diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,35365 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 58904,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.828054298642534e-08,
+      "loss": 2.6177,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.656108597285068e-08,
+      "loss": 2.5825,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.484162895927602e-08,
+      "loss": 2.5859,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1312217194570136e-07,
+      "loss": 2.4777,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.414027149321267e-07,
+      "loss": 2.2447,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6968325791855205e-07,
+      "loss": 2.1244,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9796380090497739e-07,
+      "loss": 1.8977,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2624434389140273e-07,
+      "loss": 1.8579,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5452488687782807e-07,
+      "loss": 1.6832,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.828054298642534e-07,
+      "loss": 1.6264,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1108597285067875e-07,
+      "loss": 1.5952,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.393665158371041e-07,
+      "loss": 1.5513,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6764705882352943e-07,
+      "loss": 1.611,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9592760180995477e-07,
+      "loss": 1.5037,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2420814479638017e-07,
+      "loss": 1.5756,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5248868778280546e-07,
+      "loss": 1.659,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.807692307692308e-07,
+      "loss": 1.5801,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.090497737556561e-07,
+      "loss": 1.5929,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.373303167420815e-07,
+      "loss": 1.5594,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.656108597285068e-07,
+      "loss": 1.4325,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.938914027149321e-07,
+      "loss": 1.5875,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.221719457013575e-07,
+      "loss": 1.5789,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.504524886877829e-07,
+      "loss": 1.5334,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.787330316742082e-07,
+      "loss": 1.5489,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.070135746606336e-07,
+      "loss": 1.5689,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.352941176470589e-07,
+      "loss": 1.5163,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.635746606334843e-07,
+      "loss": 1.5486,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.918552036199095e-07,
+      "loss": 1.5851,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.201357466063349e-07,
+      "loss": 1.4795,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.484162895927603e-07,
+      "loss": 1.5594,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.766968325791855e-07,
+      "loss": 1.5677,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.049773755656109e-07,
+      "loss": 1.5635,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.332579185520363e-07,
+      "loss": 1.5494,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.615384615384617e-07,
+      "loss": 1.4992,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.89819004524887e-07,
+      "loss": 1.4569,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0180995475113123e-06,
+      "loss": 1.5654,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0463800904977378e-06,
+      "loss": 1.4813,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.074660633484163e-06,
+      "loss": 1.5734,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1029411764705884e-06,
+      "loss": 1.4907,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1312217194570136e-06,
+      "loss": 1.4825,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1595022624434391e-06,
+      "loss": 1.5824,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1877828054298642e-06,
+      "loss": 1.4838,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2160633484162897e-06,
+      "loss": 1.5418,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.244343891402715e-06,
+      "loss": 1.4261,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2726244343891403e-06,
+      "loss": 1.4816,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3009049773755658e-06,
+      "loss": 1.501,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.329185520361991e-06,
+      "loss": 1.4837,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3574660633484164e-06,
+      "loss": 1.5118,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3857466063348419e-06,
+      "loss": 1.4323,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4140271493212672e-06,
+      "loss": 1.5228,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4423076923076922e-06,
+      "loss": 1.5563,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4705882352941177e-06,
+      "loss": 1.5522,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.498868778280543e-06,
+      "loss": 1.4955,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5271493212669685e-06,
+      "loss": 1.4745,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5554298642533938e-06,
+      "loss": 1.5065,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.583710407239819e-06,
+      "loss": 1.5047,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6119909502262446e-06,
+      "loss": 1.5371,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6402714932126699e-06,
+      "loss": 1.4695,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.668552036199095e-06,
+      "loss": 1.4403,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6968325791855207e-06,
+      "loss": 1.508,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7251131221719457e-06,
+      "loss": 1.59,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.753393665158371e-06,
+      "loss": 1.4488,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7816742081447965e-06,
+      "loss": 1.5784,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8099547511312218e-06,
+      "loss": 1.4555,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8382352941176473e-06,
+      "loss": 1.4555,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8665158371040726e-06,
+      "loss": 1.5309,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.894796380090498e-06,
+      "loss": 1.4982,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9230769230769234e-06,
+      "loss": 1.5335,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9513574660633485e-06,
+      "loss": 1.3938,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.979638009049774e-06,
+      "loss": 1.5495,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0079185520361995e-06,
+      "loss": 1.6057,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0361990950226245e-06,
+      "loss": 1.3612,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0644796380090496e-06,
+      "loss": 1.574,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0927601809954755e-06,
+      "loss": 1.4533,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1210407239819006e-06,
+      "loss": 1.4438,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.149321266968326e-06,
+      "loss": 1.4969,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.177601809954751e-06,
+      "loss": 1.4337,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2058823529411767e-06,
+      "loss": 1.5504,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.234162895927602e-06,
