": 2.8185,
+      "step": 23380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013882548094430722,
+      "loss": 2.7146,
+      "step": 23390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001387727320907103,
+      "loss": 2.74,
+      "step": 23400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001387199705369874,
+      "loss": 2.7476,
+      "step": 23410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013866719630042072,
+      "loss": 2.721,
+      "step": 23420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013861440939829663,
+      "loss": 2.6641,
+      "step": 23430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001385616098479057,
+      "loss": 2.7156,
+      "step": 23440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001385087976665425,
+      "loss": 2.7076,
+      "step": 23450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013845597287150588,
+      "loss": 2.745,
+      "step": 23460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001384031354800987,
+      "loss": 2.6312,
+      "step": 23470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001383502855096281,
+      "loss": 2.7598,
+      "step": 23480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013829742297740514,
+      "loss": 2.7158,
+      "step": 23490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013824454790074519,
+      "loss": 2.6798,
+      "step": 23500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001381916602969676,
+      "loss": 2.7081,
+      "step": 23510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013813876018339583,
+      "loss": 2.6004,
+      "step": 23520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013808584757735759,
+      "loss": 2.6863,
+      "step": 23530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013803292249618443,
+      "loss": 2.7632,
+      "step": 23540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013797998495721221,
+      "loss": 2.645,
+      "step": 23550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013792703497778077,
+      "loss": 2.6463,
+      "step": 23560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013787407257523398,
+      "loss": 2.6695,
+      "step": 23570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013782109776691988,
+      "loss": 2.6054,
+      "step": 23580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013776811057019054,
+      "loss": 2.7125,
+      "step": 23590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013771511100240205,
+      "loss": 2.6672,
+      "step": 23600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013766209908091455,
+      "loss": 2.7067,
+      "step": 23610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013760907482309232,
+      "loss": 2.6156,
+      "step": 23620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013755603824630363,
+      "loss": 2.843,
+      "step": 23630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013750829480883555,
+      "loss": 2.7367,
+      "step": 23640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013745523487387445,
+      "loss": 2.6752,
+      "step": 23650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013740216267033763,
+      "loss": 2.7265,
+      "step": 23660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000137349078215609,
+      "loss": 2.657,
+      "step": 23670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013729598152707655,
+      "loss": 2.7196,
+      "step": 23680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013724287262213228,
+      "loss": 2.6186,
+      "step": 23690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013718975151817215,
+      "loss": 2.5905,
+      "step": 23700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013713661823259608,
+      "loss": 2.8019,
+      "step": 23710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013708347278280818,
+      "loss": 2.682,
+      "step": 23720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013703031518621627,
+      "loss": 2.7472,
+      "step": 23730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013697714546023236,
+      "loss": 2.7288,
+      "step": 23740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001369292823506108,
+      "loss": 2.625,
+      "step": 23750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013687608962676615,
+      "loss": 2.7256,
+      "step": 23760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001368228848240465,
+      "loss": 2.5728,
+      "step": 23770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013676966795987927,
+      "loss": 2.6909,
+      "step": 23780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001367164390516958,
+      "loss": 2.6917,
+      "step": 23790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001366631981169314,
+      "loss": 2.7467,
+      "step": 23800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001366099451730253,
+      "loss": 2.6735,
+      "step": 23810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013655668023742062,
+      "loss": 2.6752,
+      "step": 23820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013650340332756447,
+      "loss": 2.5948,
+      "step": 23830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013645011446090788,
+      "loss": 2.6597,
+      "step": 23840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001363968136549058,
+      "loss": 2.6952,
+      "step": 23850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000136343500927017,
+      "loss": 2.6971,
+      "step": 23860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001362901762947043,
+      "loss": 2.6622,
+      "step": 23870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013623683977543428,
+      "loss": 2.581,
+      "step": 23880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013618349138667756,
+      "loss": 2.6647,
+      "step": 23890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013613013114590846,
+      "loss": 2.6405,
+      "step": 23900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001360767590706054,
+      "loss": 2.686,
+      "step": 23910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013602337517825048,
+      "loss": 2.7012,
+      "step": 23920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001359699794863298,
+      "loss": 2.7884,
+      "step": 23930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013591657201233328,
+      "loss": 2.7038,
+      "step": 23940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013586315277375468,
+      "loss": 2.6983,
+      "step": 23950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013580972178809162,
+      "loss": 2.5843,
+      "step": 23960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001357562790728456,
+      "loss": 2.7072,
+      "step": 23970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013570282464552198,
+      "loss": 2.6556,
+      "step": 23980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013564935852362987,
+      "loss": 2.6763,
+      "step": 23990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001355958807246823,
+      "loss": 2.6535,
+      "step": 24000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013554239126619604,
+      "loss": 2.6564,
+      "step": 24010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013548889016569174,
+      "loss": 2.7205,
+      "step": 24020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013543537744069387,
+      "loss": 2.7567,
+      "step": 24030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013538185310873067,
+      "loss": 2.7189,
+      "step": 24040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013532831718733417,
+      "loss": 2.7561,
+      "step": 24050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013527476969404028,
+      "loss": 2.6202,
+      "step": 24060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001352212106463886,
+      "loss": 2.7012,
+      "step": 24070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013516764006192257,
+      "loss": 2.6545,
+      "step": 24080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013511405795818944,
+      "loss": 2.6225,
+      "step": 24090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013506046435274012,
+      "loss": 2.6752,
+      "step": 24100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001350068592631294,
+      "loss": 2.6823,
+      "step": 24110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013495324270691573,
+      "loss": 2.7115,
+      "step": 24120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001348996147016615,
+      "loss": 2.7192,
+      "step": 24130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013484597526493256,
+      "loss": 2.7164,
+      "step": 24140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013479232441429884,
+      "loss": 2.7175,
+      "step": 24150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001347386621673337,
+      "loss": 2.6902,
+      "step": 24160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013468498854161447,
+      "loss": 2.6477,
+      "step": 24170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013463130355472207,
+      "loss": 2.6805,
+      "step": 24180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013457760722424114,
+      "loss": 2.611,
+      "step": 24190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013452389956776015,
+      "loss": 2.7243,
+      "step": 24200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013447018060287118,
+      "loss": 2.7553,
+      "step": 24210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013441645034717,
+      "loss": 2.7228,
+      "step": 24220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013436270881825616,
+      "loss": 2.6699,
+      "step": 24230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013430895603373287,
+      "loss": 2.617,
+      "step": 24240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013425519201120696,
+      "loss": 2.6809,
+      "step": 24250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013420141676828904,
+      "loss": 2.6714,
+      "step": 24260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013414763032259337,
+      "loss": 2.6926,
+      "step": 24270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013409383269173786,
+      "loss": 2.5747,
+      "step": 24280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013404002389334405,
+      "loss": 2.6385,
+      "step": 24290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013398620394503716,
+      "loss": 2.6434,
+      "step": 24300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013393237286444615,
+      "loss": 2.6385,
+      "step": 24310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013387853066920352,
+      "loss": 2.6632,
+      "step": 24320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001338246773769454,
+      "loss": 2.7371,
+      "step": 24330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001337708130053117,
+      "loss": 2.7659,
+      "step": 24340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013371693757194576,
+      "loss": 2.6295,
+      "step": 24350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001336630510944947,
+      "loss": 2.7329,
+      "step": 24360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013360915359060914,
+      "loss": 2.7251,
+      "step": 24370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013355524507794347,
+      "loss": 2.6456,
+      "step": 24380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013350132557415548,
+      "loss": 2.6582,
+      "step": 24390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013344739509690674,
+      "loss": 2.6461,
+      "step": 24400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013339345366386227,
+      "loss": 2.7416,
+      "step": 24410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013333950129269082,
+      "loss": 2.7072,
+      "step": 24420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013328553800106465,
+      "loss": 2.7799,
+      "step": 24430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013323156380665955,
+      "loss": 2.7196,
+      "step": 24440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013317757872715496,
+      "loss": 2.7262,
+      "step": 24450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013312358278023387,
+      "loss": 2.6379,
+      "step": 24460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001330695759835828,
+      "loss": 2.6867,
+      "step": 24470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001330155583548918,
+      "loss": 2.741,
+      "step": 24480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001329615299118546,
+      "loss": 2.599,
+      "step": 24490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013290749067216832,
+      "loss": 2.7022,
+      "step": 24500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013285344065353368,
+      "loss": 2.7317,
+      "step": 24510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013279937987365493,
+      "loss": 2.685,
+      "step": 24520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013274530835023982,
+      "loss": 2.7313,
+      "step": 24530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001326912261009997,
+      "loss": 2.6035,
+      "step": 24540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013263713314364933,
+      "loss": 2.6815,
+      "step": 24550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013258302949590705,
+      "loss": 2.6506,
+      "step": 24560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013252891517549466,
+      "loss": 2.6217,
+      "step": 24570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013247479020013744,
+      "loss": 2.722,
+      "step": 24580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013242065458756423,
+      "loss": 2.7081,
+      "step": 24590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001323665083555073,
+      "loss": 2.7286,
+      "step": 24600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013231235152170242,
+      "loss": 2.6095,
+      "step": 24610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013225818410388882,
+      "loss": 2.5863,
+      "step": 24620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001322040061198092,
+      "loss": 2.7587,
+      "step": 24630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001321498175872097,
+      "loss": 2.7305,
+      "step": 24640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013209561852384002,
+      "loss": 2.7458,
+      "step": 24650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013204140894745314,
+      "loss": 2.7168,
+      "step": 24660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001319871888758056,
+      "loss": 2.65,
+      "step": 24670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013193295832665737,
+      "loss": 2.6896,
+      "step": 24680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013187871731777174,
+      "loss": 2.6825,
+      "step": 24690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013182446586691566,
+      "loss": 2.7736,
+      "step": 24700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013177020399185923,
+      "loss": 2.6889,
+      "step": 24710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001317159317103762,
+      "loss": 2.7002,
+      "step": 24720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013166164904024355,
+      "loss": 2.6801,
+      "step": 24730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013160735599924173,
+      "loss": 2.6965,
+      "step": 24740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013155305260515463,
+      "loss": 2.6586,
+      "step": 24750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013149873887576945,
+      "loss": 2.6893,
+      "step": 24760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013144441482887685,
+      "loss": 2.6534,
+      "step": 24770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013139008048227086,
+      "loss": 2.7165,
+      "step": 24780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013133573585374876,
+      "loss": 2.6307,
+      "step": 24790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013128138096111142,
+      "loss": 2.6672,
+      "step": 24800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013122701582216287,
+      "loss": 2.6787,
+      "step": 24810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013117264045471057,
+      "loss": 2.592,
+      "step": 24820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013111825487656544,
+      "loss": 2.7031,
+      "step": 24830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001310638591055415,
+      "loss": 2.6763,
+      "step": 24840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013100945315945635,
+      "loss": 2.594,
+      "step": 24850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013095503705613077,
+      "loss": 2.6684,
+      "step": 24860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013090061081338895,
+      "loss": 2.6329,
+      "step": 24870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013084617444905834,
+      "loss": 2.6892,
+      "step": 24880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013079172798096975,
+      "loss": 2.7063,
+      "step": 24890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013073727142695732,
+      "loss": 2.6878,
+      "step": 24900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013068280480485844,
+      "loss": 2.6435,
+      "step": 24910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013062832813251377,
+      "loss": 2.6777,
+      "step": 24920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013057384142776737,
+      "loss": 2.6163,
+      "step": 24930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001305193447084665,
+      "loss": 2.6328,
+      "step": 24940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013046483799246175,
+      "loss": 2.7748,
+      "step": 24950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001304103212976069,
+      "loss": 2.6409,
+      "step": 24960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013035579464175907,
+      "loss": 2.5843,
+      "step": 24970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001303012580427787,
+      "loss": 2.6758,
+      "step": 24980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013024671151852938,
+      "loss": 2.7069,
+      "step": 24990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000130192155086878,
+      "loss": 2.6971,
+      "step": 25000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013013758876569465,
+      "loss": 2.6273,
+      "step": 25010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013008301257285272,
+      "loss": 2.6655,
+      "step": 25020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00013002842652622877,
+      "loss": 2.7208,
+      "step": 25030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001299738306437027,
+      "loss": 2.6248,
+      "step": 25040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001299192249431575,
+      "loss": 2.7011,
+      "step": 25050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012986460944247944,
+      "loss": 2.6428,
+      "step": 25060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012980998415955798,
+      "loss": 2.7719,
+      "step": 25070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012975534911228585,
+      "loss": 2.5946,
+      "step": 25080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001297007043185589,
+      "loss": 2.7026,
+      "step": 25090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012964604979627618,
+      "loss": 2.7008,
+      "step": 25100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001295913855633399,
+      "loss": 2.747,
+      "step": 25110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012953671163765562,
+      "loss": 2.7272,
+      "step": 25120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012948202803713188,
+      "loss": 2.703,
+      "step": 25130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001294273347796804,
+      "loss": 2.627,
+      "step": 25140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001293726318832162,
+      "loss": 2.7101,
+      "step": 25150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001293179193656574,
+      "loss": 2.5712,
+      "step": 25160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012926319724492519,
+      "loss": 2.6969,
+      "step": 25170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000129208465538944,
+      "loss": 2.7268,
+      "step": 25180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012915372426564135,
+      "loss": 2.6974,
+      "step": 25190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001290989734429479,
+      "loss": 2.688,
+      "step": 25200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012904421308879745,
+      "loss": 2.5982,
+      "step": 25210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000128989443221127,
+      "loss": 2.6046,
+      "step": 25220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012893466385787645,
+      "loss": 2.6203,
+      "step": 25230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012887987501698903,
+      "loss": 2.6056,
+      "step": 25240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000128825076716411,
+      "loss": 2.696,
+      "step": 25250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012877026897409165,
+      "loss": 2.6304,
+      "step": 25260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012871545180798347,
+      "loss": 2.6106,
+      "step": 25270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000128660625236042,
+      "loss": 2.6627,
+      "step": 25280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001286057892762258,
+      "loss": 2.753,
+      "step": 25290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001285509439464966,
+      "loss": 2.7295,
+      "step": 25300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012849608926481914,
+      "loss": 2.6748,
+      "step": 25310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012844122524916123,
+      "loss": 2.654,
+      "step": 25320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012838635191749374,
+      "loss": 2.5974,
+      "step": 25330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001283314692877906,
+      "loss": 2.6414,
+      "step": 25340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001282765773780288,
+      "loss": 2.6447,
+      "step": 25350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012822167620618832,
+      "loss": 2.7306,
+      "step": 25360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012816676579025223,
+      "loss": 2.7392,
+      "step": 25370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001281118461482066,
+      "loss": 2.556,
+      "step": 25380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012805691729804054,
+      "loss": 2.6917,
+      "step": 25390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012800197925774614,
+      "loss": 2.7279,
+      "step": 25400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001279470320453185,
+      "loss": 2.5659,
+      "step": 25410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001278920756787558,
+      "loss": 2.6185,
+      "step": 25420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012783711017605912,
+      "loss": 2.7416,
+      "step": 25430
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012778213555523254,
+      "loss": 2.694,
+      "step": 25440
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001277271518342833,
+      "loss": 2.6366,
+      "step": 25450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012767215903122136,
+      "loss": 2.8283,
+      "step": 25460
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012761715716405982,
+      "loss": 2.6175,
+      "step": 25470
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012756214625081472,
+      "loss": 2.6884,
+      "step": 25480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012750712630950502,
+      "loss": 2.6469,
+      "step": 25490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012745209735815272,
+      "loss": 2.6514,
+      "step": 25500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012739705941478265,
+      "loss": 2.7389,
+      "step": 25510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001273420124974227,
+      "loss": 2.5644,
+      "step": 25520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001272869566241037,
+      "loss": 2.7025,
+      "step": 25530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012723189181285927,
+      "loss": 2.6905,
+      "step": 25540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001271768180817261,
+      "loss": 2.6645,
+      "step": 25550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012712173544874377,
+      "loss": 2.691,
+      "step": 25560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012706664393195475,
+      "loss": 2.7878,
+      "step": 25570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012701154354940444,
+      "loss": 2.6942,
+      "step": 25580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012695643431914113,
+      "loss": 2.7493,
+      "step": 25590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012690131625921604,
+      "loss": 2.6525,
+      "step": 25600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001268461893876832,
+      "loss": 2.7048,
+      "step": 25610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012679105372259962,
+      "loss": 2.5927,
+      "step": 25620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001267359092820252,
+      "loss": 2.7784,
+      "step": 25630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001266807560840226,
+      "loss": 2.6885,
+      "step": 25640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012662559414665744,
+      "loss": 2.7298,
+      "step": 25650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012657042348799817,
+      "loss": 2.6121,
+      "step": 25660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012651524412611616,
+      "loss": 2.5925,
+      "step": 25670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001264600560790855,
+      "loss": 2.6464,
+      "step": 25680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012640485936498322,
+      "loss": 2.6422,
+      "step": 25690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012634965400188923,
+      "loss": 2.6276,
+      "step": 25700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012629444000788616,
+      "loss": 2.6272,
+      "step": 25710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012623921740105953,
+      "loss": 2.7075,
+      "step": 25720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012618398619949764,
+      "loss": 2.6633,
+      "step": 25730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012612874642129173,
+      "loss": 2.6484,
+      "step": 25740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012607349808453568,
+      "loss": 2.6245,
+      "step": 25750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012601824120732626,
+      "loss": 2.6938,
+      "step": 25760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012596297580776307,
+      "loss": 2.7927,
+      "step": 25770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001259077019039484,
+      "loss": 2.6484,
+      "step": 25780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001258524195139874,
+      "loss": 2.6999,
+      "step": 25790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012579712865598802,
+      "loss": 2.71,
+      "step": 25800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012574182934806092,
+      "loss": 2.676,
+      "step": 25810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012568652160831957,
+      "loss": 2.7379,
+      "step": 25820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012563120545488015,
+      "loss": 2.6973,
+      "step": 25830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012557588090586169,
+      "loss": 2.7221,
+      "step": 25840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012552054797938588,
+      "loss": 2.6188,
+      "step": 25850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001254652066935772,
+      "loss": 2.723,
+      "step": 25860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012540985706656284,
+      "loss": 2.6901,
+      "step": 25870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012535449911647273,
+      "loss": 2.7593,
+      "step": 25880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012529913286143956,
+      "loss": 2.7126,
+      "step": 25890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012524375831959874,
+      "loss": 2.6313,
+      "step": 25900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001251883755090883,
+      "loss": 2.7028,
+      "step": 25910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012513298444804912,
+      "loss": 2.7125,
+      "step": 25920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012507758515462465,
+      "loss": 2.6642,
+      "step": 25930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012502217764696113,
+      "loss": 2.718,
+      "step": 25940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012496676194320746,
+      "loss": 2.6765,
+      "step": 25950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012491133806151518,
+      "loss": 2.7217,
+      "step": 25960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001248559060200386,
+      "loss": 2.5944,
+      "step": 25970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012480046583693462,
+      "loss": 2.6511,
+      "step": 25980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012474501753036282,
+      "loss": 2.6934,
+      "step": 25990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012468956111848556,
+      "loss": 2.6911,
+      "step": 26000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012463409661946768,
+      "loss": 2.6613,
+      "step": 26010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001245786240514767,
+      "loss": 2.6827,
+      "step": 26020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012452314343268294,
+      "loss": 2.7109,
+      "step": 26030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001244676547812592,
+      "loss": 2.6181,
+      "step": 26040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012441215811538088,
+      "loss": 2.6611,
+      "step": 26050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001243566534532262,
+      "loss": 2.7017,
+      "step": 26060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012430114081297576,
+      "loss": 2.6021,
+      "step": 26070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012424562021281303,
+      "loss": 2.6923,
+      "step": 26080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012419009167092385,
+      "loss": 2.5941,
+      "step": 26090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012413455520549679,
+      "loss": 2.6065,
+      "step": 26100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012407901083472301,
+      "loss": 2.7595,
+      "step": 26110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001240234585767962,
+      "loss": 2.6459,
+      "step": 26120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012396789844991267,
+      "loss": 2.6779,
+      "step": 26130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012391233047227138,
+      "loss": 2.7104,
+      "step": 26140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001238567546620737,
+      "loss": 2.6531,
+      "step": 26150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012380117103752368,
+      "loss": 2.6872,
+      "step": 26160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012374557961682792,
+      "loss": 2.7101,
+      "step": 26170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012368998041819556,
+      "loss": 2.7079,
+      "step": 26180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001236343734598383,
+      "loss": 2.7023,
+      "step": 26190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001235787587599703,
+      "loss": 2.6828,
+      "step": 26200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001235231363368084,
+      "loss": 2.5824,
+      "step": 26210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012346750620857187,
+      "loss": 2.7219,
+      "step": 26220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012341186839348253,
+      "loss": 2.6649,
+      "step": 26230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012335622290976468,
+      "loss": 2.5396,
+      "step": 26240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001233005697756452,
+      "loss": 2.7617,
+      "step": 26250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012324490900935344,
+      "loss": 2.6061,
+      "step": 26260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012318924062912126,
+      "loss": 2.6095,
+      "step": 26270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012313356465318298,
+      "loss": 2.6376,
+      "step": 26280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012307788109977544,
+      "loss": 2.6361,
+      "step": 26290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000123022189987138,
+      "loss": 2.6824,
+      "step": 26300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012296649133351238,
+      "loss": 2.647,
+      "step": 26310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012291078515714294,
+      "loss": 2.6166,
+      "step": 26320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012285507147627632,
+      "loss": 2.7013,
+      "step": 26330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001227993503091617,
+      "loss": 2.7151,
+      "step": 26340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001227436216740508,
+      "loss": 2.687,
+      "step": 26350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012268788558919764,
+      "loss": 2.6977,
+      "step": 26360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012263214207285875,
+      "loss": 2.6501,
+      "step": 26370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012257639114329313,
+      "loss": 2.7009,
+      "step": 26380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012252063281876209,
+      "loss": 2.7077,
+      "step": 26390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012246486711752952,
+      "loss": 2.6515,
+      "step": 26400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001224090940578616,
+      "loss": 2.6857,
+      "step": 26410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012235331365802695,
+      "loss": 2.7082,
+      "step": 26420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012229752593629663,
+      "loss": 2.6801,
+      "step": 26430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012224173091094408,
+      "loss": 2.6895,
+      "step": 26440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012218592860024517,
+      "loss": 2.6896,
+      "step": 26450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000122130119022478,
+      "loss": 2.6268,
+      "step": 26460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001220743021959233,
+      "loss": 2.5484,
+      "step": 26470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012201847813886396,
+      "loss": 2.5964,
+      "step": 26480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012196264686958534,
+      "loss": 2.6641,
+      "step": 26490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012190680840637512,
+      "loss": 2.6655,
+      "step": 26500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012185096276752341,
+      "loss": 2.7396,
+      "step": 26510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001217951099713226,
+      "loss": 2.7122,
+      "step": 26520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012173925003606739,
+      "loss": 2.6615,
+      "step": 26530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001216833829800549,
+      "loss": 2.6762,
+      "step": 26540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001216275088215846,
+      "loss": 2.6669,
+      "step": 26550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012157162757895821,
+      "loss": 2.717,
+      "step": 26560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012151573927047977,
+      "loss": 2.7109,
+      "step": 26570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012145984391445571,
+      "loss": 2.6563,
+      "step": 26580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001214039415291947,
+      "loss": 2.646,
+      "step": 26590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012134803213300774,
+      "loss": 2.6838,
+      "step": 26600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012129211574420811,
+      "loss": 2.7103,
+      "step": 26610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012123619238111141,
+      "loss": 2.5674,
+      "step": 26620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001211802620620355,
+      "loss": 2.6598,
+      "step": 26630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012112432480530054,
+      "loss": 2.6328,
+      "step": 26640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012106838062922891,
+      "loss": 2.672,
+      "step": 26650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012101242955214536,
+      "loss": 2.6746,
+      "step": 26660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012095647159237677,
+      "loss": 2.668,
+      "step": 26670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012090050676825236,
+      "loss": 2.5858,
+      "step": 26680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001208445350981036,
+      "loss": 2.7594,
+      "step": 26690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012078855660026415,
+      "loss": 2.568,
+      "step": 26700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012073257129306993,
+      "loss": 2.6777,
+      "step": 26710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012067657919485913,
+      "loss": 2.6867,
+      "step": 26720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012062058032397214,
+      "loss": 2.6183,
+      "step": 26730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012056457469875153,
+      "loss": 2.6249,
+      "step": 26740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012050856233754212,
+      "loss": 2.6733,
+      "step": 26750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012045254325869091,
+      "loss": 2.6023,
+      "step": 26760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012039651748054717,
+      "loss": 2.6171,
+      "step": 26770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012034048502146225,
+      "loss": 2.6092,
+      "step": 26780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012028444589978978,
+      "loss": 2.6521,
+      "step": 26790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012022840013388554,
+      "loss": 2.7374,
+      "step": 26800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012017234774210754,
+      "loss": 2.6877,
+      "step": 26810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012011628874281586,
+      "loss": 2.6256,
+      "step": 26820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012006022315437277,
+      "loss": 2.5747,
+      "step": 26830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00012000415099514279,
+      "loss": 2.6197,
+      "step": 26840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001199480722834925,
+      "loss": 2.6537,
+      "step": 26850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001198919870377906,
+      "loss": 2.6736,
+      "step": 26860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011983589527640808,
+      "loss": 2.6435,
+      "step": 26870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001197797970177179,
+      "loss": 2.7251,
+      "step": 26880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001197236922800952,
+      "loss": 2.7093,
+      "step": 26890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011966758108191733,
+      "loss": 2.6704,
+      "step": 26900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001196114634415636,
+      "loss": 2.6002,
+      "step": 26910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011955533937741559,
+      "loss": 2.6509,
+      "step": 26920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011949920890785684,
+      "loss": 2.6208,
+      "step": 26930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001194430720512731,
+      "loss": 2.7251,
+      "step": 26940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011938692882605213,
+      "loss": 2.6256,
+      "step": 26950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011933077925058387,
+      "loss": 2.6589,
+      "step": 26960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011927462334326022,
+      "loss": 2.7361,
+      "step": 26970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011921846112247526,
+      "loss": 2.6666,
+      "step": 26980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011916229260662508,
+      "loss": 2.6351,
+      "step": 26990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011910611781410787,
+      "loss": 2.6939,
+      "step": 27000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011904993676332382,
+      "loss": 2.6589,
+      "step": 27010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011899374947267524,
+      "loss": 2.6417,
+      "step": 27020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011893755596056639,
+      "loss": 2.5428,
+      "step": 27030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011888135624540369,
+      "loss": 2.6297,
+      "step": 27040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001188251503455955,
+      "loss": 2.6431,
+      "step": 27050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001187689382795522,
+      "loss": 2.643,
+      "step": 27060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001187127200656863,
+      "loss": 2.654,
+      "step": 27070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011865649572241216,
+      "loss": 2.6291,
+      "step": 27080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011860026526814627,
+      "loss": 2.6327,
+      "step": 27090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011854402872130712,
+      "loss": 2.6706,
+      "step": 27100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011848778610031512,
+      "loss": 2.7475,
+      "step": 27110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001184315374235927,
+      "loss": 2.6357,
+      "step": 27120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011837528270956435,
+      "loss": 2.6363,
+      "step": 27130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011831902197665643,
+      "loss": 2.6192,
+      "step": 27140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011826275524329729,
+      "loss": 2.6591,
+      "step": 27150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011820648252791728,
+      "loss": 2.6873,
+      "step": 27160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011815020384894872,
+      "loss": 2.6547,
+      "step": 27170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011809391922482586,
+      "loss": 2.7182,
+      "step": 27180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011803762867398491,
+      "loss": 2.5819,
+      "step": 27190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011798133221486396,
+      "loss": 2.6678,
+      "step": 27200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011792502986590315,
+      "loss": 2.7201,
+      "step": 27210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011786872164554443,
+      "loss": 2.6216,
+      "step": 27220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011781240757223176,
+      "loss": 2.5959,
+      "step": 27230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011775608766441097,
+      "loss": 2.6722,
+      "step": 27240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011769976194052985,
+      "loss": 2.677,
+      "step": 27250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000117643430419038,
+      "loss": 2.5944,
+      "step": 27260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011758709311838702,
+      "loss": 2.6943,
+      "step": 27270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011753075005703038,
+      "loss": 2.6195,
+      "step": 27280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011747440125342341,
+      "loss": 2.6251,
+      "step": 27290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011741804672602333,
+      "loss": 2.8016,
+      "step": 27300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011736168649328918,
+      "loss": 2.67,
+      "step": 27310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011730532057368202,
+      "loss": 2.6414,
+      "step": 27320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011724894898566461,
+      "loss": 2.6078,
+      "step": 27330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011719257174770165,
+      "loss": 2.6762,
+      "step": 27340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011713618887825968,
+      "loss": 2.689,
+      "step": 27350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011707980039580711,
+      "loss": 2.6469,
+      "step": 27360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001170234063188141,
+      "loss": 2.7306,
+      "step": 27370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011696700666575273,
+      "loss": 2.7399,
+      "step": 27380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011691060145509691,
+      "loss": 2.654,
+      "step": 27390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011685419070532229,
+      "loss": 2.6275,
+      "step": 27400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011679777443490638,
+      "loss": 2.675,
+      "step": 27410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011674135266232855,
+      "loss": 2.6736,
+      "step": 27420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011668492540606987,
+      "loss": 2.6315,
+      "step": 27430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011662849268461326,
+      "loss": 2.6603,
+      "step": 27440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011657205451644349,
+      "loss": 2.6751,
+      "step": 27450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011651561092004699,
+      "loss": 2.6896,
+      "step": 27460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011645916191391206,
+      "loss": 2.6608,
+      "step": 27470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011640270751652875,
+      "loss": 2.6124,
+      "step": 27480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011634624774638882,
+      "loss": 2.6527,
+      "step": 27490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001162897826219859,
+      "loss": 2.6628,
+      "step": 27500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001162333121618153,
+      "loss": 2.5349,
+      "step": 27510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011617683638437404,
+      "loss": 2.6097,
+      "step": 27520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011612035530816099,
+      "loss": 2.6792,
+      "step": 27530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011606386895167669,
+      "loss": 2.6609,
+      "step": 27540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001160073773334234,
+      "loss": 2.6572,
+      "step": 27550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001159508804719051,
+      "loss": 2.6729,
+      "step": 27560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011589437838562756,
+      "loss": 2.6484,
+      "step": 27570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011583787109309819,
+      "loss": 2.6943,
+      "step": 27580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001157813586128261,
+      "loss": 2.6476,
+      "step": 27590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011572484096332215,
+      "loss": 2.632,
+      "step": 27600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011566831816309887,
+      "loss": 2.615,
+      "step": 27610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011561179023067041,
+      "loss": 2.7265,
+      "step": 27620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011555525718455278,
+      "loss": 2.6326,
+      "step": 27630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011549871904326346,
+      "loss": 2.5967,
+      "step": 27640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011544217582532175,
+      "loss": 2.6755,
+      "step": 27650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011538562754924854,
+      "loss": 2.5492,
+      "step": 27660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011532907423356631,
+      "loss": 2.6474,
+      "step": 27670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011527251589679938,
+      "loss": 2.6639,
+      "step": 27680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011521595255747355,
+      "loss": 2.616,
+      "step": 27690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011515938423411631,
+      "loss": 2.6541,
+      "step": 27700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011510281094525681,
+      "loss": 2.6276,
+      "step": 27710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011504623270942575,
+      "loss": 2.6274,
+      "step": 27720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011498964954515558,
+      "loss": 2.6978,
+      "step": 27730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011493306147098026,
+      "loss": 2.5674,
+      "step": 27740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011487646850543534,
+      "loss": 2.6579,
+      "step": 27750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011481987066705808,
+      "loss": 2.676,
+      "step": 27760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011476326797438724,
+      "loss": 2.7089,
+      "step": 27770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011470666044596319,
+      "loss": 2.6734,
+      "step": 27780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011465004810032793,
+      "loss": 2.6829,
+      "step": 27790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000114593430956025,
+      "loss": 2.6666,
+      "step": 27800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011453680903159947,
+      "loss": 2.5892,
+      "step": 27810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001144801823455981,
+      "loss": 2.6461,
+      "step": 27820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011442355091656909,
+      "loss": 2.6368,
+      "step": 27830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011436691476306222,
+      "loss": 2.6904,
+      "step": 27840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001143102739036289,
+      "loss": 2.6637,
+      "step": 27850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011425362835682192,
+      "loss": 2.6049,
+      "step": 27860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011419697814119578,
+      "loss": 2.7022,
+      "step": 27870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011414032327530641,
+      "loss": 2.686,
+      "step": 27880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011408366377771125,
+      "loss": 2.5847,
+      "step": 27890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001140269996669693,
+      "loss": 2.7132,
+      "step": 27900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011397033096164112,
+      "loss": 2.7284,
+      "step": 27910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011391365768028864,
+      "loss": 2.6108,
+      "step": 27920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011385697984147539,
+      "loss": 2.6125,
+      "step": 27930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011380029746376639,
+      "loss": 2.6682,
+      "step": 27940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011374361056572813,
+      "loss": 2.681,
+      "step": 27950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011368691916592853,
+      "loss": 2.5618,
+      "step": 27960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011363022328293707,
+      "loss": 2.607,
+      "step": 27970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011357352293532463,
+      "loss": 2.7104,
+      "step": 27980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011351681814166358,
+      "loss": 2.6076,
+      "step": 27990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011346010892052779,
+      "loss": 2.6167,
+      "step": 28000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011340339529049249,
+      "loss": 2.5345,
+      "step": 28010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011334667727013442,
+      "loss": 2.6842,
+      "step": 28020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011328995487803175,
+      "loss": 2.656,
+      "step": 28030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011323322813276404,
+      "loss": 2.5403,
+      "step": 28040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011317649705291238,
+      "loss": 2.6755,
+      "step": 28050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001131197616570591,
+      "loss": 2.5629,
+      "step": 28060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011306302196378813,
+      "loss": 2.7436,
+      "step": 28070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011300627799168468,
+      "loss": 2.6227,
+      "step": 28080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011294952975933549,
+      "loss": 2.5972,
+      "step": 28090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011289277728532852,
+      "loss": 2.6842,
+      "step": 28100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011283602058825324,
+      "loss": 2.6627,
+      "step": 28110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011277925968670053,
+      "loss": 2.5849,
+      "step": 28120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011272249459926256,
+      "loss": 2.5778,
+      "step": 28130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001126657253445329,
+      "loss": 2.6457,
+      "step": 28140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011260895194110647,
+      "loss": 2.6548,
+      "step": 28150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011255217440757964,
+      "loss": 2.6051,
+      "step": 28160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011249539276255,
+      "loss": 2.6459,
+      "step": 28170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011243860702461658,
+      "loss": 2.7059,
+      "step": 28180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011238181721237968,
+      "loss": 2.6301,
+      "step": 28190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011232502334444104,
+      "loss": 2.6867,
+      "step": 28200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011226822543940363,
+      "loss": 2.651,
+      "step": 28210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011221142351587175,
+      "loss": 2.6581,
+      "step": 28220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001121546175924511,
+      "loss": 2.5471,
+      "step": 28230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001120978076877486,
+      "loss": 2.6299,
+      "step": 28240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011204099382037249,
+      "loss": 2.703,
+      "step": 28250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011198417600893234,
+      "loss": 2.6221,
+      "step": 28260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011192735427203899,
+      "loss": 2.7261,
+      "step": 28270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011187052862830462,
+      "loss": 2.6168,
+      "step": 28280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011181369909634258,
+      "loss": 2.6417,
+      "step": 28290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011175686569476757,
+      "loss": 2.5682,
+      "step": 28300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011170002844219558,
+      "loss": 2.7134,
+      "step": 28310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011164318735724379,
+      "loss": 2.546,
+      "step": 28320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011158634245853066,
+      "loss": 2.6278,
+      "step": 28330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011152949376467594,
+      "loss": 2.5775,
+      "step": 28340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011147264129430057,
+      "loss": 2.6704,
+      "step": 28350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011141578506602672,
+      "loss": 2.6398,
+      "step": 28360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001113589250984779,
+      "loss": 2.6278,
+      "step": 28370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001113020614102787,
+      "loss": 2.5676,
+      "step": 28380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011124519402005504,
+      "loss": 2.6798,
+      "step": 28390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011118832294643398,
+      "loss": 2.6244,
+      "step": 28400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011113144820804378,
+      "loss": 2.626,
+      "step": 28410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011107456982351399,
+      "loss": 2.6561,
+      "step": 28420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011101768781147528,
+      "loss": 2.6964,
+      "step": 28430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011096080219055948,
+      "loss": 2.6686,
+      "step": 28440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001109039129793997,
+      "loss": 2.6406,
+      "step": 28450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011084702019663018,
+      "loss": 2.6336,
+      "step": 28460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001107901238608863,
+      "loss": 2.6834,
+      "step": 28470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011073322399080461,
+      "loss": 2.6357,
+      "step": 28480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011067632060502289,
+      "loss": 2.5674,
+      "step": 28490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011061941372217996,
+      "loss": 2.6129,
+      "step": 28500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011056250336091585,
+      "loss": 2.5922,
+      "step": 28510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011050558953987178,
+      "loss": 2.6232,
+      "step": 28520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011044867227768998,
+      "loss": 2.6263,
+      "step": 28530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011039175159301387,
+      "loss": 2.6401,
+      "step": 28540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011033482750448802,
+      "loss": 2.6823,
+      "step": 28550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011027790003075811,
+      "loss": 2.7083,
+      "step": 28560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011022096919047086,
+      "loss": 2.5843,
+      "step": 28570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011016403500227418,
+      "loss": 2.6861,
+      "step": 28580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011010709748481698,
+      "loss": 2.664,
+      "step": 28590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00011005015665674939,
+      "loss": 2.7157,
+      "step": 28600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010999321253672248,
+      "loss": 2.6473,
+      "step": 28610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001099362651433885,
+      "loss": 2.6262,
+      "step": 28620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010987931449540075,
+      "loss": 2.6751,
+      "step": 28630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010982236061141358,
+      "loss": 2.6807,
+      "step": 28640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010976540351008245,
+      "loss": 2.7054,
+      "step": 28650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010970844321006373,
+      "loss": 2.6956,
+      "step": 28660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010965147973001501,
+      "loss": 2.6284,
+      "step": 28670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010959451308859488,
+      "loss": 2.6781,
+      "step": 28680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010953754330446287,
+      "loss": 2.614,
+      "step": 28690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001094805703962796,
+      "loss": 2.619,
+      "step": 28700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010942359438270678,
+      "loss": 2.6963,
+      "step": 28710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010936661528240704,
+      "loss": 2.6562,
+      "step": 28720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010930963311404405,
+      "loss": 2.5649,
+      "step": 28730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010925264789628251,
+      "loss": 2.6772,
+      "step": 28740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010919565964778811,
+      "loss": 2.6073,
+      "step": 28750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010913866838722751,
+      "loss": 2.6576,
+      "step": 28760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010908167413326833,
+      "loss": 2.6806,
+      "step": 28770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001090246769045793,
+      "loss": 2.7294,
+      "step": 28780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010896767671982997,
+      "loss": 2.6458,
+      "step": 28790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010891067359769094,
+      "loss": 2.7148,
+      "step": 28800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010885366755683373,
+      "loss": 2.6724,
+      "step": 28810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010879665861593084,
+      "loss": 2.6418,
+      "step": 28820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001087396467936558,
+      "loss": 2.6238,
+      "step": 28830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010868263210868294,
+      "loss": 2.7061,
+      "step": 28840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010862561457968761,
+      "loss": 2.6605,
+      "step": 28850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010856859422534606,
+      "loss": 2.5803,
+      "step": 28860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010851157106433549,
+      "loss": 2.6371,
+      "step": 28870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010845454511533399,
+      "loss": 2.6414,
+      "step": 28880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010839751639702064,
+      "loss": 2.616,
+      "step": 28890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010834048492807531,
+      "loss": 2.5851,
+      "step": 28900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010828345072717885,
+      "loss": 2.5698,
+      "step": 28910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010822641381301297,
+      "loss": 2.6533,
+      "step": 28920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010816937420426033,
+      "loss": 2.5694,
+      "step": 28930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010811233191960439,
+      "loss": 2.6394,
+      "step": 28940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010805528697772952,
+      "loss": 2.669,
+      "step": 28950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010799823939732098,
+      "loss": 2.5596,
+      "step": 28960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010794118919706489,
+      "loss": 2.6297,
+      "step": 28970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010788413639564818,
+      "loss": 2.6304,
+      "step": 28980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010782708101175869,
+      "loss": 2.5583,
+      "step": 28990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010777002306408506,
+      "loss": 2.5967,
+      "step": 29000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010771296257131684,
+      "loss": 2.6573,
+      "step": 29010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010765589955214435,
+      "loss": 2.7197,
+      "step": 29020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010759883402525872,
+      "loss": 2.6176,
+      "step": 29030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010754176600935197,
+      "loss": 2.5794,
+      "step": 29040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010748469552311693,
+      "loss": 2.5773,
+      "step": 29050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010742762258524712,
+      "loss": 2.7293,
+      "step": 29060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010737054721443705,
+      "loss": 2.5993,
+      "step": 29070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001073134694293819,
+      "loss": 2.6393,
+      "step": 29080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010725638924877761,
+      "loss": 2.5803,
+      "step": 29090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010719930669132105,
+      "loss": 2.6177,
+      "step": 29100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010714222177570976,
+      "loss": 2.6393,
+      "step": 29110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010708513452064206,
+      "loss": 2.584,
+      "step": 29120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010702804494481708,
+      "loss": 2.6243,
+      "step": 29130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010697095306693463,
+      "loss": 2.5951,
+      "step": 29140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010691385890569541,
+      "loss": 2.6497,
+      "step": 29150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010685676247980071,
+      "loss": 2.6002,
+      "step": 29160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010679966380795268,
+      "loss": 2.6689,
+      "step": 29170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010674256290885413,
+      "loss": 2.6378,
+      "step": 29180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010668545980120868,
+      "loss": 2.7347,
+      "step": 29190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010662835450372059,
+      "loss": 2.6138,
+      "step": 29200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010657124703509489,
+      "loss": 2.6638,
+      "step": 29210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010651413741403728,
+      "loss": 2.6557,
+      "step": 29220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010645702565925424,
+      "loss": 2.6504,
+      "step": 29230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010639991178945285,
+      "loss": 2.6452,
+      "step": 29240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010634279582334095,
+      "loss": 2.6004,
+      "step": 29250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010628567777962704,
+      "loss": 2.6881,
+      "step": 29260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010622855767702029,
+      "loss": 2.6168,
+      "step": 29270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010617143553423061,
+      "loss": 2.6217,
+      "step": 29280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001061143113699685,
+      "loss": 2.6879,
+      "step": 29290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010605718520294519,
+      "loss": 2.5503,
+      "step": 29300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010600005705187248,
+      "loss": 2.5762,
+      "step": 29310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010594292693546286,
+      "loss": 2.6411,
+      "step": 29320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010588579487242953,
+      "loss": 2.5723,
+      "step": 29330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010582866088148626,
+      "loss": 2.6867,
+      "step": 29340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010577152498134743,
+      "loss": 2.6794,
+      "step": 29350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010571438719072806,
+      "loss": 2.6249,
+      "step": 29360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010565724752834385,
+      "loss": 2.6762,
+      "step": 29370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010560010601291109,
+      "loss": 2.5723,
+      "step": 29380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001055429626631466,
+      "loss": 2.6064,
+      "step": 29390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001054858174977679,
+      "loss": 2.6359,
+      "step": 29400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010542867053549301,
+      "loss": 2.6553,
+      "step": 29410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010537152179504067,
+      "loss": 2.6693,
+      "step": 29420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010531437129513005,
+      "loss": 2.6557,
+      "step": 29430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000105257219054481,
+      "loss": 2.628,
+      "step": 29440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001052000650918139,
+      "loss": 2.576,
+      "step": 29450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010514290942584972,
+      "loss": 2.6566,
+      "step": 29460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010508575207530995,
+      "loss": 2.6053,
+      "step": 29470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010502859305891668,
+      "loss": 2.5859,
+      "step": 29480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010497143239539253,
+      "loss": 2.5459,
+      "step": 29490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001049142701034606,
+      "loss": 2.6295,
+      "step": 29500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010485710620184459,
+      "loss": 2.588,
+      "step": 29510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010479994070926873,
+      "loss": 2.5839,
+      "step": 29520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010474277364445772,
+      "loss": 2.6405,
+      "step": 29530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010468560502613682,
+      "loss": 2.6253,
+      "step": 29540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010462843487303178,
+      "loss": 2.6633,
+      "step": 29550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001045712632038689,
+      "loss": 2.6782,
+      "step": 29560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010451409003737486,
+      "loss": 2.6544,
+      "step": 29570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010445691539227696,
+      "loss": 2.6215,
+      "step": 29580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010439973928730289,
+      "loss": 2.6364,
+      "step": 29590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010434256174118091,
+      "loss": 2.5883,
+      "step": 29600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010428538277263962,
+      "loss": 2.5938,
+      "step": 29610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010422820240040825,
+      "loss": 2.5529,
+      "step": 29620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010417102064321633,
+      "loss": 2.6239,
+      "step": 29630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010411383751979395,
+      "loss": 2.7011,
+      "step": 29640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010405665304887165,
+      "loss": 2.6807,
+      "step": 29650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010399946724918031,
+      "loss": 2.6333,
+      "step": 29660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010394228013945135,
+      "loss": 2.6995,
+      "step": 29670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010388509173841661,
+      "loss": 2.546,
+      "step": 29680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010382790206480823,
+      "loss": 2.607,
+      "step": 29690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010377071113735896,
+      "loss": 2.6008,
+      "step": 29700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010371351897480184,
+      "loss": 2.5819,
+      "step": 29710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010365632559587029,
+      "loss": 2.6544,
+      "step": 29720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010359913101929819,
+      "loss": 2.5333,
+      "step": 29730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010354193526381986,
+      "loss": 2.6862,
+      "step": 29740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010348473834816985,
+      "loss": 2.6102,
+      "step": 29750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010342754029108323,
+      "loss": 2.5955,
+      "step": 29760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010337034111129543,
+      "loss": 2.635,
+      "step": 29770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010331314082754218,
+      "loss": 2.5341,
+      "step": 29780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010325593945855958,
+      "loss": 2.6097,
+      "step": 29790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010319873702308413,
+      "loss": 2.5733,
+      "step": 29800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010314153353985267,
+      "loss": 2.559,
+      "step": 29810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010308432902760238,
+      "loss": 2.657,
+      "step": 29820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010302712350507078,
+      "loss": 2.6394,
+      "step": 29830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010296991699099568,
+      "loss": 2.6689,
+      "step": 29840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010291270950411527,
+      "loss": 2.6419,
+      "step": 29850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010285550106316804,
+      "loss": 2.6378,
+      "step": 29860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010279829168689274,
+      "loss": 2.5356,
+      "step": 29870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010274108139402856,
+      "loss": 2.5765,
+      "step": 29880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010268387020331485,
+      "loss": 2.5174,
+      "step": 29890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010262665813349131,
+      "loss": 2.6472,
+      "step": 29900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010256944520329789,
+      "loss": 2.6521,
+      "step": 29910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010251223143147495,
+      "loss": 2.6345,
+      "step": 29920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010245501683676298,
+      "loss": 2.6797,
+      "step": 29930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001023978014379028,
+      "loss": 2.5976,