+      "loss": 1.422,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2624434389140273e-06,
+      "loss": 1.4684,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2907239819004528e-06,
+      "loss": 1.4466,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3190045248868783e-06,
+      "loss": 1.4688,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3472850678733034e-06,
+      "loss": 1.5187,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3755656108597284e-06,
+      "loss": 1.4312,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.403846153846154e-06,
+      "loss": 1.3719,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4321266968325794e-06,
+      "loss": 1.5079,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4604072398190045e-06,
+      "loss": 1.4641,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.48868778280543e-06,
+      "loss": 1.5316,
+      "step": 880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5169683257918555e-06,
+      "loss": 1.4753,
+      "step": 890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5452488687782806e-06,
+      "loss": 1.464,
+      "step": 900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5735294117647057e-06,
+      "loss": 1.5372,
+      "step": 910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6018099547511316e-06,
+      "loss": 1.4124,
+      "step": 920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.630090497737557e-06,
+      "loss": 1.5994,
+      "step": 930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.658371040723982e-06,
+      "loss": 1.4348,
+      "step": 940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6866515837104072e-06,
+      "loss": 1.4794,
+      "step": 950
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7149321266968327e-06,
+      "loss": 1.4815,
+      "step": 960
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7432126696832582e-06,
+      "loss": 1.4002,
+      "step": 970
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7714932126696837e-06,
+      "loss": 1.5398,
+      "step": 980
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.799773755656109e-06,
+      "loss": 1.4654,
+      "step": 990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8280542986425343e-06,
+      "loss": 1.4325,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8563348416289594e-06,
+      "loss": 1.5533,
+      "step": 1010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.8846153846153845e-06,
+      "loss": 1.4209,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9128959276018104e-06,
+      "loss": 1.5601,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9411764705882355e-06,
+      "loss": 1.4349,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.969457013574661e-06,
+      "loss": 1.4997,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.997737556561086e-06,
+      "loss": 1.5154,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0260180995475115e-06,
+      "loss": 1.402,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.054298642533937e-06,
+      "loss": 1.502,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0825791855203625e-06,
+      "loss": 1.4461,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1108597285067876e-06,
+      "loss": 1.4136,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1391402714932127e-06,
+      "loss": 1.5231,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.167420814479638e-06,
+      "loss": 1.4906,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.1957013574660633e-06,
+      "loss": 1.4874,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.223981900452489e-06,
+      "loss": 1.4985,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2522624434389143e-06,
+      "loss": 1.473,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.2805429864253398e-06,
+      "loss": 1.5196,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.308823529411765e-06,
+      "loss": 1.4124,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.33710407239819e-06,
+      "loss": 1.5973,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.365384615384616e-06,
+      "loss": 1.3911,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.3936651583710413e-06,
+      "loss": 1.4382,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4219457013574664e-06,
+      "loss": 1.5346,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.4502262443438915e-06,
+      "loss": 1.4738,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.478506787330317e-06,
+      "loss": 1.4693,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.506787330316742e-06,
+      "loss": 1.4391,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.535067873303168e-06,
+      "loss": 1.418,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.563348416289593e-06,
+      "loss": 1.5279,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.5916289592760186e-06,
+      "loss": 1.4958,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6199095022624436e-06,
+      "loss": 1.5313,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6481900452488687e-06,
+      "loss": 1.5305,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.6764705882352946e-06,
+      "loss": 1.4799,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7047511312217197e-06,
+      "loss": 1.5581,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7330316742081452e-06,
+      "loss": 1.3712,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.7613122171945703e-06,
+      "loss": 1.5453,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.789592760180996e-06,
+      "loss": 1.3498,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.817873303167421e-06,
+      "loss": 1.444,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.846153846153847e-06,
+      "loss": 1.5072,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.874434389140272e-06,
+      "loss": 1.5186,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.902714932126697e-06,
+      "loss": 1.5021,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.930995475113122e-06,
+      "loss": 1.4748,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.959276018099548e-06,
+      "loss": 1.3905,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.987556561085973e-06,
+      "loss": 1.5389,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.015837104072399e-06,
+      "loss": 1.4154,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.044117647058824e-06,
+      "loss": 1.6069,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.072398190045249e-06,
+      "loss": 1.4372,
+      "step": 1440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.100678733031674e-06,
+      "loss": 1.4283,
+      "step": 1450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.128959276018099e-06,
+      "loss": 1.4517,
+      "step": 1460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.157239819004525e-06,
+      "loss": 1.4549,
+      "step": 1470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.185520361990951e-06,
+      "loss": 1.6026,
+      "step": 1480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.213800904977376e-06,
+      "loss": 1.4289,
+      "step": 1490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.242081447963801e-06,
+      "loss": 1.5206,
+      "step": 1500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.270361990950226e-06,
+      "loss": 1.6213,
+      "step": 1510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.298642533936652e-06,
+      "loss": 1.4484,
+      "step": 1520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.326923076923077e-06,
+      "loss": 1.4826,