+      "step": 29940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010234058525363548,
+      "loss": 2.6225,
+      "step": 29950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010228336830270238,
+      "loss": 2.687,
+      "step": 29960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010222615060384506,
+      "loss": 2.7049,
+      "step": 29970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010216893217580534,
+      "loss": 2.576,
+      "step": 29980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010211171303732534,
+      "loss": 2.6424,
+      "step": 29990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010205449320714733,
+      "loss": 2.629,
+      "step": 30000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001019972727040138,
+      "loss": 2.5718,
+      "step": 30010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010194005154666755,
+      "loss": 2.6925,
+      "step": 30020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010188282975385158,
+      "loss": 2.6104,
+      "step": 30030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010182560734430897,
+      "loss": 2.6072,
+      "step": 30040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010176838433678318,
+      "loss": 2.536,
+      "step": 30050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010171116075001767,
+      "loss": 2.6107,
+      "step": 30060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010165393660275633,
+      "loss": 2.5934,
+      "step": 30070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010160243440648898,
+      "loss": 2.515,
+      "step": 30080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010154520924592512,
+      "loss": 2.6037,
+      "step": 30090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010148798357922332,
+      "loss": 2.6594,
+      "step": 30100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010143075742512797,
+      "loss": 2.6848,
+      "step": 30110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010137353080238368,
+      "loss": 2.5906,
+      "step": 30120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010131630372973521,
+      "loss": 2.6901,
+      "step": 30130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010125907622592745,
+      "loss": 2.7022,
+      "step": 30140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010120184830970548,
+      "loss": 2.6331,
+      "step": 30150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010114461999981442,
+      "loss": 2.5222,
+      "step": 30160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010108739131499965,
+      "loss": 2.6212,
+      "step": 30170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010103016227400654,
+      "loss": 2.5984,
+      "step": 30180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010097293289558066,
+      "loss": 2.72,
+      "step": 30190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010091570319846764,
+      "loss": 2.5951,
+      "step": 30200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010085847320141328,
+      "loss": 2.5925,
+      "step": 30210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010080124292316342,
+      "loss": 2.6123,
+      "step": 30220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010074401238246399,
+      "loss": 2.6468,
+      "step": 30230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010068678159806104,
+      "loss": 2.724,
+      "step": 30240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010062955058870072,
+      "loss": 2.6269,
+      "step": 30250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001005723193731292,
+      "loss": 2.6155,
+      "step": 30260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010051508797009273,
+      "loss": 2.6378,
+      "step": 30270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010045785639833763,
+      "loss": 2.7504,
+      "step": 30280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010040062467661029,
+      "loss": 2.5738,
+      "step": 30290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001003433928236571,
+      "loss": 2.5874,
+      "step": 30300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010028616085822455,
+      "loss": 2.6241,
+      "step": 30310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010022892879905914,
+      "loss": 2.6537,
+      "step": 30320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00010017169666490742,
+      "loss": 2.6395,
+      "step": 30330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00010011446447451596,
+      "loss": 2.6443,
+      "step": 30340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00010005723224663126,
+      "loss": 2.6831,
+      "step": 30350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001,
+      "loss": 2.7064,
+      "step": 30360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.994276775336875e-05,
+      "loss": 2.6256,
+      "step": 30370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.988553552548408e-05,
+      "loss": 2.6226,
+      "step": 30380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.982830333509259e-05,
+      "loss": 2.5506,
+      "step": 30390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.977107120094089e-05,
+      "loss": 2.5901,
+      "step": 30400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.971383914177549e-05,
+      "loss": 2.614,
+      "step": 30410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.965660717634295e-05,
+      "loss": 2.6432,
+      "step": 30420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.959937532338976e-05,
+      "loss": 2.6487,
+      "step": 30430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.954214360166242e-05,
+      "loss": 2.6833,
+      "step": 30440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.948491202990731e-05,
+      "loss": 2.6591,
+      "step": 30450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.942768062687079e-05,
+      "loss": 2.5972,
+      "step": 30460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.937044941129928e-05,
+      "loss": 2.5943,
+      "step": 30470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.931321840193895e-05,
+      "loss": 2.6317,
+      "step": 30480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.925598761753604e-05,
+      "loss": 2.6377,
+      "step": 30490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.91987570768366e-05,
+      "loss": 2.6294,
+      "step": 30500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.914152679858675e-05,
+      "loss": 2.5864,
+      "step": 30510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.908429680153238e-05,
+      "loss": 2.6561,
+      "step": 30520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.902706710441935e-05,
+      "loss": 2.6002,
+      "step": 30530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.896983772599348e-05,
+      "loss": 2.7288,
+      "step": 30540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.891260868500037e-05,
+      "loss": 2.6492,
+      "step": 30550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.885538000018559e-05,
+      "loss": 2.6029,
+      "step": 30560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.879815169029453e-05,
+      "loss": 2.6141,
+      "step": 30570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.874092377407257e-05,
+      "loss": 2.6267,
+      "step": 30580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.868369627026484e-05,
+      "loss": 2.6424,
+      "step": 30590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.862646919761636e-05,
+      "loss": 2.5806,
+      "step": 30600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.856924257487208e-05,
+      "loss": 2.6168,
+      "step": 30610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.851201642077673e-05,
+      "loss": 2.5249,
+      "step": 30620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.84547907540749e-05,
+      "loss": 2.6657,
+      "step": 30630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.839756559351103e-05,
+      "loss": 2.605,
+      "step": 30640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.834034095782942e-05,
+      "loss": 2.5851,
+      "step": 30650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.828311686577419e-05,
+      "loss": 2.6036,
+      "step": 30660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.822589333608923e-05,
+      "loss": 2.6241,
+      "step": 30670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.816867038751826e-05,
+      "loss": 2.6053,
+      "step": 30680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.811144803880489e-05,
+      "loss": 2.6454,
+      "step": 30690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.805422630869243e-05,
+      "loss": 2.5805,
+      "step": 30700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.799700521592404e-05,
+      "loss": 2.7059,
+      "step": 30710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.793978477924266e-05,
+      "loss": 2.5708,
+      "step": 30720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.788256501739104e-05,
+      "loss": 2.5482,
+      "step": 30730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.782534594911166e-05,
+      "loss": 2.6311,
+      "step": 30740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.77681275931468e-05,
+      "loss": 2.6146,
+      "step": 30750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.771090996823852e-05,
+      "loss": 2.6053,
+      "step": 30760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.765369309312863e-05,
+      "loss": 2.6362,
+      "step": 30770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.759647698655865e-05,
+      "loss": 2.6108,
+      "step": 30780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.753926166726997e-05,
+      "loss": 2.5786,
+      "step": 30790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.748204715400356e-05,
+      "loss": 2.6747,
+      "step": 30800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.742483346550026e-05,
+      "loss": 2.6162,
+      "step": 30810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.736762062050053e-05,
+      "loss": 2.7522,
+      "step": 30820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.731040863774464e-05,
+      "loss": 2.5762,
+      "step": 30830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.72531975359726e-05,
+      "loss": 2.634,
+      "step": 30840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.719598733392403e-05,
+      "loss": 2.6933,
+      "step": 30850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.71387780503383e-05,
+      "loss": 2.5619,
+      "step": 30860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.70815697039545e-05,
+      "loss": 2.631,
+      "step": 30870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.702436231351141e-05,
+      "loss": 2.5997,
+      "step": 30880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.696715589774745e-05,
+      "loss": 2.6879,
+      "step": 30890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.690995047540077e-05,
+      "loss": 2.5759,
+      "step": 30900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.685274606520919e-05,
+      "loss": 2.6576,
+      "step": 30910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.679554268591018e-05,
+      "loss": 2.6339,
+      "step": 30920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.673834035624085e-05,
+      "loss": 2.6064,
+      "step": 30930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.668113909493807e-05,
+      "loss": 2.6188,
+      "step": 30940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.662393892073824e-05,
+      "loss": 2.6869,
+      "step": 30950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.656673985237743e-05,
+      "loss": 2.6626,
+      "step": 30960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.65095419085914e-05,
+      "loss": 2.5401,
+      "step": 30970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.64523451081155e-05,
+      "loss": 2.6519,
+      "step": 30980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.639514946968473e-05,
+      "loss": 2.5184,
+      "step": 30990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.633795501203366e-05,
+      "loss": 2.6743,
+      "step": 31000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.62807617538965e-05,
+      "loss": 2.5904,
+      "step": 31010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.622356971400716e-05,
+      "loss": 2.6042,
+      "step": 31020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.6166378911099e-05,
+      "loss": 2.6525,
+      "step": 31030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.610918936390505e-05,
+      "loss": 2.643,
+      "step": 31040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.605200109115792e-05,
+      "loss": 2.6554,
+      "step": 31050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.59948141115898e-05,
+      "loss": 2.6069,
+      "step": 31060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.593762844393249e-05,
+      "loss": 2.6292,
+      "step": 31070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.588044410691725e-05,
+      "loss": 2.6482,
+      "step": 31080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.582326111927508e-05,
+      "loss": 2.6683,
+      "step": 31090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.576607949973639e-05,
+      "loss": 2.5755,
+      "step": 31100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.570889926703119e-05,
+      "loss": 2.62,
+      "step": 31110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.565172043988903e-05,
+      "loss": 2.6012,
+      "step": 31120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.559454303703905e-05,
+      "loss": 2.6773,
+      "step": 31130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.553736707720985e-05,
+      "loss": 2.5745,
+      "step": 31140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.548019257912958e-05,
+      "loss": 2.6531,
+      "step": 31150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.542301956152597e-05,
+      "loss": 2.6368,
+      "step": 31160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.536584804312615e-05,
+      "loss": 2.6892,
+      "step": 31170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.530867804265685e-05,
+      "loss": 2.5924,
+      "step": 31180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.525150957884429e-05,
+      "loss": 2.5587,
+      "step": 31190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.519434267041416e-05,
+      "loss": 2.6102,
+      "step": 31200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.513717733609167e-05,
+      "loss": 2.623,
+      "step": 31210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.508001359460151e-05,
+      "loss": 2.5667,
+      "step": 31220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.502285146466779e-05,
+      "loss": 2.5959,
+      "step": 31230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.496569096501418e-05,
+      "loss": 2.6451,
+      "step": 31240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.490853211436375e-05,
+      "loss": 2.5376,
+      "step": 31250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.485137493143906e-05,
+      "loss": 2.6282,
+      "step": 31260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.479421943496215e-05,
+      "loss": 2.552,
+      "step": 31270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.473706564365442e-05,
+      "loss": 2.5152,
+      "step": 31280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.467991357623683e-05,
+      "loss": 2.5714,
+      "step": 31290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.462276325142963e-05,
+      "loss": 2.7117,
+      "step": 31300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.456561468795268e-05,
+      "loss": 2.6234,
+      "step": 31310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.45084679045251e-05,
+      "loss": 2.6562,
+      "step": 31320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.44513229198655e-05,
+      "loss": 2.5218,
+      "step": 31330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.43941797526919e-05,
+      "loss": 2.5663,
+      "step": 31340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.433703842172175e-05,
+      "loss": 2.5953,
+      "step": 31350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.427989894567182e-05,
+      "loss": 2.6284,
+      "step": 31360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.422276134325829e-05,
+      "loss": 2.6646,
+      "step": 31370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.416562563319682e-05,
+      "loss": 2.5863,
+      "step": 31380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.410849183420241e-05,
+      "loss": 2.687,
+      "step": 31390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.405135996498935e-05,
+      "loss": 2.6162,
+      "step": 31400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.399423004427135e-05,
+      "loss": 2.6288,
+      "step": 31410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.393710209076152e-05,
+      "loss": 2.6449,
+      "step": 31420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.38799761231723e-05,
+      "loss": 2.5917,
+      "step": 31430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.382285216021546e-05,
+      "loss": 2.5605,
+      "step": 31440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.376573022060209e-05,
+      "loss": 2.6231,
+      "step": 31450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.370861032304272e-05,
+      "loss": 2.5735,
+      "step": 31460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.365149248624711e-05,
+      "loss": 2.6417,
+      "step": 31470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.359437672892437e-05,
+      "loss": 2.7132,
+      "step": 31480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.3537263069783e-05,
+      "loss": 2.6707,
+      "step": 31490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.348015152753067e-05,
+      "loss": 2.6174,
+      "step": 31500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.34230421208745e-05,
+      "loss": 2.5828,
+      "step": 31510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.336593486852078e-05,
+      "loss": 2.6233,
+      "step": 31520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.330882978917526e-05,
+      "loss": 2.6064,
+      "step": 31530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.325172690154283e-05,
+      "loss": 2.6624,
+      "step": 31540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.319462622432767e-05,
+      "loss": 2.6192,
+      "step": 31550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.313752777623332e-05,
+      "loss": 2.6573,
+      "step": 31560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.30804315759626e-05,
+      "loss": 2.6556,
+      "step": 31570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.30233376422175e-05,
+      "loss": 2.5889,
+      "step": 31580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.296624599369932e-05,
+      "loss": 2.5177,
+      "step": 31590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.290915664910855e-05,
+      "loss": 2.6447,
+      "step": 31600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.285206962714506e-05,
+      "loss": 2.5352,
+      "step": 31610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.279498494650782e-05,
+      "loss": 2.6811,
+      "step": 31620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.27379026258951e-05,
+      "loss": 2.6113,
+      "step": 31630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.268082268400442e-05,
+      "loss": 2.6134,
+      "step": 31640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.262374513953246e-05,
+      "loss": 2.6574,
+      "step": 31650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.256667001117514e-05,
+      "loss": 2.6808,
+      "step": 31660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.250959731762759e-05,
+      "loss": 2.6889,
+      "step": 31670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.245252707758416e-05,
+      "loss": 2.6949,
+      "step": 31680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.239545930973838e-05,
+      "loss": 2.5865,
+      "step": 31690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.233839403278293e-05,
+      "loss": 2.6622,
+      "step": 31700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.228133126540978e-05,
+      "loss": 2.6317,
+      "step": 31710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.222427102630997e-05,
+      "loss": 2.6546,
+      "step": 31720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.216721333417375e-05,
+      "loss": 2.6371,
+      "step": 31730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.211015820769054e-05,
+      "loss": 2.5843,
+      "step": 31740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.205310566554895e-05,
+      "loss": 2.6576,
+      "step": 31750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.199605572643671e-05,
+      "loss": 2.599,
+      "step": 31760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.193900840904068e-05,
+      "loss": 2.5656,
+      "step": 31770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.188196373204687e-05,
+      "loss": 2.6718,
+      "step": 31780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.182492171414048e-05,
+      "loss": 2.6519,
+      "step": 31790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.176788237400578e-05,
+      "loss": 2.5336,
+      "step": 31800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.171084573032614e-05,
+      "loss": 2.6469,
+      "step": 31810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.165381180178415e-05,
+      "loss": 2.5766,
+      "step": 31820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.15967806070614e-05,
+      "loss": 2.7171,
+      "step": 31830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.153975216483864e-05,
+      "loss": 2.5885,
+      "step": 31840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.14827264937957e-05,
+      "loss": 2.6437,
+      "step": 31850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.142570361261153e-05,
+      "loss": 2.6123,
+      "step": 31860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.136868353996416e-05,
+      "loss": 2.6152,
+      "step": 31870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.131166629453062e-05,
+      "loss": 2.5959,
+      "step": 31880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.125465189498719e-05,
+      "loss": 2.6702,
+      "step": 31890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.119764036000901e-05,
+      "loss": 2.65,
+      "step": 31900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.114063170827045e-05,
+      "loss": 2.6002,
+      "step": 31910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.10836259584448e-05,
+      "loss": 2.6223,
+      "step": 31920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.10266231292045e-05,
+      "loss": 2.6062,
+      "step": 31930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.096962323922104e-05,
+      "loss": 2.594,
+      "step": 31940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.091262630716488e-05,
+      "loss": 2.6731,
+      "step": 31950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.085563235170549e-05,
+      "loss": 2.5561,
+      "step": 31960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.079864139151147e-05,
+      "loss": 2.6397,
+      "step": 31970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.074165344525037e-05,
+      "loss": 2.5176,
+      "step": 31980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.068466853158877e-05,
+      "loss": 2.6687,
+      "step": 31990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.062768666919221e-05,
+      "loss": 2.579,
+      "step": 32000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.057070787672532e-05,
+      "loss": 2.6778,
+      "step": 32010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.051373217285165e-05,
+      "loss": 2.6837,
+      "step": 32020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.045675957623375e-05,
+      "loss": 2.5402,
+      "step": 32030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.039979010553322e-05,
+      "loss": 2.6572,
+      "step": 32040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.034282377941054e-05,
+      "loss": 2.5599,
+      "step": 32050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028586061652524e-05,
+      "loss": 2.6042,
+      "step": 32060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.02289006355357e-05,
+      "loss": 2.5916,
+      "step": 32070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.017194385509941e-05,
+      "loss": 2.5841,
+      "step": 32080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.011499029387274e-05,
+      "loss": 2.6411,
+      "step": 32090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.005803997051092e-05,
+      "loss": 2.4999,
+      "step": 32100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.000109290366825e-05,
+      "loss": 2.6101,
+      "step": 32110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.994414911199795e-05,
+      "loss": 2.6506,
+      "step": 32120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.98872086141521e-05,
+      "loss": 2.5075,
+      "step": 32130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.983027142878175e-05,
+      "loss": 2.643,
+      "step": 32140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.97733375745368e-05,
+      "loss": 2.6404,
+      "step": 32150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.971640707006615e-05,
+      "loss": 2.5382,
+      "step": 32160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.965947993401757e-05,
+      "loss": 2.6133,
+      "step": 32170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.960255618503764e-05,
+      "loss": 2.6604,
+      "step": 32180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.9545635841772e-05,
+      "loss": 2.616,
+      "step": 32190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.948871892286504e-05,
+      "loss": 2.603,
+      "step": 32200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.943749663908416e-05,
+      "loss": 2.6277,
+      "step": 32210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.938058627782008e-05,
+      "loss": 2.6372,
+      "step": 32220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.932367939497712e-05,
+      "loss": 2.6144,
+      "step": 32230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.926677600919539e-05,
+      "loss": 2.6459,
+      "step": 32240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.920987613911373e-05,
+      "loss": 2.6651,
+      "step": 32250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.915297980336982e-05,
+      "loss": 2.5374,
+      "step": 32260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.909608702060031e-05,
+      "loss": 2.633,
+      "step": 32270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.903919780944054e-05,
+      "loss": 2.5547,
+      "step": 32280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.898231218852476e-05,
+      "loss": 2.6508,
+      "step": 32290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.892543017648603e-05,
+      "loss": 2.6811,
+      "step": 32300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.886855179195625e-05,
+      "loss": 2.5617,
+      "step": 32310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.881167705356606e-05,
+      "loss": 2.578,
+      "step": 32320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.875480597994499e-05,
+      "loss": 2.5723,
+      "step": 32330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.869793858972132e-05,
+      "loss": 2.622,
+      "step": 32340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.864107490152214e-05,
+      "loss": 2.6354,
+      "step": 32350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.858421493397329e-05,
+      "loss": 2.569,
+      "step": 32360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.852735870569948e-05,
+      "loss": 2.5618,
+      "step": 32370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.847050623532411e-05,
+      "loss": 2.5059,
+      "step": 32380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.841365754146938e-05,
+      "loss": 2.6164,
+      "step": 32390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.835681264275626e-05,
+      "loss": 2.5865,
+      "step": 32400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.829997155780447e-05,
+      "loss": 2.5607,
+      "step": 32410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.824313430523242e-05,
+      "loss": 2.6653,
+      "step": 32420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.818630090365743e-05,
+      "loss": 2.4724,
+      "step": 32430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.812947137169539e-05,
+      "loss": 2.6331,
+      "step": 32440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.8072645727961e-05,
+      "loss": 2.642,
+      "step": 32450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.801582399106767e-05,
+      "loss": 2.5862,
+      "step": 32460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.795900617962752e-05,
+      "loss": 2.5325,
+      "step": 32470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.790219231225144e-05,
+      "loss": 2.5955,
+      "step": 32480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.784538240754893e-05,
+      "loss": 2.6089,
+      "step": 32490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.778857648412828e-05,
+      "loss": 2.6552,
+      "step": 32500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.77317745605964e-05,
+      "loss": 2.5352,
+      "step": 32510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.767497665555899e-05,
+      "loss": 2.5628,
+      "step": 32520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.761818278762034e-05,
+      "loss": 2.5662,
+      "step": 32530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.756139297538345e-05,
+      "loss": 2.6669,
+      "step": 32540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.750460723745003e-05,
+      "loss": 2.4996,
+      "step": 32550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.74478255924204e-05,
+      "loss": 2.5825,
+      "step": 32560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.739104805889354e-05,
+      "loss": 2.6205,
+      "step": 32570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.733427465546714e-05,
+      "loss": 2.6435,
+      "step": 32580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.727750540073748e-05,
+      "loss": 2.5927,
+      "step": 32590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.722074031329947e-05,
+      "loss": 2.6716,
+      "step": 32600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.716397941174674e-05,
+      "loss": 2.6018,
+      "step": 32610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.710722271467147e-05,
+      "loss": 2.6438,
+      "step": 32620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705047024066452e-05,
+      "loss": 2.5614,
+      "step": 32630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699372200831532e-05,
+      "loss": 2.6428,
+      "step": 32640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.69369780362119e-05,
+      "loss": 2.6942,
+      "step": 32650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.68802383429409e-05,
+      "loss": 2.5096,
+      "step": 32660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.682350294708766e-05,
+      "loss": 2.57,
+      "step": 32670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.676677186723597e-05,
+      "loss": 2.5938,
+      "step": 32680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.671004512196827e-05,
+      "loss": 2.6658,
+      "step": 32690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.665332272986559e-05,
+      "loss": 2.6253,
+      "step": 32700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.659660470950755e-05,
+      "loss": 2.5927,
+      "step": 32710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.653989107947223e-05,
+      "loss": 2.6732,
+      "step": 32720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.648318185833643e-05,
+      "loss": 2.571,
+      "step": 32730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.642647706467542e-05,
+      "loss": 2.6967,
+      "step": 32740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.6369776717063e-05,
+      "loss": 2.6214,
+      "step": 32750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.63130808340715e-05,
+      "loss": 2.542,
+      "step": 32760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.625638943427191e-05,
+      "loss": 2.5753,
+      "step": 32770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.61997025362336e-05,
+      "loss": 2.5876,
+      "step": 32780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.61430201585246e-05,
+      "loss": 2.4853,
+      "step": 32790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.608634231971137e-05,
+      "loss": 2.5875,
+      "step": 32800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.602966903835889e-05,
+      "loss": 2.671,
+      "step": 32810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.597300033303071e-05,
+      "loss": 2.6479,
+      "step": 32820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.591633622228879e-05,
+      "loss": 2.6245,
+      "step": 32830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.585967672469362e-05,
+      "loss": 2.5774,
+      "step": 32840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.580302185880424e-05,
+      "loss": 2.6292,
+      "step": 32850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.574637164317811e-05,
+      "loss": 2.6067,
+      "step": 32860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.568972609637112e-05,
+      "loss": 2.6024,
+      "step": 32870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.563308523693779e-05,
+      "loss": 2.6396,
+      "step": 32880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.557644908343095e-05,
+      "loss": 2.5759,
+      "step": 32890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.551981765440194e-05,
+      "loss": 2.5823,
+      "step": 32900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.546319096840054e-05,
+      "loss": 2.5526,
+      "step": 32910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.540656904397506e-05,
+      "loss": 2.6044,
+      "step": 32920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.534995189967212e-05,
+      "loss": 2.574,
+      "step": 32930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.529333955403685e-05,
+      "loss": 2.6452,
+      "step": 32940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.523673202561281e-05,
+      "loss": 2.5687,
+      "step": 32950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.518012933294197e-05,
+      "loss": 2.5957,
+      "step": 32960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.512353149456467e-05,
+      "loss": 2.6516,
+      "step": 32970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.506693852901977e-05,
+      "loss": 2.6285,
+      "step": 32980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.501035045484441e-05,
+      "loss": 2.5496,
+      "step": 32990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.495376729057426e-05,
+      "loss": 2.5975,
+      "step": 33000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.489718905474323e-05,
+      "loss": 2.5877,
+      "step": 33010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.48406157658837e-05,
+      "loss": 2.5408,
+      "step": 33020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.478404744252648e-05,
+      "loss": 2.5708,
+      "step": 33030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.472748410320065e-05,
+      "loss": 2.7015,
+      "step": 33040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.46709257664337e-05,
+      "loss": 2.6317,
+      "step": 33050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.46143724507515e-05,
+      "loss": 2.5631,
+      "step": 33060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.455782417467826e-05,
+      "loss": 2.6817,
+      "step": 33070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.450128095673655e-05,
+      "loss": 2.6192,
+      "step": 33080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.444474281544723e-05,
+      "loss": 2.7026,
+      "step": 33090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.43882097693296e-05,
+      "loss": 2.5941,
+      "step": 33100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.433168183690119e-05,
+      "loss": 2.5896,
+      "step": 33110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.427515903667789e-05,
+      "loss": 2.6453,
+      "step": 33120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.421864138717395e-05,
+      "loss": 2.6121,
+      "step": 33130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.416212890690186e-05,
+      "loss": 2.4978,
+      "step": 33140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.410562161437244e-05,
+      "loss": 2.6365,
+      "step": 33150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.404911952809491e-05,
+      "loss": 2.6208,
+      "step": 33160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.399262266657662e-05,
+      "loss": 2.5717,
+      "step": 33170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.393613104832332e-05,
+      "loss": 2.5737,
+      "step": 33180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387964469183902e-05,
+      "loss": 2.5346,
+      "step": 33190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.382316361562597e-05,
+      "loss": 2.6137,
+      "step": 33200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376668783818472e-05,
+      "loss": 2.5206,
+      "step": 33210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37102173780141e-05,
+      "loss": 2.5274,
+      "step": 33220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.365375225361119e-05,
+      "loss": 2.5671,
+      "step": 33230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.359729248347127e-05,
+      "loss": 2.5631,
+      "step": 33240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.354083808608796e-05,
+      "loss": 2.641,
+      "step": 33250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.348438907995303e-05,
+      "loss": 2.5801,
+      "step": 33260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.342794548355652e-05,
+      "loss": 2.5264,
+      "step": 33270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.337150731538675e-05,
+      "loss": 2.6721,
+      "step": 33280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.331507459393017e-05,
+      "loss": 2.5475,
+      "step": 33290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.32586473376715e-05,
+      "loss": 2.564,
+      "step": 33300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.320222556509364e-05,
+      "loss": 2.4893,
+      "step": 33310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.314580929467776e-05,
+      "loss": 2.5932,
+      "step": 33320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.30893985449031e-05,
+      "loss": 2.6126,
+      "step": 33330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.303299333424727e-05,
+      "loss": 2.6231,
+      "step": 33340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.29765936811859e-05,
+      "loss": 2.6864,
+      "step": 33350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.29201996041929e-05,
+      "loss": 2.5874,
+      "step": 33360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.286381112174033e-05,
+      "loss": 2.4811,
+      "step": 33370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.280742825229836e-05,
+      "loss": 2.5521,
+      "step": 33380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.27510510143354e-05,
+      "loss": 2.5402,
+      "step": 33390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.269467942631801e-05,
+      "loss": 2.6696,
+      "step": 33400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.263831350671084e-05,
+      "loss": 2.6744,
+      "step": 33410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.25819532739767e-05,
+      "loss": 2.5722,
+      "step": 33420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.252559874657661e-05,
+      "loss": 2.5401,
+      "step": 33430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.246924994296963e-05,
+      "loss": 2.7006,
+      "step": 33440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.2412906881613e-05,
+      "loss": 2.5845,
+      "step": 33450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.235656958096202e-05,
+      "loss": 2.6693,
+      "step": 33460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.230023805947019e-05,
+      "loss": 2.5728,
+      "step": 33470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.224391233558905e-05,
+      "loss": 2.6132,
+      "step": 33480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.218759242776826e-05,
+      "loss": 2.5298,
+      "step": 33490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.21312783544556e-05,
+      "loss": 2.6541,
+      "step": 33500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.20749701340969e-05,
+      "loss": 2.5383,
+      "step": 33510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.201866778513604e-05,
+      "loss": 2.6517,
+      "step": 33520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.196237132601511e-05,
+      "loss": 2.5983,
+      "step": 33530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.190608077517413e-05,
+      "loss": 2.6201,
+      "step": 33540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.184979615105129e-05,
+      "loss": 2.6422,
+      "step": 33550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.179351747208274e-05,
+      "loss": 2.5995,
+      "step": 33560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.173724475670273e-05,
+      "loss": 2.6187,
+      "step": 33570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.16809780233436e-05,
+      "loss": 2.6193,
+      "step": 33580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.162471729043567e-05,
+      "loss": 2.6017,
+      "step": 33590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.156846257640732e-05,
+      "loss": 2.6517,
+      "step": 33600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.151221389968491e-05,
+      "loss": 2.582,
+      "step": 33610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.145597127869292e-05,
+      "loss": 2.6222,
+      "step": 33620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.139973473185377e-05,
+      "loss": 2.6,
+      "step": 33630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.134350427758786e-05,
+      "loss": 2.6229,
+      "step": 33640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.128727993431375e-05,
+      "loss": 2.64,
+      "step": 33650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.123106172044783e-05,
+      "loss": 2.6772,
+      "step": 33660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.117484965440454e-05,
+      "loss": 2.5327,
+      "step": 33670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.111864375459635e-05,
+      "loss": 2.617,
+      "step": 33680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.106244403943364e-05,
+      "loss": 2.5626,
+      "step": 33690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.100625052732477e-05,
+      "loss": 2.5397,
+      "step": 33700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.095006323667619e-05,
+      "loss": 2.561,
+      "step": 33710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.089388218589213e-05,
+      "loss": 2.5349,
+      "step": 33720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.08377073933749e-05,
+      "loss": 2.5757,
+      "step": 33730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.078153887752474e-05,
+      "loss": 2.6291,
+      "step": 33740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.07253766567398e-05,
+      "loss": 2.7031,
+      "step": 33750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066922074941615e-05,
+      "loss": 2.5763,
+      "step": 33760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.061307117394787e-05,
+      "loss": 2.4859,
+      "step": 33770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055692794872694e-05,
+      "loss": 2.5904,
+      "step": 33780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.050079109214317e-05,
+      "loss": 2.6739,
+      "step": 33790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044466062258445e-05,
+      "loss": 2.5431,
+      "step": 33800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.038853655843642e-05,
+      "loss": 2.6639,
+      "step": 33810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.03324189180827e-05,
+      "loss": 2.536,
+      "step": 33820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.02763077199048e-05,
+      "loss": 2.6177,
+      "step": 33830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.022020298228215e-05,
+      "loss": 2.5787,
+      "step": 33840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.016410472359195e-05,
+      "loss": 2.5849,
+      "step": 33850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.01080129622094e-05,
+      "loss": 2.6167,
+      "step": 33860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 8.005192771650755e-05,
+      "loss": 2.6675,
+      "step": 33870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.99958490048572e-05,
+      "loss": 2.637,
+      "step": 33880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.993977684562722e-05,
+      "loss": 2.5651,
+      "step": 33890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.988371125718417e-05,
+      "loss": 2.6825,
+      "step": 33900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.982765225789247e-05,
+      "loss": 2.5359,
+      "step": 33910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.977159986611444e-05,
+      "loss": 2.6403,
+      "step": 33920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.971555410021023e-05,
+      "loss": 2.57,
+      "step": 33930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.965951497853776e-05,
+      "loss": 2.6122,
+      "step": 33940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.960348251945286e-05,
+      "loss": 2.5871,
+      "step": 33950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.954745674130911e-05,
+      "loss": 2.6274,
+      "step": 33960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.94914376624579e-05,
+      "loss": 2.5834,
+      "step": 33970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.943542530124848e-05,
+      "loss": 2.5951,
+      "step": 33980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.937941967602787e-05,
+      "loss": 2.4752,
+      "step": 33990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.932342080514088e-05,
+      "loss": 2.5861,
+      "step": 34000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.926742870693008e-05,
+      "loss": 2.6264,
+      "step": 34010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.921144339973588e-05,
+      "loss": 2.5943,
+      "step": 34020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.915546490189645e-05,
+      "loss": 2.5767,
+      "step": 34030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.909949323174765e-05,
+      "loss": 2.558,
+      "step": 34040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.904352840762326e-05,
+      "loss": 2.5735,
+      "step": 34050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.898757044785469e-05,
+      "loss": 2.5535,
+      "step": 34060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.893161937077107e-05,
+      "loss": 2.5727,
+      "step": 34070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.887567519469947e-05,
+      "loss": 2.5929,
+      "step": 34080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.881973793796448e-05,
+      "loss": 2.5686,
+      "step": 34090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.87638076188886e-05,
+      "loss": 2.6726,
+      "step": 34100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.870788425579191e-05,
+      "loss": 2.6218,
+      "step": 34110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.865196786699227e-05,
+      "loss": 2.6214,
+      "step": 34120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.859605847080531e-05,
+      "loss": 2.569,
+      "step": 34130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.854015608554431e-05,
+      "loss": 2.5777,
+      "step": 34140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.848426072952025e-05,
+      "loss": 2.7291,
+      "step": 34150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.842837242104182e-05,
+      "loss": 2.6412,
+      "step": 34160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.837249117841542e-05,
+      "loss": 2.58,
+      "step": 34170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.831661701994512e-05,
+      "loss": 2.503,
+      "step": 34180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.826074996393264e-05,
+      "loss": 2.5948,
+      "step": 34190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.820489002867745e-05,
+      "loss": 2.58,
+      "step": 34200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.814903723247662e-05,
+      "loss": 2.5378,
+      "step": 34210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.809319159362489e-05,
+      "loss": 2.5758,
+      "step": 34220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.80373531304147e-05,
+      "loss": 2.5994,
+      "step": 34230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.798152186113609e-05,
+      "loss": 2.4807,
+      "step": 34240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.792569780407671e-05,
+      "loss": 2.535,
+      "step": 34250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7869880977522e-05,
+      "loss": 2.596,
+      "step": 34260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.781407139975485e-05,
+      "loss": 2.5779,
+      "step": 34270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.775826908905591e-05,
+      "loss": 2.597,
+      "step": 34280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.770247406370338e-05,
+      "loss": 2.5891,
+      "step": 34290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.764668634197308e-05,
+      "loss": 2.5256,
+      "step": 34300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.759090594213843e-05,
+      "loss": 2.588,
+      "step": 34310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.75351328824705e-05,
+      "loss": 2.6114,
+      "step": 34320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747936718123791e-05,
+      "loss": 2.6108,
+      "step": 34330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742360885670688e-05,
+      "loss": 2.6,
+      "step": 34340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736785792714125e-05,
+      "loss": 2.5836,
+      "step": 34350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731211441080239e-05,
+      "loss": 2.6074,
+      "step": 34360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72563783259492e-05,
+      "loss": 2.5523,
+      "step": 34370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72006496908383e-05,
+      "loss": 2.599,
+      "step": 34380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.714492852372372e-05,
+      "loss": 2.595,
+      "step": 34390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.708921484285712e-05,
+      "loss": 2.6365,
+      "step": 34400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.703350866648763e-05,
+      "loss": 2.5226,
+      "step": 34410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.697781001286203e-05,
+      "loss": 2.6495,
+      "step": 34420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.692211890022455e-05,
+      "loss": 2.5864,
+      "step": 34430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.686643534681703e-05,
+      "loss": 2.5863,
+      "step": 34440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.681075937087876e-05,
+      "loss": 2.6029,
+      "step": 34450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.675509099064656e-05,
+      "loss": 2.4836,
+      "step": 34460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.66994302243548e-05,
+      "loss": 2.5914,
+      "step": 34470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.664377709023533e-05,
+      "loss": 2.6024,
+      "step": 34480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.658813160651749e-05,
+      "loss": 2.6303,
+      "step": 34490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.653249379142813e-05,
+      "loss": 2.5376,
+      "step": 34500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.647686366319161e-05,
+      "loss": 2.5722,
+      "step": 34510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.64212412400297e-05,
+      "loss": 2.5708,
+      "step": 34520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.636562654016173e-05,
+      "loss": 2.5913,
+      "step": 34530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.631001958180446e-05,
+      "loss": 2.6042,
+      "step": 34540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.625442038317211e-05,
+      "loss": 2.5326,
+      "step": 34550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.620438775401977e-05,
+      "loss": 2.6521,
+      "step": 34560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.614880334903584e-05,
+      "loss": 2.6514,
+      "step": 34570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.609322675658373e-05,
+      "loss": 2.5918,
+      "step": 34580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.603765799486766e-05,
+      "loss": 2.5627,
+      "step": 34590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.598209708208936e-05,
+      "loss": 2.6356,
+      "step": 34600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.592654403644802e-05,
+      "loss": 2.6433,
+      "step": 34610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.587099887614017e-05,
+      "loss": 2.5665,
+      "step": 34620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.581546161935978e-05,
+      "loss": 2.6039,
+      "step": 34630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.57599322842983e-05,
+      "loss": 2.6104,
+      "step": 34640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.57044108891445e-05,
+      "loss": 2.5843,
+      "step": 34650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.564889745208459e-05,
+      "loss": 2.6189,
+      "step": 34660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.559339199130212e-05,
+      "loss": 2.594,
+      "step": 34670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.553789452497814e-05,
+      "loss": 2.6001,
+      "step": 34680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.5482405071291e-05,
+      "loss": 2.6883,
+      "step": 34690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.542692364841638e-05,
+      "loss": 2.5357,
+      "step": 34700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.537145027452746e-05,
+      "loss": 2.6797,
+      "step": 34710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.531598496779468e-05,
+      "loss": 2.5746,
+      "step": 34720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.526052774638585e-05,
+      "loss": 2.5722,
+      "step": 34730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.52050786284662e-05,
+      "loss": 2.6161,
+      "step": 34740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.51496376321982e-05,
+      "loss": 2.5552,
+      "step": 34750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.509420477574177e-05,
+      "loss": 2.5169,
+      "step": 34760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.503878007725409e-05,
+      "loss": 2.566,
+      "step": 34770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.498336355488967e-05,
+      "loss": 2.6428,
+      "step": 34780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.492795522680038e-05,
+      "loss": 2.5193,
+      "step": 34790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.48725551111354e-05,
+      "loss": 2.5901,
+      "step": 34800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.481716322604118e-05,
+      "loss": 2.5946,
+      "step": 34810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.476177958966149e-05,
+      "loss": 2.573,
+      "step": 34820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.470640422013744e-05,
+      "loss": 2.6475,
+      "step": 34830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.465103713560739e-05,
+      "loss": 2.6166,
+      "step": 34840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.459567835420695e-05,
+      "loss": 2.5889,
+      "step": 34850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.454032789406914e-05,
+      "loss": 2.5249,
+      "step": 34860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.448498577332412e-05,
+      "loss": 2.5442,
+      "step": 34870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.442965201009937e-05,
+      "loss": 2.6616,
+      "step": 34880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437432662251964e-05,
+      "loss": 2.5627,
+      "step": 34890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431900962870695e-05,
+      "loss": 2.6493,
+      "step": 34900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.426370104678051e-05,
+      "loss": 2.5267,
+      "step": 34910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420840089485683e-05,
+      "loss": 2.6105,
+      "step": 34920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.415310919104961e-05,
+      "loss": 2.6312,
+      "step": 34930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409782595346988e-05,
+      "loss": 2.637,
+      "step": 34940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404255120022582e-05,
+      "loss": 2.5854,
+      "step": 34950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.398728494942282e-05,
+      "loss": 2.6318,
+      "step": 34960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.393202721916348e-05,
+      "loss": 2.6505,
+      "step": 34970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.387677802754771e-05,
+      "loss": 2.588,
+      "step": 34980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.38215373926725e-05,
+      "loss": 2.6385,
+      "step": 34990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.376630533263207e-05,
+      "loss": 2.684,
+      "step": 35000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.37110818655179e-05,
+      "loss": 2.5807,
+      "step": 35010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.365586700941858e-05,
+      "loss": 2.5373,
+      "step": 35020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.36006607824199e-05,
+      "loss": 2.5346,
+      "step": 35030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.354546320260483e-05,
+      "loss": 2.5268,
+      "step": 35040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.349027428805351e-05,
+      "loss": 2.6294,
+      "step": 35050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.343509405684324e-05,
+      "loss": 2.5228,
+      "step": 35060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.337992252704842e-05,
+      "loss": 2.5283,
+      "step": 35070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.332475971674075e-05,
+      "loss": 2.4926,
+      "step": 35080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.32696056439889e-05,
+      "loss": 2.6291,
+      "step": 35090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.321446032685875e-05,
+      "loss": 2.5814,
+      "step": 35100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.315932378341335e-05,
+      "loss": 2.6037,
+      "step": 35110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.310419603171283e-05,
+      "loss": 2.6279,
+      "step": 35120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.304907708981448e-05,
+      "loss": 2.5788,
+      "step": 35130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.299396697577266e-05,
+      "loss": 2.5565,
+      "step": 35140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.293886570763882e-05,
+      "loss": 2.5908,
+      "step": 35150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.288377330346158e-05,
+      "loss": 2.5817,
+      "step": 35160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.282868978128663e-05,
+      "loss": 2.6622,
+      "step": 35170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.27736151591567e-05,
+      "loss": 2.5478,
+      "step": 35180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.271854945511172e-05,
+      "loss": 2.5753,
+      "step": 35190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.266349268718859e-05,
+      "loss": 2.6001,
+      "step": 35200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.260844487342131e-05,
+      "loss": 2.5585,
+      "step": 35210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.255340603184096e-05,
+      "loss": 2.5817,
+      "step": 35220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.249837618047571e-05,
+      "loss": 2.5876,
+      "step": 35230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.244335533735073e-05,
+      "loss": 2.5905,
+      "step": 35240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.238834352048824e-05,
+      "loss": 2.676,
+      "step": 35250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.233334074790759e-05,
+      "loss": 2.4738,
+      "step": 35260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.227834703762507e-05,
+      "loss": 2.5569,
+      "step": 35270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.2223362407654e-05,
+      "loss": 2.6139,
+      "step": 35280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.216838687600484e-05,
+      "loss": 2.596,
+      "step": 35290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.21134204606849e-05,
+      "loss": 2.602,
+      "step": 35300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.205846317969869e-05,
+      "loss": 2.6326,
+      "step": 35310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.20035150510476e-05,
+      "loss": 2.5807,
+      "step": 35320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.194857609273002e-05,
+      "loss": 2.7027,
+      "step": 35330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.189364632274143e-05,
+      "loss": 2.5356,
+      "step": 35340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.18387257590742e-05,
+      "loss": 2.6058,
+      "step": 35350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.178381441971774e-05,
+      "loss": 2.5583,
+      "step": 35360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.172891232265842e-05,
+      "loss": 2.6578,
+      "step": 35370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.167401948587959e-05,
+      "loss": 2.4681,
+      "step": 35380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.161913592736159e-05,
+      "loss": 2.5298,
+      "step": 35390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.156426166508166e-05,
+      "loss": 2.5655,
+      "step": 35400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.150939671701404e-05,
+      "loss": 2.5263,
+      "step": 35410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.145454110112995e-05,
+      "loss": 2.505,
+      "step": 35420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.139969483539743e-05,
+      "loss": 2.5579,
+      "step": 35430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.134485793778162e-05,
+      "loss": 2.7049,
+      "step": 35440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.12900304262445e-05,
+      "loss": 2.6231,
+      "step": 35450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123521231874495e-05,
+      "loss": 2.5835,
+      "step": 35460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118040363323882e-05,
+      "loss": 2.4762,
+      "step": 35470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112560438767884e-05,
+      "loss": 2.5857,
+      "step": 35480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107081460001473e-05,
+      "loss": 2.5826,
+      "step": 35490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101603428819304e-05,
+      "loss": 2.62,
+      "step": 35500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096126347015719e-05,
+      "loss": 2.5589,
+      "step": 35510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090650216384753e-05,
+      "loss": 2.6172,
+      "step": 35520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085175038720133e-05,
+      "loss": 2.5633,
+      "step": 35530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.079700815815267e-05,
+      "loss": 2.5318,
+      "step": 35540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.07422754946325e-05,
+      "loss": 2.5164,
+      "step": 35550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.068755241456876e-05,
+      "loss": 2.6291,
+      "step": 35560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.06328389358861e-05,
+      "loss": 2.5749,
+      "step": 35570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.057813507650607e-05,
+      "loss": 2.5037,
+      "step": 35580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.052344085434713e-05,
+      "loss": 2.6721,
+      "step": 35590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.046875628732452e-05,
+      "loss": 2.6177,
+      "step": 35600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.041408139335035e-05,
+      "loss": 2.5694,
+      "step": 35610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.03594161903335e-05,
+      "loss": 2.549,
+      "step": 35620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.030476069617978e-05,
+      "loss": 2.5129,
+      "step": 35630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.025011492879172e-05,
+      "loss": 2.5913,
+      "step": 35640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.019547890606872e-05,
+      "loss": 2.5305,
+      "step": 35650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.014085264590698e-05,
+      "loss": 2.5755,
+      "step": 35660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.008623616619949e-05,
+      "loss": 2.6259,
+      "step": 35670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.003162948483608e-05,
+      "loss": 2.6205,
+      "step": 35680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.99770326197033e-05,
+      "loss": 2.5844,
+      "step": 35690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.99224455886845e-05,
+      "loss": 2.5366,
+      "step": 35700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.986786840965987e-05,
+      "loss": 2.6098,
+      "step": 35710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.981330110050632e-05,
+      "loss": 2.4961,
+      "step": 35720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.975874367909749e-05,
+      "loss": 2.6201,
+      "step": 35730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.97041961633039e-05,
+      "loss": 2.5613,
+      "step": 35740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.964965857099273e-05,
+      "loss": 2.5434,
+      "step": 35750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.959513092002791e-05,
+      "loss": 2.5496,
+      "step": 35760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.954061322827014e-05,
+      "loss": 2.5976,
+      "step": 35770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.94861055135769e-05,
+      "loss": 2.5743,
+      "step": 35780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.943160779380231e-05,
+      "loss": 2.6126,
+      "step": 35790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.937712008679726e-05,
+      "loss": 2.5781,
+      "step": 35800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.932264241040941e-05,
+      "loss": 2.6386,
+      "step": 35810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.926817478248306e-05,
+      "loss": 2.6004,
+      "step": 35820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.92137172208592e-05,
+      "loss": 2.5086,
+      "step": 35830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.915926974337567e-05,
+      "loss": 2.6832,
+      "step": 35840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.91048323678668e-05,
+      "loss": 2.5596,
+      "step": 35850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.905040511216382e-05,
+      "loss": 2.5465,
+      "step": 35860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.899598799409448e-05,
+      "loss": 2.5644,
+      "step": 35870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.894158103148326e-05,
+      "loss": 2.5368,
+      "step": 35880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.88871842421514e-05,
+      "loss": 2.5657,
+      "step": 35890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.883279764391666e-05,
+      "loss": 2.5646,
+      "step": 35900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.877842125459357e-05,
+      "loss": 2.6155,
+      "step": 35910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.872405509199324e-05,
+      "loss": 2.5773,
+      "step": 35920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.866969917392353e-05,
+      "loss": 2.5641,
+      "step": 35930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.861535351818885e-05,
+      "loss": 2.5506,
+      "step": 35940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.856101814259028e-05,
+      "loss": 2.4932,
+      "step": 35950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.850669306492557e-05,
+      "loss": 2.541,
+      "step": 35960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.845237830298903e-05,
+      "loss": 2.5564,
+      "step": 35970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.839807387457163e-05,
+      "loss": 2.555,
+      "step": 35980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.834377979746096e-05,
+      "loss": 2.5823,
+      "step": 35990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.828949608944122e-05,
+      "loss": 2.5998,
+      "step": 36000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.823522276829319e-05,
+      "loss": 2.5527,
+      "step": 36010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.818095985179425e-05,
+      "loss": 2.6562,
+      "step": 36020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812670735771837e-05,
+      "loss": 2.5415,
+      "step": 36030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.807246530383617e-05,
+      "loss": 2.7046,
+      "step": 36040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801823370791478e-05,
+      "loss": 2.5594,
+      "step": 36050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796401258771792e-05,
+      "loss": 2.6434,
+      "step": 36060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790980196100587e-05,
+      "loss": 2.5518,
+      "step": 36070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78556018455355e-05,
+      "loss": 2.5493,
+      "step": 36080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.780141225906022e-05,
+      "loss": 2.5773,
+      "step": 36090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774723321932997e-05,
+      "loss": 2.5247,
+      "step": 36100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769306474409129e-05,
+      "loss": 2.5798,
+      "step": 36110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.763890685108723e-05,
+      "loss": 2.5597,
+      "step": 36120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.758475955805731e-05,
+      "loss": 2.5723,
+      "step": 36130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.753062288273774e-05,
+      "loss": 2.549,
+      "step": 36140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.747649684286109e-05,
+      "loss": 2.6087,
+      "step": 36150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.742238145615652e-05,
+      "loss": 2.5242,
+      "step": 36160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.736827674034967e-05,
+      "loss": 2.6114,
+      "step": 36170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.731418271316271e-05,
+      "loss": 2.596,
+      "step": 36180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.726009939231437e-05,
+      "loss": 2.5613,
+      "step": 36190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.72060267955197e-05,
+      "loss": 2.5341,
+      "step": 36200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.715196494049038e-05,
+      "loss": 2.6262,
+      "step": 36210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.709791384493458e-05,
+      "loss": 2.5625,
+      "step": 36220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.704387352655685e-05,
+      "loss": 2.5614,
+      "step": 36230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.698984400305832e-05,
+      "loss": 2.5329,
+      "step": 36240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.693582529213644e-05,
+      "loss": 2.5221,
+      "step": 36250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.688181741148527e-05,
+      "loss": 2.6425,
+      "step": 36260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.682782037879523e-05,
+      "loss": 2.6439,
+      "step": 36270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.677383421175319e-05,
+      "loss": 2.6683,
+      "step": 36280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.671985892804251e-05,
+      "loss": 2.5385,
+      "step": 36290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.666589454534298e-05,
+      "loss": 2.5179,
+      "step": 36300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.661194108133076e-05,
+      "loss": 2.6157,
+      "step": 36310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.655799855367844e-05,
+      "loss": 2.5276,
+      "step": 36320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.650406698005514e-05,
+      "loss": 2.584,
+      "step": 36330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.645014637812627e-05,
+      "loss": 2.6152,
+      "step": 36340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.639623676555365e-05,
+      "loss": 2.4669,
+      "step": 36350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.634233815999561e-05,
+      "loss": 2.5811,
+      "step": 36360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.628845057910675e-05,
+      "loss": 2.5907,
+      "step": 36370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.623457404053812e-05,
+      "loss": 2.5852,
+      "step": 36380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.618070856193714e-05,
+      "loss": 2.5621,
+      "step": 36390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.612685416094763e-05,
+      "loss": 2.5695,
+      "step": 36400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.607301085520976e-05,
+      "loss": 2.6433,
+      "step": 36410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.601917866236007e-05,
+      "loss": 2.573,
+      "step": 36420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.596535760003142e-05,
+      "loss": 2.5371,
+      "step": 36430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.591154768585312e-05,
+      "loss": 2.6182,
+      "step": 36440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.585774893745074e-05,
+      "loss": 2.5921,
+      "step": 36450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.580396137244622e-05,
+      "loss": 2.5038,
+      "step": 36460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.575018500845781e-05,
+      "loss": 2.5698,
+      "step": 36470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.56964198631002e-05,
+      "loss": 2.5809,
+      "step": 36480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.564266595398425e-05,
+      "loss": 2.5646,
+      "step": 36490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.558892329871725e-05,
+      "loss": 2.5122,
+      "step": 36500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.553519191490278e-05,
+      "loss": 2.564,
+      "step": 36510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.548147182014072e-05,
+      "loss": 2.581,
+      "step": 36520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.54277630320272e-05,
+      "loss": 2.5356,
+      "step": 36530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.537406556815475e-05,
+      "loss": 2.5634,
+      "step": 36540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.532037944611214e-05,
+      "loss": 2.6559,
+      "step": 36550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.526670468348437e-05,
+      "loss": 2.5757,
+      "step": 36560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.521304129785279e-05,
+      "loss": 2.5909,
+      "step": 36570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5159389306795e-05,
+      "loss": 2.607,
+      "step": 36580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510574872788493e-05,
+      "loss": 2.5227,
+      "step": 36590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505211957869268e-05,
+      "loss": 2.5489,
+      "step": 36600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499850187678463e-05,
+      "loss": 2.4845,
+      "step": 36610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494489563972339e-05,
+      "loss": 2.6684,
+      "step": 36620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489130088506792e-05,
+      "loss": 2.5734,
+      "step": 36630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483771763037329e-05,
+      "loss": 2.6273,
+      "step": 36640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.478414589319085e-05,
+      "loss": 2.5202,
+      "step": 36650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.473058569106823e-05,
+      "loss": 2.578,
+      "step": 36660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467703704154922e-05,
+      "loss": 2.6493,
+      "step": 36670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46234999621738e-05,
+      "loss": 2.5492,
+      "step": 36680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456997447047828e-05,
+      "loss": 2.5267,
+      "step": 36690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.451646058399505e-05,
+      "loss": 2.5268,
+      "step": 36700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.446295832025274e-05,
+      "loss": 2.5746,
+      "step": 36710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.440946769677618e-05,
+      "loss": 2.6034,
+      "step": 36720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.435598873108641e-05,
+      "loss": 2.595,
+      "step": 36730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.430252144070062e-05,
+      "loss": 2.5384,
+      "step": 36740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.424906584313215e-05,
+      "loss": 2.5901,
+      "step": 36750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.419562195589055e-05,
+      "loss": 2.6025,
+      "step": 36760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.414218979648159e-05,
+      "loss": 2.6463,
+      "step": 36770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.408876938240708e-05,
+      "loss": 2.4858,
+      "step": 36780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.403536073116508e-05,
+      "loss": 2.6421,
+      "step": 36790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.398196386024968e-05,
+      "loss": 2.5706,
+      "step": 36800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.392857878715126e-05,
+      "loss": 2.556,
+      "step": 36810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.387520552935624e-05,
+      "loss": 2.5829,
+      "step": 36820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.382184410434717e-05,
+      "loss": 2.6171,
+      "step": 36830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.376849452960279e-05,
+      "loss": 2.6515,
+      "step": 36840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.371515682259789e-05,
+      "loss": 2.5681,
+      "step": 36850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.366183100080338e-05,
+      "loss": 2.5628,
+      "step": 36860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.360851708168631e-05,
+      "loss": 2.5906,
+      "step": 36870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.355521508270985e-05,
+      "loss": 2.6443,
+      "step": 36880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.350192502133319e-05,
+      "loss": 2.5287,
+      "step": 36890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.344864691501162e-05,
+      "loss": 2.6674,
+      "step": 36900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.339538078119663e-05,
+      "loss": 2.5651,
+      "step": 36910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.334212663733566e-05,
+      "loss": 2.5261,
+      "step": 36920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.328888450087222e-05,
+      "loss": 2.6076,
+      "step": 36930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.323565438924601e-05,
+      "loss": 2.6554,
+      "step": 36940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.318243631989267e-05,
+      "loss": 2.4875,
+      "step": 36950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.312923031024398e-05,
+      "loss": 2.5971,
+      "step": 36960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.30760363777277e-05,
+      "loss": 2.5634,
+      "step": 36970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.302285453976765e-05,
+      "loss": 2.5727,
+      "step": 36980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.296968481378374e-05,
+      "loss": 2.5701,
+      "step": 36990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.291652721719185e-05,
+      "loss": 2.6221,
+      "step": 37000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.286338176740392e-05,
+      "loss": 2.496,
+      "step": 37010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.281024848182788e-05,
+      "loss": 2.5735,
+      "step": 37020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.275712737786774e-05,
+      "loss": 2.5896,
+      "step": 37030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.270401847292347e-05,
+      "loss": 2.5677,
+      "step": 37040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.265092178439101e-05,
+      "loss": 2.5314,
+      "step": 37050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.259783732966238e-05,
+      "loss": 2.5964,
+      "step": 37060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.254476512612557e-05,
+      "loss": 2.5033,
+      "step": 37070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.249170519116448e-05,
+      "loss": 2.5392,
+      "step": 37080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.243865754215912e-05,
+      "loss": 2.6184,
+      "step": 37090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.238562219648537e-05,
+      "loss": 2.637,
+      "step": 37100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.233259917151512e-05,
+      "loss": 2.6307,
+      "step": 37110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.227958848461623e-05,
+      "loss": 2.5424,
+      "step": 37120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.222659015315246e-05,
+      "loss": 2.5822,
+      "step": 37130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.217360419448369e-05,
+      "loss": 2.5492,
+      "step": 37140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.212063062596556e-05,
+      "loss": 2.5797,
+      "step": 37150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206766946494973e-05,
+      "loss": 2.5748,
+      "step": 37160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201472072878376e-05,
+      "loss": 2.6574,
+      "step": 37170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196178443481123e-05,
+      "loss": 2.5532,
+      "step": 37180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190886060037154e-05,
+      "loss": 2.579,
+      "step": 37190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185594924280004e-05,
+      "loss": 2.6252,
+      "step": 37200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180305037942806e-05,
+      "loss": 2.6023,
+      "step": 37210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175016402758275e-05,
+      "loss": 2.6692,
+      "step": 37220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169729020458717e-05,
+      "loss": 2.5516,
+      "step": 37230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164442892776034e-05,
+      "loss": 2.5852,
+      "step": 37240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159158021441715e-05,
+      "loss": 2.6332,
+      "step": 37250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.15387440818683e-05,
+      "loss": 2.557,
+      "step": 37260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.148592054742043e-05,
+      "loss": 2.527,
+      "step": 37270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.14331096283761e-05,
+      "loss": 2.5868,
+      "step": 37280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.138031134203364e-05,
+      "loss": 2.6475,
+      "step": 37290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.13275257056873e-05,
+      "loss": 2.5719,
+      "step": 37300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.127475273662716e-05,
+      "loss": 2.5748,
+      "step": 37310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.122199245213922e-05,
+      "loss": 2.6365,
+      "step": 37320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.116924486950524e-05,
+      "loss": 2.6182,
+      "step": 37330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.111651000600286e-05,
+      "loss": 2.5192,
+      "step": 37340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.10637878789055e-05,
+      "loss": 2.6013,
+      "step": 37350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.101107850548251e-05,
+      "loss": 2.6308,
+      "step": 37360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0958381902998986e-05,
+      "loss": 2.6071,
+      "step": 37370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.090569808871582e-05,
+      "loss": 2.6105,
+      "step": 37380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.085302707988981e-05,
+      "loss": 2.6545,
+      "step": 37390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.080036889377349e-05,
+      "loss": 2.6144,
+      "step": 37400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.074772354761521e-05,
+      "loss": 2.5806,
+      "step": 37410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.069509105865907e-05,
+      "loss": 2.5822,
+      "step": 37420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0642471444145075e-05,
+      "loss": 2.5347,
+      "step": 37430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0589864721308886e-05,
+      "loss": 2.5384,
+      "step": 37440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.053727090738199e-05,
+      "loss": 2.6357,
+      "step": 37450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0484690019591696e-05,
+      "loss": 2.4609,
+      "step": 37460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.0432122075161015e-05,
+      "loss": 2.4859,
+      "step": 37470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.03795670913087e-05,
+      "loss": 2.4926,
+      "step": 37480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.032702508524932e-05,
+      "loss": 2.6422,
+      "step": 37490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.027449607419318e-05,
+      "loss": 2.485,
+      "step": 37500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.022198007534632e-05,
+      "loss": 2.5795,
+      "step": 37510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.016947710591051e-05,
+      "loss": 2.6305,
+      "step": 37520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.011698718308323e-05,
+      "loss": 2.614,
+      "step": 37530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.006451032405775e-05,
+      "loss": 2.559,
+      "step": 37540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.001204654602302e-05,
+      "loss": 2.5698,
+      "step": 37550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.995959586616369e-05,
+      "loss": 2.5074,
+      "step": 37560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9907158301660125e-05,
+      "loss": 2.5574,
+      "step": 37570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.985473386968845e-05,
+      "loss": 2.6022,
+      "step": 37580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.980232258742045e-05,
+      "loss": 2.5784,
+      "step": 37590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9749924472023524e-05,
+      "loss": 2.5776,
+      "step": 37600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.969753954066093e-05,
+      "loss": 2.6277,
+      "step": 37610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.964516781049145e-05,
+      "loss": 2.6681,
+      "step": 37620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9592809298669605e-05,
+      "loss": 2.5396,
+      "step": 37630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9540464022345636e-05,
+      "loss": 2.5541,
+      "step": 37640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9488131998665345e-05,
+      "loss": 2.5115,
+      "step": 37650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9435813244770255e-05,
+      "loss": 2.4931,
+      "step": 37660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.938350777779753e-05,
+      "loss": 2.4796,
+      "step": 37670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9331215614879966e-05,
+      "loss": 2.5555,
+      "step": 37680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9278936773146085e-05,
+      "loss": 2.4983,
+      "step": 37690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.922667126971995e-05,
+      "loss": 2.5692,
+      "step": 37700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.917441912172127e-05,
+      "loss": 2.5328,
+      "step": 37710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9122180346265375e-05,
+      "loss": 2.4906,
+      "step": 37720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906995496046329e-05,
+      "loss": 2.5057,
+      "step": 37730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901774298142157e-05,
+      "loss": 2.5872,
+      "step": 37740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896554442624238e-05,
+      "loss": 2.5632,
+      "step": 37750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.891335931202355e-05,
+      "loss": 2.5524,
+      "step": 37760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886118765585849e-05,
+      "loss": 2.5227,
+      "step": 37770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.880902947483612e-05,
+      "loss": 2.4414,
+      "step": 37780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.875688478604107e-05,
+      "loss": 2.4927,
+      "step": 37790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870475360655346e-05,
+      "loss": 2.4679,
+      "step": 37800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865263595344904e-05,
+      "loss": 2.6439,
+      "step": 37810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.860053184379906e-05,
+      "loss": 2.5945,
+      "step": 37820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854844129467043e-05,
+      "loss": 2.5555,
+      "step": 37830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849636432312555e-05,
+      "loss": 2.5763,
+      "step": 37840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8444300946222374e-05,
+      "loss": 2.6156,
+      "step": 37850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.839225118101439e-05,
+      "loss": 2.5677,
+      "step": 37860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.834021504455075e-05,
+      "loss": 2.5279,
+      "step": 37870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8288192553875976e-05,
+      "loss": 2.5518,
+      "step": 37880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8236183726030256e-05,
+      "loss": 2.516,
+      "step": 37890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.818418857804915e-05,
+      "loss": 2.5574,
+      "step": 37900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.8132207126963915e-05,
+      "loss": 2.568,
+      "step": 37910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.808023938980114e-05,
+      "loss": 2.6242,
+      "step": 37920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.802828538358307e-05,
+      "loss": 2.5777,
+      "step": 37930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7976345125327434e-05,
+      "loss": 2.5008,
+      "step": 37940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.792441863204734e-05,
+      "loss": 2.4446,
+      "step": 37950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7872505920751504e-05,
+      "loss": 2.5487,
+      "step": 37960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.782060700844414e-05,
+      "loss": 2.6023,
+      "step": 37970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.776872191212481e-05,
+      "loss": 2.555,
+      "step": 37980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.771685064878868e-05,
+      "loss": 2.5337,
+      "step": 37990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.766499323542637e-05,
+      "loss": 2.5177,
+      "step": 38000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.761314968902385e-05,
+      "loss": 2.5643,
+      "step": 38010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7561320026562684e-05,
+      "loss": 2.5264,
+      "step": 38020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7509504265019865e-05,
+      "loss": 2.5389,
+      "step": 38030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7457702421367676e-05,
+      "loss": 2.4855,
+      "step": 38040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.740591451257415e-05,
+      "loss": 2.5738,
+      "step": 38050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7354140555602456e-05,
+      "loss": 2.6184,
+      "step": 38060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.730238056741132e-05,
+      "loss": 2.6678,
+      "step": 38070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.725063456495494e-05,
+      "loss": 2.5479,
+      "step": 38080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.71989025651828e-05,
+      "loss": 2.5564,
+      "step": 38090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.7147184585039914e-05,
+      "loss": 2.6261,
+      "step": 38100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.709548064146668e-05,
+      "loss": 2.519,
+      "step": 38110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.704379075139884e-05,
+      "loss": 2.5923,
+      "step": 38120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6992114931767593e-05,
+      "loss": 2.4984,
+      "step": 38130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.694045319949954e-05,
+      "loss": 2.5886,
+      "step": 38140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6888805571516565e-05,
+      "loss": 2.5002,
+      "step": 38150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.683717206473611e-05,
+      "loss": 2.5459,
+      "step": 38160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6785552696070774e-05,
+      "loss": 2.5428,
+      "step": 38170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.673394748242868e-05,
+      "loss": 2.5834,
+      "step": 38180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.668235644071335e-05,
+      "loss": 2.6371,
+      "step": 38190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.663077958782346e-05,
+      "loss": 2.557,
+      "step": 38200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.657921694065322e-05,
+      "loss": 2.5347,
+      "step": 38210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.652766851609219e-05,
+      "loss": 2.6268,
+      "step": 38220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.647613433102506e-05,
+      "loss": 2.568,
+      "step": 38230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.64246144023322e-05,
+      "loss": 2.5543,
+      "step": 38240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6373108746888966e-05,
+      "loss": 2.5791,
+      "step": 38250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.632161738156627e-05,
+      "loss": 2.5715,
+      "step": 38260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.627014032323028e-05,
+      "loss": 2.6108,
+      "step": 38270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6218677588742394e-05,
+      "loss": 2.5312,
+      "step": 38280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6167229194959404e-05,
+      "loss": 2.5703,
+      "step": 38290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611579515873346e-05,
+      "loss": 2.552,
+      "step": 38300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6064375496911834e-05,
+      "loss": 2.6514,
+      "step": 38310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601297022633724e-05,
+      "loss": 2.5467,
+      "step": 38320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5961579363847663e-05,
+      "loss": 2.5322,
+      "step": 38330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5910202926276265e-05,
+      "loss": 2.5829,
+      "step": 38340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5858840930451595e-05,
+      "loss": 2.5866,
+      "step": 38350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580749339319746e-05,
+      "loss": 2.5347,
+      "step": 38360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.575616033133283e-05,
+      "loss": 2.5481,
+      "step": 38370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5704841761672044e-05,
+      "loss": 2.5428,
+      "step": 38380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.565353770102468e-05,
+      "loss": 2.6407,
+      "step": 38390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.56022481661955e-05,
+      "loss": 2.6803,
+      "step": 38400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555097317398454e-05,
+      "loss": 2.6337,
+      "step": 38410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5499712741187106e-05,
+      "loss": 2.4687,
+      "step": 38420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5448466884593685e-05,
+      "loss": 2.5581,
+      "step": 38430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5397235620990084e-05,
+      "loss": 2.5476,
+      "step": 38440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5346018967157165e-05,
+      "loss": 2.5992,
+      "step": 38450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5294816939871175e-05,
+      "loss": 2.508,
+      "step": 38460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.524362955590341e-05,
+      "loss": 2.5747,
+      "step": 38470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.519245683202049e-05,
+      "loss": 2.564,
+      "step": 38480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.514129878498424e-05,
+      "loss": 2.6326,
+      "step": 38490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.5090155431551584e-05,
+      "loss": 2.4538,
+      "step": 38500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.503902678847467e-05,
+      "loss": 2.6087,
+      "step": 38510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.498791287250091e-05,
+      "loss": 2.5598,
+      "step": 38520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.493681370037274e-05,
+      "loss": 2.5342,
+      "step": 38530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.488572928882789e-05,
+      "loss": 2.5022,
+      "step": 38540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.483465965459924e-05,
+      "loss": 2.4754,
+      "step": 38550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4783604814414745e-05,
+      "loss": 2.5718,
+      "step": 38560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.473766812097825e-05,
+      "loss": 2.5611,
+      "step": 38570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.468664143554605e-05,
+      "loss": 2.5326,
+      "step": 38580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.463562959264181e-05,
+      "loss": 2.5991,
+      "step": 38590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.45846326089747e-05,
+      "loss": 2.595,
+      "step": 38600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4533650501248856e-05,
+      "loss": 2.5596,
+      "step": 38610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4482683286163664e-05,
+      "loss": 2.5562,
+      "step": 38620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.44317309804136e-05,
+      "loss": 2.527,
+      "step": 38630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4380793600688194e-05,
+      "loss": 2.5473,
+      "step": 38640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.432987116367219e-05,
+      "loss": 2.5953,
+      "step": 38650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.427896368604539e-05,
+      "loss": 2.561,
+      "step": 38660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.422807118448264e-05,
+      "loss": 2.545,
+      "step": 38670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.417719367565397e-05,
+      "loss": 2.6195,
+      "step": 38680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4126331176224495e-05,
+      "loss": 2.5392,
+      "step": 38690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.4075483702854314e-05,
+      "loss": 2.4845,
+      "step": 38700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.402465127219869e-05,
+      "loss": 2.5936,
+      "step": 38710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3973833900907954e-05,
+      "loss": 2.5265,
+      "step": 38720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.392303160562748e-05,
+      "loss": 2.5953,
+      "step": 38730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.387224440299779e-05,
+      "loss": 2.5274,
+      "step": 38740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.382147230965424e-05,
+      "loss": 2.538,
+      "step": 38750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3770715342227485e-05,
+      "loss": 2.5551,
+      "step": 38760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.371997351734313e-05,
+      "loss": 2.5216,
+      "step": 38770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3669246851621755e-05,
+      "loss": 2.6477,
+      "step": 38780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.361853536167905e-05,
+      "loss": 2.5628,
+      "step": 38790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.356783906412577e-05,
+      "loss": 2.5367,
+      "step": 38800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.351715797556755e-05,
+      "loss": 2.5349,
+      "step": 38810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3466492112605236e-05,
+      "loss": 2.5072,
+      "step": 38820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3415841491834495e-05,
+      "loss": 2.5564,
+      "step": 38830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3365206129846124e-05,
+      "loss": 2.5445,
+      "step": 38840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3314586043225936e-05,
+      "loss": 2.5431,
+      "step": 38850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326398124855461e-05,
+      "loss": 2.6477,
+      "step": 38860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3213391762407935e-05,
+      "loss": 2.4726,
+      "step": 38870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3162817601356704e-05,
+      "loss": 2.5858,
+      "step": 38880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311225878196655e-05,
+      "loss": 2.6749,
+      "step": 38890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306171532079822e-05,
+      "loss": 2.6043,
+      "step": 38900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.301118723440736e-05,
+      "loss": 2.599,
+      "step": 38910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296067453934459e-05,
+      "loss": 2.5472,
+      "step": 38920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2910177252155567e-05,
+      "loss": 2.5572,
+      "step": 38930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2859695389380725e-05,
+      "loss": 2.5146,
+      "step": 38940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2809228967555604e-05,
+      "loss": 2.6815,
+      "step": 38950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2758778003210675e-05,
+      "loss": 2.5716,
+      "step": 38960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27083425128712e-05,
+      "loss": 2.5543,
+      "step": 38970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2657922513057556e-05,
+      "loss": 2.5059,
+      "step": 38980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2607518020284986e-05,
+      "loss": 2.6026,
+      "step": 38990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2557129051063535e-05,
+      "loss": 2.5651,
+      "step": 39000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.250675562189835e-05,
+      "loss": 2.5059,
+      "step": 39010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.245639774928939e-05,
+      "loss": 2.5109,
+      "step": 39020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.2406055449731496e-05,
+      "loss": 2.6496,
+      "step": 39030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.2355728739714436e-05,
+      "loss": 2.6393,
+      "step": 39040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.230541763572294e-05,
+      "loss": 2.5683,
+      "step": 39050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.2255122154236494e-05,
+      "loss": 2.5315,
+      "step": 39060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.22048423117296e-05,
+      "loss": 2.5936,
+      "step": 39070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.215457812467148e-05,
+      "loss": 2.5391,
+      "step": 39080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.210432960952634e-05,
+      "loss": 2.512,
+      "step": 39090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.205409678275335e-05,
+      "loss": 2.5955,
+      "step": 39100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.200387966080631e-05,
+      "loss": 2.6164,
+      "step": 39110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.195367826013402e-05,
+      "loss": 2.6131,
+      "step": 39120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.190349259718015e-05,
+      "loss": 2.5641,
+      "step": 39130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1853322688383074e-05,
+      "loss": 2.5694,
+      "step": 39140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.180316855017618e-05,
+      "loss": 2.6144,
+      "step": 39150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.175303019898754e-05,
+      "loss": 2.5567,
+      "step": 39160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.170791919439211e-05,
+      "loss": 2.5719,
+      "step": 39170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1657810883779255e-05,
+      "loss": 2.5176,
+      "step": 39180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.160771840779699e-05,
+      "loss": 2.5915,
+      "step": 39190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.15576417828533e-05,
+      "loss": 2.4794,
+      "step": 39200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.150758102535098e-05,
+      "loss": 2.5073,
+      "step": 39210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.145753615168747e-05,
+      "loss": 2.5807,
+      "step": 39220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.14075071782552e-05,
+      "loss": 2.6019,
+      "step": 39230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.135749412144135e-05,
+      "loss": 2.554,
+      "step": 39240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.130749699762776e-05,
+      "loss": 2.5807,
+      "step": 39250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1257515823191216e-05,
+      "loss": 2.5399,
+      "step": 39260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.120755061450313e-05,
+      "loss": 2.6573,
+      "step": 39270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1157601387929776e-05,
+      "loss": 2.4933,
+      "step": 39280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.1107668159832255e-05,
+      "loss": 2.5627,
+      "step": 39290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.105775094656622e-05,
+      "loss": 2.5683,
+      "step": 39300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.100784976448227e-05,
+      "loss": 2.5203,
+      "step": 39310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.095796462992575e-05,
+      "loss": 2.5224,
+      "step": 39320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.090809555923657e-05,
+      "loss": 2.6065,
+      "step": 39330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.085824256874954e-05,
+      "loss": 2.568,
+      "step": 39340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.080840567479421e-05,
+      "loss": 2.5081,
+      "step": 39350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.075858489369474e-05,
+      "loss": 2.5722,
+      "step": 39360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.070878024177008e-05,
+      "loss": 2.5219,
+      "step": 39370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.065899173533397e-05,
+      "loss": 2.5827,
+      "step": 39380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0609219390694694e-05,
+      "loss": 2.5009,
+      "step": 39390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0559463224155365e-05,
+      "loss": 2.4868,
+      "step": 39400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0509723252013774e-05,
+      "loss": 2.4914,
+      "step": 39410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0459999490562416e-05,
+      "loss": 2.5647,
+      "step": 39420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041029195608846e-05,
+      "loss": 2.5804,
+      "step": 39430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0360600664873715e-05,
+      "loss": 2.5962,
+      "step": 39440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031092563319474e-05,
+      "loss": 2.6384,
+      "step": 39450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026126687732278e-05,
+      "loss": 2.6219,
+      "step": 39460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.021162441352365e-05,
+      "loss": 2.5512,
+      "step": 39470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.01619982580579e-05,
+      "loss": 2.4981,
+      "step": 39480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011238842718078e-05,
+      "loss": 2.5452,
+      "step": 39490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0062794937142075e-05,
+      "loss": 2.5068,
+      "step": 39500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001321780418635e-05,
+      "loss": 2.5089,
+      "step": 39510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996365704455266e-05,
+      "loss": 2.539,
+      "step": 39520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9914112674474836e-05,
+      "loss": 2.5271,
+      "step": 39530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986458471018132e-05,
+      "loss": 2.5514,
+      "step": 39540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.981507316789505e-05,
+      "loss": 2.5272,
+      "step": 39550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976557806383376e-05,
+      "loss": 2.5713,
+      "step": 39560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971609941420973e-05,
+      "loss": 2.5701,
+      "step": 39570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966663723522973e-05,
+      "loss": 2.5218,
+      "step": 39580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.961719154309541e-05,
+      "loss": 2.5632,
+      "step": 39590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.9567762354002736e-05,
+      "loss": 2.5626,
+      "step": 39600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.951834968414242e-05,
+      "loss": 2.5008,
+      "step": 39610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.946895354969978e-05,
+      "loss": 2.5908,
+      "step": 39620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.941957396685459e-05,
+      "loss": 2.5044,
+      "step": 39630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.937021095178133e-05,
+      "loss": 2.6042,
+      "step": 39640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.932086452064903e-05,
+      "loss": 2.5444,
+      "step": 39650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.927153468962119e-05,
+      "loss": 2.5784,
+      "step": 39660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.9222221474855976e-05,
+      "loss": 2.6056,
+      "step": 39670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.917292489250613e-05,
+      "loss": 2.5466,
+      "step": 39680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.912364495871883e-05,
+      "loss": 2.5618,
+      "step": 39690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.907438168963588e-05,
+      "loss": 2.4998,
+      "step": 39700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.902513510139368e-05,
+      "loss": 2.6429,
+      "step": 39710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.897590521012301e-05,
+      "loss": 2.6028,
+      "step": 39720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.892669203194933e-05,
+      "loss": 2.4407,
+      "step": 39730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.887749558299252e-05,
+      "loss": 2.5261,
+      "step": 39740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8828315879367037e-05,
+      "loss": 2.5019,
+      "step": 39750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8779152937181896e-05,
+      "loss": 2.6259,
+      "step": 39760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.873000677254045e-05,
+      "loss": 2.5693,
+      "step": 39770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8680877401540794e-05,
+      "loss": 2.5027,
+      "step": 39780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8631764840275385e-05,
+      "loss": 2.5702,
+      "step": 39790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8582669104831135e-05,
+      "loss": 2.6094,
+      "step": 39800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.853359021128957e-05,
+      "loss": 2.5109,
+      "step": 39810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.848452817572655e-05,
+      "loss": 2.5858,
+      "step": 39820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8435483014212536e-05,
+      "loss": 2.6025,
+      "step": 39830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.838645474281245e-05,
+      "loss": 2.6243,
+      "step": 39840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.833744337758559e-05,
+      "loss": 2.6158,
+      "step": 39850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8288448934585804e-05,
+      "loss": 2.4409,
+      "step": 39860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8239471429861406e-05,
+      "loss": 2.4969,
+      "step": 39870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.8190510879455064e-05,
+      "loss": 2.5068,
+      "step": 39880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.814156729940398e-05,
+      "loss": 2.5942,
+      "step": 39890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.809264070573981e-05,
+      "loss": 2.5064,
+      "step": 39900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.804373111448853e-05,
+      "loss": 2.5615,
+      "step": 39910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7994838541670686e-05,
+      "loss": 2.5664,
+      "step": 39920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7945963003301196e-05,
+      "loss": 2.5089,
+      "step": 39930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7897104515389337e-05,
+      "loss": 2.6441,
+      "step": 39940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7848263093938874e-05,
+      "loss": 2.4369,
+      "step": 39950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7799438754947976e-05,
+      "loss": 2.5307,
+      "step": 39960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.77506315144092e-05,
+      "loss": 2.5907,
+      "step": 39970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.770184138830955e-05,
+      "loss": 2.5584,
+      "step": 39980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76530683926303e-05,
+      "loss": 2.5969,
+      "step": 39990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760431254334722e-05,
+      "loss": 2.5448,
+      "step": 40000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755557385643048e-05,
+      "loss": 2.6246,
+      "step": 40010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.750685234784451e-05,
+      "loss": 2.4884,
+      "step": 40020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7458148033548234e-05,
+      "loss": 2.5273,
+      "step": 40030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740946092949492e-05,
+      "loss": 2.5347,
+      "step": 40040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736079105163212e-05,
+      "loss": 2.4027,
+      "step": 40050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7312138415901865e-05,
+      "loss": 2.6107,
+      "step": 40060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726350303824039e-05,
+      "loss": 2.4698,
+      "step": 40070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721488493457841e-05,
+      "loss": 2.4835,
+      "step": 40080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7166284120840965e-05,
+      "loss": 2.5226,
+      "step": 40090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7117700612947326e-05,
+      "loss": 2.5353,
+      "step": 40100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.70691344268112e-05,
+      "loss": 2.5832,
+      "step": 40110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7020585578340626e-05,
+      "loss": 2.5831,
+      "step": 40120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.697205408343783e-05,
+      "loss": 2.5876,
+      "step": 40130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692353995799958e-05,
+      "loss": 2.5869,
+      "step": 40140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6875043217916734e-05,
+      "loss": 2.5864,
+      "step": 40150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.682656387907456e-05,
+      "loss": 2.5448,
+      "step": 40160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6778101957352675e-05,
+      "loss": 2.6073,
+      "step": 40170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6729657468624823e-05,
+      "loss": 2.5901,
+      "step": 40180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6681230428759206e-05,
+      "loss": 2.5684,
+      "step": 40190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.663282085361826e-05,
+      "loss": 2.5544,
+      "step": 40200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.658442875905865e-05,
+      "loss": 2.4368,
+      "step": 40210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6536054160931365e-05,
+      "loss": 2.5639,
+      "step": 40220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6487697075081696e-05,
+      "loss": 2.5146,
+      "step": 40230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6439357517349066e-05,
+      "loss": 2.5331,
+      "step": 40240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.639103550356731e-05,
+      "loss": 2.627,
+      "step": 40250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.634273104956445e-05,
+      "loss": 2.6719,
+      "step": 40260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.629444417116271e-05,
+      "loss": 2.5787,
+      "step": 40270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.624617488417863e-05,
+      "loss": 2.5577,
+      "step": 40280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6197923204423e-05,
+      "loss": 2.5863,
+      "step": 40290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6149689147700726e-05,
+      "loss": 2.5451,
+      "step": 40300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.6101472729811105e-05,
+      "loss": 2.5393,
+      "step": 40310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.605327396654745e-05,
+      "loss": 2.528,
+      "step": 40320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.600509287369752e-05,
+      "loss": 2.6322,
+      "step": 40330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5956929467043186e-05,
+      "loss": 2.5989,
+      "step": 40340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.590878376236044e-05,
+      "loss": 2.5601,
+      "step": 40350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.586065577541963e-05,
+      "loss": 2.4454,
+      "step": 40360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.581254552198515e-05,
+      "loss": 2.5779,
+      "step": 40370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.576445301781569e-05,
+      "loss": 2.5565,
+      "step": 40380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.571637827866414e-05,
+      "loss": 2.5499,
+      "step": 40390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5668321320277464e-05,
+      "loss": 2.6262,
+      "step": 40400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.562028215839687e-05,
+      "loss": 2.5089,
+      "step": 40410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.557226080875779e-05,
+      "loss": 2.5647,
+      "step": 40420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.552425728708969e-05,
+      "loss": 2.4954,
+      "step": 40430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.54762716091163e-05,
+      "loss": 2.6293,
+      "step": 40440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.54283037905555e-05,
+      "loss": 2.5226,
+      "step": 40450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.538035384711925e-05,
+      "loss": 2.5026,
+      "step": 40460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.533242179451371e-05,
+      "loss": 2.4615,
+      "step": 40470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.528450764843921e-05,
+      "loss": 2.582,
+      "step": 40480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5236611424590115e-05,
+      "loss": 2.6032,
+      "step": 40490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5188733138655004e-05,
+      "loss": 2.4671,
+      "step": 40500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5140872806316545e-05,
+      "loss": 2.546,
+      "step": 40510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.509303044325156e-05,
+      "loss": 2.6072,
+      "step": 40520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.5045206065130976e-05,
+      "loss": 2.5282,
+      "step": 40530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.499739968761975e-05,
+      "loss": 2.5678,
+      "step": 40540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4949611326377025e-05,
+      "loss": 2.5644,
+      "step": 40550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.490184099705609e-05,
+      "loss": 2.6047,
+      "step": 40560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485408871530417e-05,
+      "loss": 2.5434,
+      "step": 40570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4806354496762705e-05,
+      "loss": 2.5188,
+      "step": 40580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4758638357067226e-05,
+      "loss": 2.6543,
+      "step": 40590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4710940311847216e-05,
+      "loss": 2.5529,
+      "step": 40600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46632603767264e-05,
+      "loss": 2.4907,
+      "step": 40610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4615598567322415e-05,
+      "loss": 2.6039,
+      "step": 40620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4567954899247045e-05,
+      "loss": 2.607,
+      "step": 40630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4520329388106174e-05,
+      "loss": 2.5726,
+      "step": 40640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4472722049499615e-05,
+      "loss": 2.5993,
+      "step": 40650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442513289902131e-05,
+      "loss": 2.5562,
+      "step": 40660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437756195225928e-05,
+      "loss": 2.545,
+      "step": 40670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4330009224795424e-05,
+      "loss": 2.5415,
+      "step": 40680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4282474732205926e-05,
+      "loss": 2.5465,
+      "step": 40690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.423495849006074e-05,
+      "loss": 2.5231,
+      "step": 40700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418746051392401e-05,
+      "loss": 2.5101,
+      "step": 40710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413998081935384e-05,
+      "loss": 2.5986,
+      "step": 40720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409251942190231e-05,
+      "loss": 2.497,
+      "step": 40730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.404507633711557e-05,
+      "loss": 2.5457,
+      "step": 40740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.39976515805338e-05,
+      "loss": 2.6033,
+      "step": 40750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3950245167691e-05,
+      "loss": 2.624,
+      "step": 40760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.390285711411538e-05,
+      "loss": 2.512,
+      "step": 40770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.385548743532904e-05,
+      "loss": 2.4653,
+      "step": 40780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3808136146848e-05,
+      "loss": 2.5625,
+      "step": 40790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.376080326418235e-05,
+      "loss": 2.5819,
+      "step": 40800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.371348880283616e-05,
+      "loss": 2.571,
+      "step": 40810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.366619277830736e-05,
+      "loss": 2.4933,
+      "step": 40820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.361891520608793e-05,
+      "loss": 2.5741,
+      "step": 40830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3571656101663816e-05,
+      "loss": 2.5211,
+      "step": 40840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3524415480514826e-05,
+      "loss": 2.5517,
+      "step": 40850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.347719335811482e-05,
+      "loss": 2.5616,
+      "step": 40860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3429989749931454e-05,
+      "loss": 2.5276,
+      "step": 40870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.338280467142652e-05,
+      "loss": 2.4806,
+      "step": 40880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.333563813805563e-05,
+      "loss": 2.4652,
+      "step": 40890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.328849016526826e-05,
+      "loss": 2.475,
+      "step": 40900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.324136076850793e-05,
+      "loss": 2.4632,
+      "step": 40910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3194249963211954e-05,
+      "loss": 2.4914,
+      "step": 40920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3147157764811644e-05,
+      "loss": 2.5548,
+      "step": 40930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.310008418873225e-05,
+      "loss": 2.5067,
+      "step": 40940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.3053029250392774e-05,
+      "loss": 2.5514,
+      "step": 40950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.300599296520624e-05,
+      "loss": 2.5586,
+      "step": 40960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2958975348579575e-05,
+      "loss": 2.518,
+      "step": 40970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.291197641591346e-05,
+      "loss": 2.5352,
+      "step": 40980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2864996182602576e-05,
+      "loss": 2.4868,
+      "step": 40990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.281803466403548e-05,
+      "loss": 2.4602,
+      "step": 41000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.277109187559448e-05,
+      "loss": 2.4807,
+      "step": 41010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.272416783265587e-05,
+      "loss": 2.4934,
+      "step": 41020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.26772625505898e-05,
+      "loss": 2.4467,
+      "step": 41030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.263037604476017e-05,
+      "loss": 2.4611,
+      "step": 41040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.258350833052481e-05,
+      "loss": 2.4866,
+      "step": 41050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.253665942323541e-05,
+      "loss": 2.5404,
+      "step": 41060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.248982933823746e-05,
+      "loss": 2.5712,
+      "step": 41070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2443018090870334e-05,
+      "loss": 2.598,
+      "step": 41080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2396225696467116e-05,
+      "loss": 2.5122,
+      "step": 41090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.234945217035485e-05,
+      "loss": 2.5433,
+      "step": 41100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2302697527854374e-05,
+      "loss": 2.5835,
+      "step": 41110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225596178428024e-05,
+      "loss": 2.4946,
+      "step": 41120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2209244954940916e-05,
+      "loss": 2.5683,
+      "step": 41130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216254705513868e-05,
+      "loss": 2.5374,
+      "step": 41140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21158681001695e-05,
+      "loss": 2.5115,
+      "step": 41150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.206920810532328e-05,
+      "loss": 2.5109,
+      "step": 41160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202256708588357e-05,
+      "loss": 2.5624,
+      "step": 41170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.197594505712782e-05,
+      "loss": 2.5716,
+      "step": 41180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192934203432724e-05,
+      "loss": 2.4938,
+      "step": 41190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1882758032746724e-05,
+      "loss": 2.5062,
+      "step": 41200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1836193067645046e-05,
+      "loss": 2.5256,
+      "step": 41210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178964715427474e-05,
+      "loss": 2.5038,
+      "step": 41220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.174312030788193e-05,
+      "loss": 2.6479,
+      "step": 41230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169661254370679e-05,
+      "loss": 2.5062,
+      "step": 41240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1650123876982974e-05,
+      "loss": 2.5452,
+      "step": 41250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160365432293801e-05,
+      "loss": 2.522,
+      "step": 41260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155720389679318e-05,
+      "loss": 2.4831,
+      "step": 41270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151077261376341e-05,
+      "loss": 2.4916,
+      "step": 41280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.146436048905741e-05,
+      "loss": 2.6026,
+      "step": 41290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1417967537877657e-05,
+      "loss": 2.6227,
+      "step": 41300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1371593775420245e-05,
+      "loss": 2.5314,
+      "step": 41310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.132523921687507e-05,
+      "loss": 2.5105,
+      "step": 41320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.127890387742575e-05,
+      "loss": 2.5371,
+      "step": 41330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.123258777224948e-05,
+      "loss": 2.5588,
+      "step": 41340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.118629091651728e-05,
+      "loss": 2.5319,
+      "step": 41350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.1140013325393864e-05,
+      "loss": 2.5258,
+      "step": 41360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.1093755014037524e-05,
+      "loss": 2.5627,
+      "step": 41370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.104751599760036e-05,
+      "loss": 2.5694,
+      "step": 41380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.100129629122812e-05,
+      "loss": 2.5746,
+      "step": 41390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.095509591006015e-05,
+      "loss": 2.5987,
+      "step": 41400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0908914869229583e-05,
+      "loss": 2.6081,
+      "step": 41410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0862753183863066e-05,
+      "loss": 2.4699,
+      "step": 41420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.08166108690811e-05,
+      "loss": 2.5514,
+      "step": 41430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.077048793999774e-05,
+      "loss": 2.4442,
+      "step": 41440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.072438441172061e-05,
+      "loss": 2.5195,
+      "step": 41450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.067830029935115e-05,
+      "loss": 2.5111,
+      "step": 41460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.063223561798426e-05,
+      "loss": 2.5682,
+      "step": 41470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.058619038270862e-05,
+      "loss": 2.5892,
+      "step": 41480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.054016460860651e-05,
+      "loss": 2.4836,
+      "step": 41490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.049415831075375e-05,
+      "loss": 2.5167,
+      "step": 41500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.044817150421988e-05,
+      "loss": 2.3881,
+      "step": 41510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.040220420406804e-05,
+      "loss": 2.5733,
+      "step": 41520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0356256425354886e-05,
+      "loss": 2.5212,
+      "step": 41530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.031032818313081e-05,
+      "loss": 2.6043,
+      "step": 41540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0264419492439753e-05,
+      "loss": 2.4941,
+      "step": 41550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0218530368319196e-05,
+      "loss": 2.486,
+      "step": 41560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.017266082580029e-05,
+      "loss": 2.5237,
+      "step": 41570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0126810879907784e-05,
+      "loss": 2.5369,
+      "step": 41580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.008098054565989e-05,
+      "loss": 2.5052,
+      "step": 41590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.00351698380685e-05,
+      "loss": 2.632,
+      "step": 41600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.998937877213907e-05,
+      "loss": 2.5443,
+      "step": 41610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.994360736287058e-05,
+      "loss": 2.5009,
+      "step": 41620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.989785562525565e-05,
+      "loss": 2.4294,
+      "step": 41630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.985212357428032e-05,
+      "loss": 2.5329,
+      "step": 41640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.980641122492429e-05,
+      "loss": 2.553,
+      "step": 41650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.976071859216083e-05,
+      "loss": 2.4872,
+      "step": 41660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.971504569095662e-05,
+      "loss": 2.5605,
+      "step": 41670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.966939253627199e-05,
+      "loss": 2.4913,
+      "step": 41680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9623759143060815e-05,
+      "loss": 2.5338,
+      "step": 41690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957814552627037e-05,
+      "loss": 2.5516,
+      "step": 41700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9532551700841614e-05,
+      "loss": 2.6489,
+      "step": 41710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.948697768170886e-05,
+      "loss": 2.5624,
+      "step": 41720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9441423483800054e-05,
+      "loss": 2.5125,
+      "step": 41730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9395889122036655e-05,
+      "loss": 2.5168,
+      "step": 41740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935037461133352e-05,
+      "loss": 2.5346,
+      "step": 41750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9304879966599074e-05,
+      "loss": 2.5768,
+      "step": 41760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925940520273527e-05,
+      "loss": 2.5788,
+      "step": 41770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9213950334637406e-05,
+      "loss": 2.5487,
+      "step": 41780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91685153771945e-05,
+      "loss": 2.5972,
+      "step": 41790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912310034528881e-05,
+      "loss": 2.55,
+      "step": 41800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9077705253796205e-05,
+      "loss": 2.4776,
+      "step": 41810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.903233011758599e-05,
+      "loss": 2.5435,
+      "step": 41820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898697495152089e-05,
+      "loss": 2.5646,
+      "step": 41830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8941639770457164e-05,
+      "loss": 2.5006,
+      "step": 41840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88963245892445e-05,
+      "loss": 2.5832,
+      "step": 41850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.885102942272597e-05,
+      "loss": 2.5452,
+      "step": 41860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880575428573817e-05,
+      "loss": 2.5623,
+      "step": 41870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8760499193111145e-05,
+      "loss": 2.4972,
+      "step": 41880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8715264159668265e-05,
+      "loss": 2.4585,
+      "step": 41890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.867004920022644e-05,
+      "loss": 2.524,
+      "step": 41900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.8624854329596015e-05,
+      "loss": 2.5401,
+      "step": 41910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.857967956258062e-05,
+      "loss": 2.6167,
+      "step": 41920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.853452491397743e-05,
+      "loss": 2.5592,
+      "step": 41930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.848939039857701e-05,
+      "loss": 2.5763,
+      "step": 41940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.844427603116325e-05,
+      "loss": 2.5027,
+      "step": 41950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.839918182651357e-05,
+      "loss": 2.5501,
+      "step": 41960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.83541077993986e-05,
+      "loss": 2.5798,
+      "step": 41970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.830905396458259e-05,
+      "loss": 2.5428,
+      "step": 41980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.826402033682304e-05,
+      "loss": 2.5047,
+      "step": 41990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.8219006930870795e-05,
+      "loss": 2.5302,
+      "step": 42000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.8174013761470205e-05,
+      "loss": 2.5816,
+      "step": 42010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.812904084335884e-05,
+      "loss": 2.5802,
+      "step": 42020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.808408819126776e-05,
+      "loss": 2.5459,
+      "step": 42030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.803915581992138e-05,
+      "loss": 2.582,
+      "step": 42040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.799424374403736e-05,
+      "loss": 2.5856,
+      "step": 42050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.794935197832681e-05,
+      "loss": 2.5668,
+      "step": 42060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.790448053749421e-05,
+      "loss": 2.4686,
+      "step": 42070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.785962943623727e-05,
+      "loss": 2.5662,
+      "step": 42080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.781479868924713e-05,
+      "loss": 2.5257,
+      "step": 42090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.776998831120828e-05,
+      "loss": 2.5115,
+      "step": 42100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.772519831679842e-05,
+      "loss": 2.5581,
+      "step": 42110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.768042872068869e-05,
+      "loss": 2.5704,
+      "step": 42120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.763567953754352e-05,
+      "loss": 2.5948,
+      "step": 42130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.75909507820206e-05,
+      "loss": 2.509,
+      "step": 42140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7550712379776254e-05,
+      "loss": 2.5495,
+      "step": 42150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.750602247709373e-05,
+      "loss": 2.5668,
+      "step": 42160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.746135304450304e-05,
+      "loss": 2.5256,
+      "step": 42170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7416704096635835e-05,
+      "loss": 2.5821,
+      "step": 42180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.737207564811694e-05,
+      "loss": 2.6688,
+      "step": 42190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.732746771356457e-05,
+      "loss": 2.5864,
+      "step": 42200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.728288030759022e-05,
+      "loss": 2.5649,
+      "step": 42210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7238313444798564e-05,
+      "loss": 2.5278,
+      "step": 42220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.719376713978766e-05,
+      "loss": 2.5238,
+      "step": 42230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.714924140714881e-05,
+      "loss": 2.5411,
+      "step": 42240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.710473626146647e-05,
+      "loss": 2.5614,
+      "step": 42250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7060251717318485e-05,
+      "loss": 2.5058,
+      "step": 42260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701578778927595e-05,
+      "loss": 2.5624,
+      "step": 42270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6971344491903024e-05,
+      "loss": 2.5472,
+      "step": 42280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.69269218397574e-05,
+      "loss": 2.6085,
+      "step": 42290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6882519847389755e-05,
+      "loss": 2.5912,
+      "step": 42300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.683813852934411e-05,
+      "loss": 2.5217,
+      "step": 42310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6793777900157767e-05,
+      "loss": 2.5269,
+      "step": 42320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.674943797436109e-05,
+      "loss": 2.5583,
+      "step": 42330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.670511876647779e-05,
+      "loss": 2.505,
+      "step": 42340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66608202910248e-05,
+      "loss": 2.5887,
+      "step": 42350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.661654256251216e-05,
+      "loss": 2.6308,
+      "step": 42360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6572285595443224e-05,
+      "loss": 2.4946,
+      "step": 42370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.652804940431445e-05,
+      "loss": 2.5464,
+      "step": 42380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.648383400361555e-05,
+      "loss": 2.4897,
+      "step": 42390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.643963940782945e-05,
+      "loss": 2.5281,
+      "step": 42400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.639546563143216e-05,
+      "loss": 2.5476,
+      "step": 42410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.635131268889297e-05,
+      "loss": 2.6139,
+      "step": 42420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.630718059467435e-05,
+      "loss": 2.5136,
+      "step": 42430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6263069363231814e-05,
+      "loss": 2.4971,
+      "step": 42440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.621897900901419e-05,
+      "loss": 2.5659,
+      "step": 42450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6174909546463377e-05,
+      "loss": 2.5226,
+      "step": 42460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.613086099001448e-05,
+      "loss": 2.5552,
+      "step": 42470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.608683335409577e-05,
+      "loss": 2.4933,
+      "step": 42480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.604282665312857e-05,
+      "loss": 2.4952,
+      "step": 42490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.599884090152743e-05,
+      "loss": 2.5154,
+      "step": 42500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.595487611370006e-05,
+      "loss": 2.5272,
+      "step": 42510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.591093230404718e-05,
+      "loss": 2.5415,
+      "step": 42520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5867009486962764e-05,
+      "loss": 2.5929,
+      "step": 42530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.582310767683389e-05,
+      "loss": 2.5775,
+      "step": 42540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.577922688804066e-05,
+      "loss": 2.5516,
+      "step": 42550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.573536713495641e-05,
+      "loss": 2.5205,
+      "step": 42560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5691528431947554e-05,
+      "loss": 2.5192,
+      "step": 42570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.564771079337351e-05,
+      "loss": 2.6706,
+      "step": 42580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5603914233586945e-05,
+      "loss": 2.5791,
+      "step": 42590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.556013876693357e-05,
+      "loss": 2.4661,
+      "step": 42600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5516384407752114e-05,
+      "loss": 2.5346,
+      "step": 42610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.547265117037449e-05,
+      "loss": 2.5704,
+      "step": 42620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.542893906912562e-05,
+      "loss": 2.4463,
+      "step": 42630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5385248118323556e-05,
+      "loss": 2.5976,
+      "step": 42640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5341578332279404e-05,
+      "loss": 2.6044,
+      "step": 42650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.529792972529732e-05,
+      "loss": 2.5707,
+      "step": 42660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.525430231167459e-05,
+      "loss": 2.5339,
+      "step": 42670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5210696105701425e-05,
+      "loss": 2.5022,
+      "step": 42680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.516711112166121e-05,
+      "loss": 2.5007,
+      "step": 42690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.5123547373830367e-05,
+      "loss": 2.5258,
+      "step": 42700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.508000487647828e-05,
+      "loss": 2.5698,
+      "step": 42710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.503648364386745e-05,
+      "loss": 2.4969,
+      "step": 42720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4992983690253426e-05,
+      "loss": 2.6348,
+      "step": 42730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4949505029884666e-05,
+      "loss": 2.5165,
+      "step": 42740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4906047677002795e-05,
+      "loss": 2.5093,
+      "step": 42750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.486261164584241e-05,
+      "loss": 2.5023,
+      "step": 42760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.481919695063108e-05,
+      "loss": 2.5034,
+      "step": 42770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.477580360558942e-05,
+      "loss": 2.5827,
+      "step": 42780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.473243162493111e-05,
+      "loss": 2.5877,
+      "step": 42790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.468908102286269e-05,
+      "loss": 2.6437,
+      "step": 42800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4645751813583825e-05,
+      "loss": 2.5462,
+      "step": 42810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.460244401128716e-05,
+      "loss": 2.5516,
+      "step": 42820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.455915763015819e-05,
+      "loss": 2.4767,
+      "step": 42830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.451589268437564e-05,
+      "loss": 2.559,
+      "step": 42840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4472649188110974e-05,
+      "loss": 2.5323,
+      "step": 42850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.442942715552876e-05,
+      "loss": 2.4977,
+      "step": 42860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4386226600786555e-05,
+      "loss": 2.5363,
+      "step": 42870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.434304753803473e-05,
+      "loss": 2.5781,
+      "step": 42880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.429988998141679e-05,
+      "loss": 2.5571,
+      "step": 42890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.425675394506914e-05,
+      "loss": 2.4743,
+      "step": 42900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4213639443121045e-05,
+      "loss": 2.5738,
+      "step": 42910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.417054648969488e-05,
+      "loss": 2.5596,
+      "step": 42920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.412747509890577e-05,
+      "loss": 2.5569,
+      "step": 42930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4084425284861954e-05,
+      "loss": 2.444,
+      "step": 42940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4041397061664535e-05,
+      "loss": 2.5074,
+      "step": 42950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.399839044340748e-05,
+      "loss": 2.5952,
+      "step": 42960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3955405444177766e-05,
+      "loss": 2.5129,
+      "step": 42970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3912442078055305e-05,
+      "loss": 2.5912,
+      "step": 42980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3869500359112796e-05,
+      "loss": 2.5266,
+      "step": 42990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.382658030141598e-05,
+      "loss": 2.5444,
+      "step": 43000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3783681919023424e-05,
+      "loss": 2.4989,
+      "step": 43010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374080522598666e-05,
+      "loss": 2.4919,
+      "step": 43020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.369795023635008e-05,
+      "loss": 2.4714,
+      "step": 43030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3655116964150925e-05,
+      "loss": 2.5057,
+      "step": 43040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.361230542341937e-05,
+      "loss": 2.5288,
+      "step": 43050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.356951562817852e-05,
+      "loss": 2.5934,
+      "step": 43060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.352674759244423e-05,
+      "loss": 2.5628,
+      "step": 43070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3484001330225325e-05,
+      "loss": 2.5307,
+      "step": 43080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3441276855523506e-05,
+      "loss": 2.544,
+      "step": 43090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.339857418233325e-05,
+      "loss": 2.5073,
+      "step": 43100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3355893324641983e-05,
+      "loss": 2.4149,
+      "step": 43110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3313234296429965e-05,
+      "loss": 2.5447,
+      "step": 43120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.327059711167023e-05,
+      "loss": 2.5465,
+      "step": 43130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.322798178432877e-05,
+      "loss": 2.4919,
+      "step": 43140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.3185388328364366e-05,
+      "loss": 2.6218,
+      "step": 43150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.314281675772858e-05,
+      "loss": 2.5094,
+      "step": 43160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.310026708636592e-05,
+      "loss": 2.5572,
+      "step": 43170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.305773932821361e-05,
+      "loss": 2.5184,
+      "step": 43180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.301523349720178e-05,
+      "loss": 2.5419,
+      "step": 43190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.297274960725332e-05,
+      "loss": 2.5569,
+      "step": 43200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2930287672283975e-05,
+      "loss": 2.6031,
+      "step": 43210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.288784770620231e-05,
+      "loss": 2.4799,
+      "step": 43220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2845429722909595e-05,
+      "loss": 2.4468,
+      "step": 43230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2803033736300006e-05,
+      "loss": 2.5641,
+      "step": 43240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2760659760260514e-05,
+      "loss": 2.5145,
+      "step": 43250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2718307808670765e-05,
+      "loss": 2.5873,
+      "step": 43260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.267597789540332e-05,
+      "loss": 2.3454,
+      "step": 43270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.263367003432348e-05,
+      "loss": 2.517,
+      "step": 43280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.259138423928927e-05,
+      "loss": 2.4778,
+      "step": 43290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.254912052415154e-05,
+      "loss": 2.4511,
+      "step": 43300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.250687890275396e-05,
+      "loss": 2.6253,
+      "step": 43310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.246465938893282e-05,
+      "loss": 2.5494,
+      "step": 43320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.242246199651727e-05,
+      "loss": 2.4524,
+      "step": 43330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2380286739329246e-05,
+      "loss": 2.4994,
+      "step": 43340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.233813363118331e-05,
+      "loss": 2.5236,
+      "step": 43350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.229600268588687e-05,
+      "loss": 2.5285,
+      "step": 43360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.225389391724007e-05,
+      "loss": 2.5124,
+      "step": 43370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.221180733903569e-05,
+      "loss": 2.6231,
+      "step": 43380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.216974296505943e-05,
+      "loss": 2.5323,
+      "step": 43390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.212770080908951e-05,
+      "loss": 2.4653,
+      "step": 43400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2085680884896996e-05,
+      "loss": 2.5722,
+      "step": 43410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204368320624568e-05,
+      "loss": 2.592,
+      "step": 43420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2001707786891956e-05,
+      "loss": 2.5583,
+      "step": 43430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1959754640585024e-05,
+      "loss": 2.5034,
+      "step": 43440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1917823781066825e-05,
+      "loss": 2.4903,
+      "step": 43450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.187591522207185e-05,
+      "loss": 2.5095,
+      "step": 43460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.183402897732745e-05,
+      "loss": 2.5087,
+      "step": 43470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.179216506055351e-05,
+      "loss": 2.4659,
+      "step": 43480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.175032348546274e-05,
+      "loss": 2.5296,
+      "step": 43490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17085042657605e-05,
+      "loss": 2.4899,
+      "step": 43500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1666707415144724e-05,
+      "loss": 2.601,
+      "step": 43510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1624932947306155e-05,
+      "loss": 2.588,
+      "step": 43520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.158318087592814e-05,
+      "loss": 2.6258,
+      "step": 43530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1541451214686676e-05,
+      "loss": 2.5327,
+      "step": 43540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.149974397725044e-05,
+      "loss": 2.4026,
+      "step": 43550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.145805917728079e-05,
+      "loss": 2.4658,
+      "step": 43560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.141639682843167e-05,
+      "loss": 2.5559,
+      "step": 43570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.137475694434978e-05,
+      "loss": 2.5742,
+      "step": 43580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13331395386743e-05,
+      "loss": 2.5556,
+      "step": 43590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1291544625037196e-05,
+      "loss": 2.5352,
+      "step": 43600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.124997221706302e-05,
+      "loss": 2.6444,
+      "step": 43610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.120842232836888e-05,
+      "loss": 2.5285,
+      "step": 43620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.11668949725646e-05,
+      "loss": 2.5036,
+      "step": 43630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.112539016325262e-05,
+      "loss": 2.5238,
+      "step": 43640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.108390791402791e-05,
+      "loss": 2.4695,
+      "step": 43650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.104244823847814e-05,
+      "loss": 2.5193,
+      "step": 43660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.100101115018357e-05,
+      "loss": 2.5707,
+      "step": 43670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0959596662716996e-05,
+      "loss": 2.4961,
+      "step": 43680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.091820478964388e-05,
+      "loss": 2.5322,
+      "step": 43690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0876835544522284e-05,
+      "loss": 2.4882,
+      "step": 43700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.083548894090279e-05,
+      "loss": 2.5493,
+      "step": 43710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.079416499232865e-05,
+      "loss": 2.5787,
+      "step": 43720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0752863712335566e-05,
+      "loss": 2.4588,
+      "step": 43730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.071158511445194e-05,
+      "loss": 2.4853,
+      "step": 43740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.067032921219876e-05,
+      "loss": 2.6028,
+      "step": 43750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.062909601908944e-05,
+      "loss": 2.5888,
+      "step": 43760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0587885548630116e-05,
+      "loss": 2.4672,
+      "step": 43770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.054669781431932e-05,
+      "loss": 2.5515,
+      "step": 43780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0505532829648242e-05,
+      "loss": 2.4605,
+      "step": 43790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0464390608100655e-05,
+      "loss": 2.5269,
+      "step": 43800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0423271163152743e-05,
+      "loss": 2.4804,
+      "step": 43810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0382174508273342e-05,
+      "loss": 2.6437,
+      "step": 43820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0341100656923805e-05,
+      "loss": 2.5205,
+      "step": 43830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0300049622557936e-05,
+      "loss": 2.5718,
+      "step": 43840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.025902141862217e-05,
+      "loss": 2.5009,
+      "step": 43850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0218016058555443e-05,
+      "loss": 2.5628,
+      "step": 43860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0177033555789125e-05,
+      "loss": 2.5749,
+      "step": 43870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0136073923747177e-05,
+      "loss": 2.533,
+      "step": 43880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.00951371758461e-05,
+      "loss": 2.5292,
+      "step": 43890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.005422332549478e-05,
+      "loss": 2.53,
+      "step": 43900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.0013332386094707e-05,
+      "loss": 2.6632,
+      "step": 43910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9972464371039855e-05,
+      "loss": 2.548,
+      "step": 43920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9931619293716584e-05,
+      "loss": 2.4937,
+      "step": 43930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9890797167503937e-05,
+      "loss": 2.5871,
+      "step": 43940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.984999800577324e-05,
+      "loss": 2.542,
+      "step": 43950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.98092218218884e-05,
+      "loss": 2.4845,
+      "step": 43960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9768468629205825e-05,
+      "loss": 2.5843,
+      "step": 43970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.972773844107429e-05,
+      "loss": 2.5331,
+      "step": 43980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.968703127083511e-05,
+      "loss": 2.5705,
+      "step": 43990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9646347131822074e-05,
+      "loss": 2.519,
+      "step": 44000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9605686037361324e-05,
+      "loss": 2.5676,
+      "step": 44010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.956504800077161e-05,
+      "loss": 2.586,
+      "step": 44020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9524433035363973e-05,
+      "loss": 2.5514,
+      "step": 44030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9483841154441993e-05,
+      "loss": 2.4804,
+      "step": 44040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.944327237130169e-05,
+      "loss": 2.566,
+      "step": 44050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.940272669923144e-05,
+      "loss": 2.5655,
+      "step": 44060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936220415151214e-05,
+      "loss": 2.471,
+      "step": 44070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.932170474141709e-05,
+      "loss": 2.5666,
+      "step": 44080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.928122848221193e-05,
+      "loss": 2.5233,
+      "step": 44090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9240775387154818e-05,
+      "loss": 2.6073,
+      "step": 44100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9200345469496294e-05,
+      "loss": 2.4669,
+      "step": 44110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.91599387424793e-05,
+      "loss": 2.3816,
+      "step": 44120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.91195552193392e-05,
+      "loss": 2.5423,
+      "step": 44130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9079194913303688e-05,
+      "loss": 2.4817,
+      "step": 44140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9038857837592938e-05,
+      "loss": 2.576,
+      "step": 44150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8998544005419504e-05,
+      "loss": 2.5877,
+      "step": 44160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8958253429988237e-05,
+      "loss": 2.6272,
+      "step": 44170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8917986124496488e-05,
+      "loss": 2.5159,
+      "step": 44180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8877742102133953e-05,
+      "loss": 2.5218,
+      "step": 44190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8837521376082622e-05,
+      "loss": 2.5569,
+      "step": 44200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8797323959516943e-05,
+      "loss": 2.5387,
+      "step": 44210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8757149865603738e-05,
+      "loss": 2.5285,
+      "step": 44220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8716999107502095e-05,
+      "loss": 2.6267,
+      "step": 44230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.867687169836355e-05,
+      "loss": 2.5713,
+      "step": 44240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8636767651331975e-05,
+      "loss": 2.5927,
+      "step": 44250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.859668697954352e-05,
+      "loss": 2.5991,
+      "step": 44260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.855662969612679e-05,
+      "loss": 2.5914,
+      "step": 44270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8516595814202594e-05,
+      "loss": 2.4942,
+      "step": 44280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8476585346884187e-05,
+      "loss": 2.5438,
+      "step": 44290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8436598307277173e-05,
+      "loss": 2.5113,
+      "step": 44300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8396634708479365e-05,
+      "loss": 2.6933,
+      "step": 44310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8356694563580997e-05,
+      "loss": 2.497,
+      "step": 44320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8316777885664548e-05,
+      "loss": 2.491,
+      "step": 44330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.827688468780485e-05,
+      "loss": 2.4912,
+      "step": 44340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8237014983069097e-05,
+      "loss": 2.481,
+      "step": 44350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8197168784516658e-05,
+      "loss": 2.5018,
+      "step": 44360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.815734610519931e-05,
+      "loss": 2.5099,
+      "step": 44370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8117546958161124e-05,
+      "loss": 2.6286,
+      "step": 44380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8077771356438364e-05,
+      "loss": 2.5131,
+      "step": 44390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.8038019313059695e-05,
+      "loss": 2.5061,
+      "step": 44400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7998290841046025e-05,
+      "loss": 2.5672,
+      "step": 44410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.79585859534105e-05,
+      "loss": 2.5632,
+      "step": 44420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7918904663158583e-05,
+      "loss": 2.4887,
+      "step": 44430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7879246983288044e-05,
+      "loss": 2.5168,
+      "step": 44440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7839612926788827e-05,
+      "loss": 2.4818,
+      "step": 44450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7800002506643196e-05,
+      "loss": 2.5514,
+      "step": 44460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7760415735825707e-05,
+      "loss": 2.4487,
+      "step": 44470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.772085262730303e-05,
+      "loss": 2.4403,
+      "step": 44480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7681313194034307e-05,
+      "loss": 2.5543,
+      "step": 44490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7641797448970707e-05,
+      "loss": 2.5828,
+      "step": 44500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7602305405055762e-05,
+      "loss": 2.4763,
+      "step": 44510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7562837075225234e-05,
+      "loss": 2.509,
+      "step": 44520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.752339247240705e-05,
+      "loss": 2.5391,
+      "step": 44530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.748397160952143e-05,
+      "loss": 2.4682,
+      "step": 44540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7444574499480825e-05,
+      "loss": 2.5257,
+      "step": 44550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7405201155189818e-05,
+      "loss": 2.6108,
+      "step": 44560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7365851589545334e-05,
+      "loss": 2.5159,
+      "step": 44570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7326525815436377e-05,
+      "loss": 2.5781,
+      "step": 44580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.728722384574426e-05,
+      "loss": 2.5066,
+      "step": 44590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7247945693342468e-05,
+      "loss": 2.5795,
+      "step": 44600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7208691371096662e-05,
+      "loss": 2.4539,
+      "step": 44610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7169460891864706e-05,
+      "loss": 2.4463,
+      "step": 44620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7130254268496713e-05,
+      "loss": 2.5823,
+      "step": 44630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7091071513834854e-05,
+      "loss": 2.5011,
+      "step": 44640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7051912640713604e-05,
+      "loss": 2.5509,
+      "step": 44650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7012777661959576e-05,
+      "loss": 2.6694,
+      "step": 44660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.697366659039153e-05,
+      "loss": 2.5975,
+      "step": 44670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6934579438820472e-05,
+      "loss": 2.5635,
+      "step": 44680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6895516220049433e-05,
+      "loss": 2.5176,
+      "step": 44690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.685647694687373e-05,
+      "loss": 2.5055,
+      "step": 44700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6817461632080832e-05,
+      "loss": 2.4566,
+      "step": 44710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6778470288450242e-05,
+      "loss": 2.5184,
+      "step": 44720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6739502928753747e-05,
+      "loss": 2.5605,
+      "step": 44730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6700559565755234e-05,
+      "loss": 2.5384,
+      "step": 44740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.666164021221067e-05,
+      "loss": 2.4972,
+      "step": 44750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6622744880868243e-05,
+      "loss": 2.4705,
+      "step": 44760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6583873584468256e-05,
+      "loss": 2.4942,
+      "step": 44770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6545026335743062e-05,
+      "loss": 2.5046,
+      "step": 44780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.650620314741723e-05,
+      "loss": 2.4633,
+      "step": 44790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6467404032207443e-05,
+      "loss": 2.5678,
+      "step": 44800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6428629002822404e-05,
+      "loss": 2.5394,
+      "step": 44810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6389878071963048e-05,
+      "loss": 2.5427,
+      "step": 44820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6351151252322305e-05,
+      "loss": 2.5705,
+      "step": 44830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.631244855658527e-05,
+      "loss": 2.4733,
+      "step": 44840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6273769997429208e-05,
+      "loss": 2.506,
+      "step": 44850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6235115587523306e-05,
+      "loss": 2.5642,
+      "step": 44860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6196485339529008e-05,
+      "loss": 2.5215,
+      "step": 44870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6157879266099695e-05,
+      "loss": 2.562,
+      "step": 44880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6119297379880935e-05,
+      "loss": 2.5564,
+      "step": 44890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.608073969351037e-05,
+      "loss": 2.639,
+      "step": 44900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6042206219617638e-05,
+      "loss": 2.528,
+      "step": 44910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.600369697082452e-05,
+      "loss": 2.4309,
+      "step": 44920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.596521195974486e-05,
+      "loss": 2.5085,
+      "step": 44930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.592675119898449e-05,
+      "loss": 2.4627,
+      "step": 44940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.588831470114137e-05,
+      "loss": 2.5464,
+      "step": 44950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5849902478805533e-05,
+      "loss": 2.5323,
+      "step": 44960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5811514544558946e-05,
+      "loss": 2.5573,
+      "step": 44970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.577315091097574e-05,
+      "loss": 2.6269,
+      "step": 44980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5734811590622054e-05,
+      "loss": 2.5644,
+      "step": 44990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5696496596055987e-05,
+      "loss": 2.5012,
+      "step": 45000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5658205939827773e-05,
+      "loss": 2.5699,
+      "step": 45010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.561993963447965e-05,
+      "loss": 2.5635,
+      "step": 45020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.558169769254579e-05,
+      "loss": 2.527,
+      "step": 45030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5543480126552566e-05,
+      "loss": 2.5896,
+      "step": 45040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.550528694901817e-05,
+      "loss": 2.5352,
+      "step": 45050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.546711817245292e-05,
+      "loss": 2.4838,
+      "step": 45060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5428973809359147e-05,
+      "loss": 2.5489,
+      "step": 45070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.539085387223109e-05,
+      "loss": 2.5443,
+      "step": 45080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5352758373555073e-05,
+      "loss": 2.557,
+      "step": 45090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531468732580945e-05,
+      "loss": 2.4796,
+      "step": 45100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.527664074146442e-05,
+      "loss": 2.4994,
+      "step": 45110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5238618632982326e-05,
+      "loss": 2.5368,
+      "step": 45120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.520062101281736e-05,
+      "loss": 2.4717,
+      "step": 45130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.51626478934158e-05,
+      "loss": 2.5297,
+      "step": 45140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.512469928721588e-05,
+      "loss": 2.4936,
+      "step": 45150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5086775206647718e-05,
+      "loss": 2.6056,
+      "step": 45160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5048875664133497e-05,
+      "loss": 2.5431,
+      "step": 45170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5011000672087347e-05,
+      "loss": 2.4713,
+      "step": 45180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.497315024291529e-05,
+      "loss": 2.5066,
+      "step": 45190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493532438901538e-05,
+      "loss": 2.4532,
+      "step": 45200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4897523122777568e-05,
+      "loss": 2.527,
+      "step": 45210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.48597464565838e-05,
+      "loss": 2.5489,
+      "step": 45220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4821994402807945e-05,
+      "loss": 2.5762,
+      "step": 45230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.478426697381576e-05,
+      "loss": 2.5533,
+      "step": 45240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4746564181965004e-05,
+      "loss": 2.5348,
+      "step": 45250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.470888603960536e-05,
+      "loss": 2.5625,
+      "step": 45260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4671232559078374e-05,
+      "loss": 2.4244,
+      "step": 45270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4633603752717572e-05,
+      "loss": 2.5248,
+      "step": 45280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4595999632848432e-05,
+      "loss": 2.502,
+      "step": 45290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4558420211788225e-05,
+      "loss": 2.52,
+      "step": 45300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4520865501846234e-05,
+      "loss": 2.5576,
+      "step": 45310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4483335515323646e-05,
+      "loss": 2.5337,
+      "step": 45320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.444583026451348e-05,
+      "loss": 2.538,
+      "step": 45330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44083497617007e-05,
+      "loss": 2.5528,
+      "step": 45340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4370894019162204e-05,
+      "loss": 2.6211,
+      "step": 45350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.433346304916667e-05,
+      "loss": 2.5674,
+      "step": 45360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.429605686397479e-05,
+      "loss": 2.4276,
+      "step": 45370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4258675475839e-05,
+      "loss": 2.5568,
+      "step": 45380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4221318897003707e-05,
+      "loss": 2.52,
+      "step": 45390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4183987139705246e-05,
+      "loss": 2.6002,
+      "step": 45400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4146680216171657e-05,
+      "loss": 2.5582,
+      "step": 45410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4109398138622998e-05,
+      "loss": 2.4476,
+      "step": 45420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.407586552223956e-05,
+      "loss": 2.5274,
+      "step": 45430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4038630685699127e-05,
+      "loss": 2.5803,
+      "step": 45440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.400142073053555e-05,
+      "loss": 2.5087,
+      "step": 45450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3964235668937097e-05,
+      "loss": 2.5109,
+      "step": 45460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3927075513083775e-05,
+      "loss": 2.4902,
+      "step": 45470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.388994027514757e-05,
+      "loss": 2.5645,
+      "step": 45480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.385282996729219e-05,
+      "loss": 2.5272,
+      "step": 45490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3815744601673252e-05,
+      "loss": 2.5292,
+      "step": 45500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3778684190438204e-05,
+      "loss": 2.5316,
+      "step": 45510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.374164874572622e-05,
+      "loss": 2.6085,
+      "step": 45520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3704638279668457e-05,
+      "loss": 2.5362,
+      "step": 45530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3667652804387795e-05,
+      "loss": 2.5266,
+      "step": 45540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3630692331998892e-05,
+      "loss": 2.6989,
+      "step": 45550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.359375687460832e-05,
+      "loss": 2.5881,
+      "step": 45560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.355684644431434e-05,
+      "loss": 2.4495,
+      "step": 45570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3519961053207108e-05,
+      "loss": 2.5047,
+      "step": 45580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3483100713368555e-05,
+      "loss": 2.5724,
+      "step": 45590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.344626543687236e-05,
+      "loss": 2.506,
+      "step": 45600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3409455235784062e-05,
+      "loss": 2.5196,
+      "step": 45610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3372670122160957e-05,
+      "loss": 2.546,
+      "step": 45620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3335910108052095e-05,
+      "loss": 2.485,
+      "step": 45630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3299175205498335e-05,
+      "loss": 2.5376,
+      "step": 45640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3262465426532342e-05,
+      "loss": 2.4808,
+      "step": 45650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3225780783178463e-05,
+      "loss": 2.5199,
+      "step": 45660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3189121287452888e-05,
+      "loss": 2.507,
+      "step": 45670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.315248695136357e-05,
+      "loss": 2.5011,
+      "step": 45680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3115877786910146e-05,
+      "loss": 2.4669,
+      "step": 45690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.307929380608409e-05,
+      "loss": 2.5593,
+      "step": 45700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.304273502086859e-05,
+      "loss": 2.4753,
+      "step": 45710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3006201443238585e-05,
+      "loss": 2.4593,
+      "step": 45720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2969693085160803e-05,
+      "loss": 2.5658,
+      "step": 45730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2933209958593594e-05,
+      "loss": 2.5008,
+      "step": 45740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.289675207548715e-05,
+      "loss": 2.5614,
+      "step": 45750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2860319447783385e-05,
+      "loss": 2.4891,
+      "step": 45760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2823912087415867e-05,
+      "loss": 2.5366,
+      "step": 45770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2787530006309986e-05,
+      "loss": 2.4776,
+      "step": 45780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2751173216382772e-05,
+      "loss": 2.4869,
+      "step": 45790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.271484172954299e-05,
+      "loss": 2.5097,
+      "step": 45800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2678535557691184e-05,
+      "loss": 2.5385,
+      "step": 45810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.264225471271949e-05,
+      "loss": 2.5438,
+      "step": 45820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.260599920651183e-05,
+      "loss": 2.5339,
+      "step": 45830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2569769050943845e-05,
+      "loss": 2.5376,
+      "step": 45840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2533564257882754e-05,
+      "loss": 2.503,
+      "step": 45850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2497384839187608e-05,
+      "loss": 2.4333,
+      "step": 45860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2461230806709076e-05,
+      "loss": 2.4657,
+      "step": 45870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2425102172289502e-05,
+      "loss": 2.5299,
+      "step": 45880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.238899894776293e-05,
+      "loss": 2.4558,
+      "step": 45890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2352921144955084e-05,
+      "loss": 2.574,
+      "step": 45900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2316868775683377e-05,
+      "loss": 2.4309,
+      "step": 45910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2280841851756874e-05,
+      "loss": 2.4905,
+      "step": 45920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.224484038497626e-05,
+      "loss": 2.4555,
+      "step": 45930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2208864387133964e-05,
+      "loss": 2.4848,
+      "step": 45940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2172913870014043e-05,
+      "loss": 2.4735,
+      "step": 45950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.213698884539215e-05,
+      "loss": 2.5167,
+      "step": 45960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2101089325035664e-05,
+      "loss": 2.4934,
+      "step": 45970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.2065215320703602e-05,
+      "loss": 2.5275,
+      "step": 45980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.202936684414656e-05,
+      "loss": 2.5345,
+      "step": 45990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1993543907106828e-05,
+      "loss": 2.6249,
+      "step": 46000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1957746521318358e-05,
+      "loss": 2.5658,
+      "step": 46010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.192197469850663e-05,
+      "loss": 2.5489,
+      "step": 46020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.188622845038888e-05,
+      "loss": 2.5934,
+      "step": 46030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1850507788673823e-05,
+      "loss": 2.4515,
+      "step": 46040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1814812725061917e-05,
+      "loss": 2.5006,
+      "step": 46050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1779143271245216e-05,
+      "loss": 2.538,
+      "step": 46060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1743499438907255e-05,
+      "loss": 2.4532,
+      "step": 46070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1707881239723383e-05,
+      "loss": 2.4935,
+      "step": 46080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1672288685360432e-05,
+      "loss": 2.5318,
+      "step": 46090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.163672178747682e-05,
+      "loss": 2.5548,
+      "step": 46100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.160118055772262e-05,
+      "loss": 2.495,
+      "step": 46110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.156566500773943e-05,
+      "loss": 2.5218,
+      "step": 46120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1530175149160504e-05,
+      "loss": 2.5489,
+      "step": 46130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1494710993610677e-05,
+      "loss": 2.4966,
+      "step": 46140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1459272552706277e-05,
+      "loss": 2.4817,
+      "step": 46150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1423859838055317e-05,
+      "loss": 2.5772,
+      "step": 46160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1388472861257358e-05,
+      "loss": 2.5615,
+      "step": 46170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1353111633903456e-05,
+      "loss": 2.633,
+      "step": 46180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.131777616757632e-05,
+      "loss": 2.5249,
+      "step": 46190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1282466473850206e-05,
+      "loss": 2.5948,
+      "step": 46200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.124718256429087e-05,
+      "loss": 2.5282,
+      "step": 46210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.121192445045569e-05,
+      "loss": 2.5079,
+      "step": 46220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1176692143893572e-05,
+      "loss": 2.5068,
+      "step": 46230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1141485656144945e-05,
+      "loss": 2.4683,
+      "step": 46240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110630499874181e-05,
+      "loss": 2.5872,
+      "step": 46250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.107115018320769e-05,
+      "loss": 2.53,
+      "step": 46260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1036021221057666e-05,
+      "loss": 2.5642,
+      "step": 46270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1000918123798365e-05,
+      "loss": 2.4543,
+      "step": 46280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0965840902927846e-05,
+      "loss": 2.4975,
+      "step": 46290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0930789569935795e-05,
+      "loss": 2.4953,
+      "step": 46300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.08957641363034e-05,
+      "loss": 2.5204,
+      "step": 46310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0860764613503304e-05,
+      "loss": 2.5749,
+      "step": 46320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0825791012999752e-05,
+      "loss": 2.4214,
+      "step": 46330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0790843346248412e-05,
+      "loss": 2.555,
+      "step": 46340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0755921624696515e-05,
+      "loss": 2.5707,
+      "step": 46350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0721025859782804e-05,
+      "loss": 2.4581,
+      "step": 46360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0686156062937446e-05,
+      "loss": 2.4691,
+      "step": 46370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0651312245582166e-05,
+      "loss": 2.5005,
+      "step": 46380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0616494419130193e-05,
+      "loss": 2.5222,
+      "step": 46390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0581702594986163e-05,
+      "loss": 2.5501,
+      "step": 46400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0546936784546266e-05,
+      "loss": 2.3928,
+      "step": 46410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0512196999198173e-05,
+      "loss": 2.5047,
+      "step": 46420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0477483250320952e-05,
+      "loss": 2.5531,
+      "step": 46430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.044279554928522e-05,
+      "loss": 2.5472,
+      "step": 46440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0408133907453053e-05,
+      "loss": 2.5784,
+      "step": 46450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0373498336177977e-05,
+      "loss": 2.5483,
+      "step": 46460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0338888846804983e-05,
+      "loss": 2.5909,
+      "step": 46470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.030430545067048e-05,
+      "loss": 2.5005,
+      "step": 46480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0269748159102388e-05,
+      "loss": 2.4953,
+      "step": 46490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0235216983420067e-05,
+      "loss": 2.4482,
+      "step": 46500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020071193493427e-05,
+      "loss": 2.5316,
+      "step": 46510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0166233024947255e-05,
+      "loss": 2.5154,
+      "step": 46520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.013178026475271e-05,
+      "loss": 2.5296,
+      "step": 46530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0097353665635698e-05,
+      "loss": 2.4914,
+      "step": 46540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.0062953238872784e-05,
+      "loss": 2.5747,
+      "step": 46550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.002857899573195e-05,
+      "loss": 2.4536,
+      "step": 46560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9994230947472526e-05,
+      "loss": 2.5603,
+      "step": 46570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.99599091053454e-05,
+      "loss": 2.5688,
+      "step": 46580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9925613480592708e-05,
+      "loss": 2.5443,
+      "step": 46590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.989134408444813e-05,
+      "loss": 2.505,
+      "step": 46600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9857100928136728e-05,
+      "loss": 2.511,
+      "step": 46610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.982288402287489e-05,
+      "loss": 2.543,
+      "step": 46620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9788693379870536e-05,
+      "loss": 2.5477,
+      "step": 46630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9754529010322907e-05,
+      "loss": 2.5597,
+      "step": 46640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9720390925422605e-05,
+      "loss": 2.5216,
+      "step": 46650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9686279136351705e-05,
+      "loss": 2.5307,
+      "step": 46660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.965219365428359e-05,
+      "loss": 2.5298,
+      "step": 46670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9618134490383068e-05,
+      "loss": 2.5805,
+      "step": 46680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9584101655806363e-05,
+      "loss": 2.5044,
+      "step": 46690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.955009516170099e-05,
+      "loss": 2.4197,
+      "step": 46700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.951611501920588e-05,
+      "loss": 2.549,
+      "step": 46710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9482161239451367e-05,
+      "loss": 2.4986,
+      "step": 46720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.944823383355907e-05,
+      "loss": 2.6847,
+      "step": 46730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9414332812642043e-05,
+      "loss": 2.5001,
+      "step": 46740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9380458187804674e-05,
+      "loss": 2.4404,
+      "step": 46750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.934660997014267e-05,
+      "loss": 2.4275,
+      "step": 46760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9312788170743136e-05,
+      "loss": 2.5151,
+      "step": 46770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9278992800684515e-05,
+      "loss": 2.5229,
+      "step": 46780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9245223871036556e-05,
+      "loss": 2.473,
+      "step": 46790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9211481392860375e-05,
+      "loss": 2.5356,
+      "step": 46800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9177765377208446e-05,
+      "loss": 2.546,
+      "step": 46810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9144075835124542e-05,
+      "loss": 2.4378,
+      "step": 46820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9110412777643793e-05,
+      "loss": 2.422,
+      "step": 46830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.90767762157926e-05,
+      "loss": 2.4726,
+      "step": 46840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9043166160588734e-05,
+      "loss": 2.4896,
+      "step": 46850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.900958262304131e-05,
+      "loss": 2.5855,
+      "step": 46860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8976025614150648e-05,
+      "loss": 2.5565,
+      "step": 46870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8942495144908513e-05,
+      "loss": 2.4976,
+      "step": 46880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8908991226297845e-05,
+      "loss": 2.4944,
+      "step": 46890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8875513869293005e-05,
+      "loss": 2.5116,
+      "step": 46900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8842063084859605e-05,
+      "loss": 2.4379,
+      "step": 46910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.880863888395451e-05,
+      "loss": 2.5218,
+      "step": 46920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8775241277525946e-05,
+      "loss": 2.5143,
+      "step": 46930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8741870276513417e-05,
+      "loss": 2.4415,
+      "step": 46940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8708525891847652e-05,
+      "loss": 2.4104,
+      "step": 46950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8675208134450728e-05,
+      "loss": 2.573,
+      "step": 46960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.864191701523601e-05,
+      "loss": 2.5403,
+      "step": 46970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8608652545108042e-05,
+      "loss": 2.5791,
+      "step": 46980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8575414734962736e-05,
+      "loss": 2.5238,
+      "step": 46990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8542203595687235e-05,
+      "loss": 2.5465,
+      "step": 47000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8509019138159944e-05,
+      "loss": 2.5392,
+      "step": 47010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8475861373250568e-05,
+      "loss": 2.4853,
+      "step": 47020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8442730311819968e-05,
+      "loss": 2.5202,
+      "step": 47030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8409625964720368e-05,
+      "loss": 2.5706,
+      "step": 47040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8376548342795207e-05,
+      "loss": 2.49,
+      "step": 47050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.834349745687911e-05,
+      "loss": 2.4857,
+      "step": 47060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8310473317798037e-05,
+      "loss": 2.4578,
+      "step": 47070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.827747593636916e-05,
+      "loss": 2.509,
+      "step": 47080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8244505323400806e-05,
+      "loss": 2.5089,
+      "step": 47090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.821156148969265e-05,
+      "loss": 2.5338,
+      "step": 47100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8178644446035552e-05,
+      "loss": 2.4574,
+      "step": 47110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.814575420321154e-05,
+      "loss": 2.5145,
+      "step": 47120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8112890771993973e-05,
+      "loss": 2.5978,
+      "step": 47130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8080054163147318e-05,
+      "loss": 2.4177,
+      "step": 47140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.804724438742732e-05,
+      "loss": 2.5606,
+      "step": 47150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8014461455580955e-05,
+      "loss": 2.5167,
+      "step": 47160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.798170537834628e-05,
+      "loss": 2.4608,
+      "step": 47170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7948976166452746e-05,
+      "loss": 2.4138,
+      "step": 47180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.791627383062089e-05,
+      "loss": 2.4895,
+      "step": 47190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7883598381562417e-05,
+      "loss": 2.5031,
+      "step": 47200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7850949829980324e-05,
+      "loss": 2.5539,
+      "step": 47210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7818328186568678e-05,
+      "loss": 2.4521,
+      "step": 47220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7785733462012832e-05,
+      "loss": 2.4766,
+      "step": 47230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7753165666989314e-05,
+      "loss": 2.5322,
+      "step": 47240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7720624812165754e-05,
+      "loss": 2.5655,
+      "step": 47250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7688110908201027e-05,
+      "loss": 2.4969,
+      "step": 47260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7655623965745195e-05,
+      "loss": 2.4679,
+      "step": 47270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7623163995439384e-05,
+      "loss": 2.4941,
+      "step": 47280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7590731007916006e-05,
+      "loss": 2.5208,
+      "step": 47290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7558325013798603e-05,
+      "loss": 2.557,
+      "step": 47300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7525946023701788e-05,
+      "loss": 2.4897,
+      "step": 47310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7493594048231443e-05,
+      "loss": 2.5451,
+      "step": 47320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.746126909798458e-05,
+      "loss": 2.5248,
+      "step": 47330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.742897118354927e-05,
+      "loss": 2.5054,
+      "step": 47340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7396700315504834e-05,
+      "loss": 2.5184,
+      "step": 47350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.736445650442168e-05,
+      "loss": 2.5579,
+      "step": 47360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7332239760861368e-05,
+      "loss": 2.4726,
+      "step": 47370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.730005009537663e-05,
+      "loss": 2.5119,
+      "step": 47380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7267887518511218e-05,
+      "loss": 2.6059,
+      "step": 47390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7235752040800114e-05,
+      "loss": 2.4903,
+      "step": 47400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.720364367276942e-05,
+      "loss": 2.4533,
+      "step": 47410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7171562424936273e-05,
+      "loss": 2.5459,
+      "step": 47420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7139508307809038e-05,
+      "loss": 2.5567,
+      "step": 47430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.71074813318871e-05,
+      "loss": 2.5515,
+      "step": 47440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7075481507660985e-05,
+      "loss": 2.4951,
+      "step": 47450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7043508845612387e-05,
+      "loss": 2.4426,
+      "step": 47460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7011563356213987e-05,
+      "loss": 2.5018,
+      "step": 47470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.697964504992965e-05,
+      "loss": 2.522,
+      "step": 47480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6947753937214352e-05,
+      "loss": 2.5495,
+      "step": 47490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6915890028514047e-05,
+      "loss": 2.5229,
+      "step": 47500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6884053334265904e-05,
+      "loss": 2.5123,
+      "step": 47510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6852243864898155e-05,
+      "loss": 2.5311,
+      "step": 47520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.682046163083002e-05,
+      "loss": 2.5249,
+      "step": 47530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6788706642471896e-05,
+      "loss": 2.463,
+      "step": 47540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.675697891022524e-05,
+      "loss": 2.4819,
+      "step": 47550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6725278444482552e-05,
+      "loss": 2.4526,
+      "step": 47560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6693605255627455e-05,
+      "loss": 2.5096,
+      "step": 47570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.666195935403453e-05,
+      "loss": 2.4901,
+      "step": 47580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6630340750069516e-05,
+      "loss": 2.4723,
+      "step": 47590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6598749454089225e-05,
+      "loss": 2.578,
+      "step": 47600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6567185476441415e-05,
+      "loss": 2.542,
+      "step": 47610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6535648827465e-05,
+      "loss": 2.4866,
+      "step": 47620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6504139517489914e-05,
+      "loss": 2.4678,
+      "step": 47630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.647265755683709e-05,
+      "loss": 2.4989,
+      "step": 47640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6441202955818557e-05,
+      "loss": 2.5396,
+      "step": 47650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6409775724737397e-05,
+      "loss": 2.4865,
+      "step": 47660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6378375873887643e-05,
+      "loss": 2.5655,
+      "step": 47670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.634700341355446e-05,
+      "loss": 2.5263,
+      "step": 47680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.631565835401394e-05,
+      "loss": 2.4884,
+      "step": 47690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6284340705533273e-05,
+      "loss": 2.5125,
+      "step": 47700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6253050478370692e-05,
+      "loss": 2.4956,
+      "step": 47710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.622491272762242e-05,
+      "loss": 2.4779,
+      "step": 47720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6193674629193244e-05,
+      "loss": 2.5453,
+      "step": 47730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6162463981780096e-05,
+      "loss": 2.5055,
+      "step": 47740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.613128079560606e-05,
+      "loss": 2.5322,
+      "step": 47750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.610012508088532e-05,
+      "loss": 2.5337,
+      "step": 47760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6068996847823036e-05,
+      "loss": 2.4255,
+      "step": 47770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6037896106615312e-05,
+      "loss": 2.4172,
+      "step": 47780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.6006822867449322e-05,
+      "loss": 2.4624,
+      "step": 47790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.597577714050321e-05,
+      "loss": 2.473,
+      "step": 47800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.594475893594606e-05,
+      "loss": 2.5448,
+      "step": 47810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.591376826393801e-05,
+      "loss": 2.6252,
+      "step": 47820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5882805134630117e-05,
+      "loss": 2.4294,
+      "step": 47830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5851869558164478e-05,
+      "loss": 2.579,
+      "step": 47840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5820961544674118e-05,
+      "loss": 2.5872,
+      "step": 47850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5790081104283062e-05,
+      "loss": 2.5012,
+      "step": 47860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5759228247106307e-05,
+      "loss": 2.5293,
+      "step": 47870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5728402983249756e-05,
+      "loss": 2.5059,
+      "step": 47880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.569760532281034e-05,
+      "loss": 2.5112,
+      "step": 47890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5666835275875957e-05,
+      "loss": 2.5532,
+      "step": 47900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5636092852525376e-05,
+      "loss": 2.459,
+      "step": 47910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5605378062828392e-05,
+      "loss": 2.4963,
+      "step": 47920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5574690916845756e-05,
+      "loss": 2.5077,
+      "step": 47930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.554403142462908e-05,
+      "loss": 2.5688,
+      "step": 47940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5513399596221013e-05,
+      "loss": 2.5913,
+      "step": 47950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5482795441655118e-05,
+      "loss": 2.4515,
+      "step": 47960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5452218970955833e-05,
+      "loss": 2.5822,
+      "step": 47970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5421670194138582e-05,
+      "loss": 2.5282,
+      "step": 47980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5391149121209757e-05,
+      "loss": 2.6064,
+      "step": 47990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5360655762166575e-05,
+      "loss": 2.5353,
+      "step": 48000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5330190126997246e-05,
+      "loss": 2.4505,
+      "step": 48010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5299752225680898e-05,
+      "loss": 2.5093,
+      "step": 48020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.52693420681875e-05,
+      "loss": 2.5375,
+      "step": 48030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5238959664478081e-05,
+      "loss": 2.4839,
+      "step": 48040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5208605024504418e-05,
+      "loss": 2.5704,
+      "step": 48050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5178278158209291e-05,
+      "loss": 2.4686,
+      "step": 48060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5147979075526374e-05,
+      "loss": 2.4342,
+      "step": 48070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5117707786380186e-05,
+      "loss": 2.5121,
+      "step": 48080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5087464300686206e-05,
+      "loss": 2.5265,
+      "step": 48090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5057248628350784e-05,
+      "loss": 2.4519,
+      "step": 48100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5027060779271119e-05,
+      "loss": 2.5077,
+      "step": 48110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4996900763335386e-05,
+      "loss": 2.5558,
+      "step": 48120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4966768590422532e-05,
+      "loss": 2.4916,
+      "step": 48130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4936664270402479e-05,
+      "loss": 2.5275,
+      "step": 48140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4906587813135986e-05,
+      "loss": 2.5077,
+      "step": 48150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4876539228474673e-05,
+      "loss": 2.5265,
+      "step": 48160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4846518526261043e-05,
+      "loss": 2.599,
+      "step": 48170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.481652571632851e-05,
+      "loss": 2.573,
+      "step": 48180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4786560808501248e-05,
+      "loss": 2.5243,
+      "step": 48190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4756623812594383e-05,
+      "loss": 2.5333,
+      "step": 48200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4726714738413905e-05,
+      "loss": 2.4746,
+      "step": 48210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.469683359575652e-05,
+      "loss": 2.4178,
+      "step": 48220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4666980394410013e-05,
+      "loss": 2.4978,
+      "step": 48230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4637155144152803e-05,
+      "loss": 2.5781,
+      "step": 48240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4607357854754267e-05,
+      "loss": 2.4974,
+      "step": 48250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4577588535974629e-05,
+      "loss": 2.4957,
+      "step": 48260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4547847197564856e-05,
+      "loss": 2.535,
+      "step": 48270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4518133849266857e-05,
+      "loss": 2.5558,
+      "step": 48280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4488448500813334e-05,
+      "loss": 2.5067,
+      "step": 48290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4458791161927786e-05,
+      "loss": 2.5502,
+      "step": 48300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4429161842324568e-05,
+      "loss": 2.4576,
+      "step": 48310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439956055170889e-05,
+      "loss": 2.4175,
+      "step": 48320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4369987299776689e-05,
+      "loss": 2.5984,
+      "step": 48330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4340442096214802e-05,
+      "loss": 2.5909,
+      "step": 48340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4310924950700877e-05,
+      "loss": 2.334,
+      "step": 48350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4281435872903293e-05,
+      "loss": 2.5335,
+      "step": 48360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4251974872481333e-05,
+      "loss": 2.5722,
+      "step": 48370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.422254195908499e-05,
+      "loss": 2.5308,
+      "step": 48380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4193137142355128e-05,
+      "loss": 2.5538,
+      "step": 48390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.416376043192339e-05,
+      "loss": 2.5012,
+      "step": 48400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4134411837412188e-05,
+      "loss": 2.4503,
+      "step": 48410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4105091368434787e-05,
+      "loss": 2.5163,
+      "step": 48420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.407579903459515e-05,
+      "loss": 2.5544,
+      "step": 48430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4046534845488069e-05,
+      "loss": 2.4609,
+      "step": 48440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4017298810699164e-05,
+      "loss": 2.4856,
+      "step": 48450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3988090939804743e-05,
+      "loss": 2.5197,
+      "step": 48460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.395891124237193e-05,
+      "loss": 2.4552,
+      "step": 48470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.392975972795867e-05,
+      "loss": 2.4796,
+      "step": 48480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3900636406113576e-05,
+      "loss": 2.4535,
+      "step": 48490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.38715412863761e-05,
+      "loss": 2.5788,
+      "step": 48500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3842474378276472e-05,
+      "loss": 2.6014,
+      "step": 48510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3813435691335586e-05,
+      "loss": 2.4451,
+      "step": 48520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.378442523506518e-05,
+      "loss": 2.5426,
+      "step": 48530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3755443018967739e-05,
+      "loss": 2.5106,
+      "step": 48540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3726489052536429e-05,
+      "loss": 2.453,
+      "step": 48550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3697563345255238e-05,
+      "loss": 2.5374,
+      "step": 48560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3668665906598887e-05,
+      "loss": 2.4687,
+      "step": 48570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363979674603275e-05,
+      "loss": 2.5745,
+      "step": 48580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3610955873013109e-05,
+      "loss": 2.5078,
+      "step": 48590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3582143296986793e-05,
+      "loss": 2.3621,
+      "step": 48600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3553359027391488e-05,
+      "loss": 2.5373,
+      "step": 48610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3524603073655584e-05,
+      "loss": 2.5549,
+      "step": 48620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3495875445198148e-05,
+      "loss": 2.5551,
+      "step": 48630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3467176151428996e-05,
+      "loss": 2.5041,
+      "step": 48640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3438505201748719e-05,
+      "loss": 2.5193,
+      "step": 48650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3409862605548517e-05,
+      "loss": 2.5457,
+      "step": 48660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3381248372210409e-05,
+      "loss": 2.4669,
+      "step": 48670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3352662511107017e-05,
+      "loss": 2.39,
+      "step": 48680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3324105031601763e-05,
+      "loss": 2.51,
+      "step": 48690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3295575943048755e-05,
+      "loss": 2.3754,
+      "step": 48700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3267075254792739e-05,
+      "loss": 2.4566,
+      "step": 48710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3238602976169211e-05,
+      "loss": 2.4808,
+      "step": 48720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3210159116504395e-05,
+      "loss": 2.5548,
+      "step": 48730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31817436851151e-05,
+      "loss": 2.5511,
+      "step": 48740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3153356691308916e-05,
+      "loss": 2.4522,
+      "step": 48750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3124998144384115e-05,
+      "loss": 2.4619,
+      "step": 48760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3096668053629558e-05,
+      "loss": 2.4771,
+      "step": 48770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3068366428324941e-05,
+      "loss": 2.5325,
+      "step": 48780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3040093277740472e-05,
+      "loss": 2.5046,
+      "step": 48790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3011848611137134e-05,
+      "loss": 2.5942,
+      "step": 48800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.298363243776658e-05,
+      "loss": 2.5246,
+      "step": 48810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.295544476687106e-05,
+      "loss": 2.5173,
+      "step": 48820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2927285607683549e-05,
+      "loss": 2.5275,
+      "step": 48830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2899154969427684e-05,
+      "loss": 2.4934,
+      "step": 48840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.28710528613177e-05,
+      "loss": 2.4827,
+      "step": 48850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2842979292558544e-05,
+      "loss": 2.5016,
+      "step": 48860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2814934272345836e-05,
+      "loss": 2.4563,
+      "step": 48870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.278691780986575e-05,
+      "loss": 2.4724,
+      "step": 48880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2758929914295192e-05,
+      "loss": 2.4508,
+      "step": 48890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2730970594801694e-05,
+      "loss": 2.4539,
+      "step": 48900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2703039860543387e-05,
+      "loss": 2.6043,
+      "step": 48910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2675137720669095e-05,
+      "loss": 2.4805,
+      "step": 48920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.264726418431822e-05,
+      "loss": 2.4925,
+      "step": 48930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2619419260620835e-05,
+      "loss": 2.4963,
+      "step": 48940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2591602958697635e-05,
+      "loss": 2.4975,
+      "step": 48950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.256381528765993e-05,
+      "loss": 2.4697,
+      "step": 48960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2536056256609675e-05,
+      "loss": 2.4954,
+      "step": 48970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2508325874639393e-05,
+      "loss": 2.4859,
+      "step": 48980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2480624150832266e-05,
+      "loss": 2.5084,
+      "step": 48990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2452951094262099e-05,
+      "loss": 2.416,
+      "step": 49000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2425306713993246e-05,
+      "loss": 2.4958,
+      "step": 49010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2397691019080737e-05,
+      "loss": 2.4938,
+      "step": 49020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2370104018570184e-05,
+      "loss": 2.4997,
+      "step": 49030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2342545721497756e-05,
+      "loss": 2.5336,
+      "step": 49040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2315016136890279e-05,
+      "loss": 2.5574,
+      "step": 49050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2287515273765172e-05,
+      "loss": 2.5439,
+      "step": 49060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2260043141130384e-05,
+      "loss": 2.5115,
+      "step": 49070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2232599747984509e-05,
+      "loss": 2.4628,
+      "step": 49080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2205185103316753e-05,
+      "loss": 2.509,
+      "step": 49090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.21777992161068e-05,
+      "loss": 2.4924,
+      "step": 49100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2150442095325032e-05,
+      "loss": 2.4964,
+      "step": 49110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2123113749932346e-05,
+      "loss": 2.5316,
+      "step": 49120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2095814188880184e-05,
+      "loss": 2.4942,
+      "step": 49130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2068543421110667e-05,
+      "loss": 2.4776,
+      "step": 49140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2041301455556364e-05,
+      "loss": 2.4401,
+      "step": 49150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2014088301140492e-05,
+      "loss": 2.5714,
+      "step": 49160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1986903966776808e-05,
+      "loss": 2.5306,
+      "step": 49170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1962462714353707e-05,
+      "loss": 2.5619,
+      "step": 49180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1935333162612705e-05,
+      "loss": 2.5539,
+      "step": 49190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190823245672038e-05,
+      "loss": 2.5177,
+      "step": 49200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.188116060555362e-05,
+      "loss": 2.5042,
+      "step": 49210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1854117617979899e-05,
+      "loss": 2.4575,
+      "step": 49220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.182710350285725e-05,
+      "loss": 2.4445,
+      "step": 49230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18001182690342e-05,
+      "loss": 2.4278,
+      "step": 49240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1773161925349852e-05,
+      "loss": 2.6486,
+      "step": 49250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1746234480633844e-05,
+      "loss": 2.5331,
+      "step": 49260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1719335943706344e-05,
+      "loss": 2.4793,
+      "step": 49270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.169246632337807e-05,
+      "loss": 2.5264,
+      "step": 49280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1665625628450194e-05,
+      "loss": 2.4717,
+      "step": 49290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1638813867714503e-05,
+      "loss": 2.5463,
+      "step": 49300
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 57782,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 50,
+  "total_flos": 1.671676728138793e+19,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}