+      "step": 1530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.355203619909502e-06,
+      "loss": 1.4997,
+      "step": 1540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.383484162895928e-06,
+      "loss": 1.4951,
+      "step": 1550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.411764705882353e-06,
+      "loss": 1.5391,
+      "step": 1560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4400452488687785e-06,
+      "loss": 1.4632,
+      "step": 1570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.468325791855204e-06,
+      "loss": 1.5417,
+      "step": 1580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4966063348416295e-06,
+      "loss": 1.4523,
+      "step": 1590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.5248868778280546e-06,
+      "loss": 1.4525,
+      "step": 1600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.55316742081448e-06,
+      "loss": 1.5519,
+      "step": 1610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.5814479638009056e-06,
+      "loss": 1.4964,
+      "step": 1620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.609728506787331e-06,
+      "loss": 1.5018,
+      "step": 1630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.6380090497737566e-06,
+      "loss": 1.4033,
+      "step": 1640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.666289592760182e-06,
+      "loss": 1.4996,
+      "step": 1650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.694570135746607e-06,
+      "loss": 1.544,
+      "step": 1660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.722850678733032e-06,
+      "loss": 1.4364,
+      "step": 1670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.751131221719457e-06,
+      "loss": 1.4763,
+      "step": 1680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.779411764705883e-06,
+      "loss": 1.4401,
+      "step": 1690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.807692307692308e-06,
+      "loss": 1.5479,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.835972850678734e-06,
+      "loss": 1.5661,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.864253393665159e-06,
+      "loss": 1.5403,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.892533936651584e-06,
+      "loss": 1.5201,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.920814479638009e-06,
+      "loss": 1.4231,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.949095022624435e-06,
+      "loss": 1.5133,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.97737556561086e-06,
+      "loss": 1.5098,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999999984883541e-06,
+      "loss": 1.4656,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99999945580748e-06,
+      "loss": 1.5685,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999998170908628e-06,
+      "loss": 1.4709,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999961301873736e-06,
+      "loss": 1.4098,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999993333644336e-06,
+      "loss": 1.5516,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999989781280359e-06,
+      "loss": 1.4509,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999854730965166e-06,
+      "loss": 1.4538,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999980409094112e-06,
+      "loss": 1.4752,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999745892746754e-06,
+      "loss": 1.4818,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999968013639967e-06,
+      "loss": 1.5078,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999960682191975e-06,
+      "loss": 1.4188,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999952594932914e-06,
+      "loss": 1.4883,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999943751865231e-06,
+      "loss": 1.3966,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999934152991599e-06,
+      "loss": 1.405,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999237983149196e-06,
+      "loss": 1.542,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999912687838324e-06,
+      "loss": 1.4335,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9999008215651705e-06,
+      "loss": 1.5425,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9998881994990475e-06,
+      "loss": 1.432,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99987482164377e-06,
+      "loss": 1.4181,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999860688003384e-06,
+      "loss": 1.6119,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99984579858216e-06,
+      "loss": 1.4544,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999830153384602e-06,
+      "loss": 1.457,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999813752415439e-06,
+      "loss": 1.4031,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999796595679629e-06,
+      "loss": 1.5013,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.99977868318236e-06,
+      "loss": 1.5033,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999760014929047e-06,
+      "loss": 1.4345,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999740590925334e-06,
+      "loss": 1.5744,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999720411177094e-06,
+      "loss": 1.4683,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.999699475690426e-06,
+      "loss": 1.472,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9996777844716615e-06,
+      "loss": 1.48,
+      "step": 2060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9996553375273575e-06,
+      "loss": 1.5202,
+      "step": 2070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9996321348643e-06,
+      "loss": 1.5497,
+      "step": 2080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999608176489505e-06,
+      "loss": 1.5599,
+      "step": 2090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9995834624102146e-06,
+      "loss": 1.5187,
+      "step": 2100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999557992633901e-06,
+      "loss": 1.5564,
+      "step": 2110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9995317671682655e-06,
+      "loss": 1.4715,
+      "step": 2120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9995047860212355e-06,
+      "loss": 1.5699,
+      "step": 2130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999477049200968e-06,
+      "loss": 1.335,
+      "step": 2140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99944855671585e-06,
+      "loss": 1.4631,
+      "step": 2150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999419308574494e-06,
+      "loss": 1.4774,
+      "step": 2160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999389304785744e-06,
+      "loss": 1.3558,
+      "step": 2170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999358545358671e-06,
+      "loss": 1.5612,
+      "step": 2180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9993270303025734e-06,
+      "loss": 1.4538,
+      "step": 2190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99929475962698e-06,
+      "loss": 1.3841,
+      "step": 2200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999261733341645e-06,
+      "loss": 1.4141,
+      "step": 2210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9992279514565565e-06,
+      "loss": 1.4458,
+      "step": 2220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999193413981926e-06,
+      "loss": 1.5233,
+      "step": 2230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999158120928196e-06,
+      "loss": 1.5308,
+      "step": 2240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999122072306036e-06,
+      "loss": 1.5113,
+      "step": 2250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999085268126344e-06,
+      "loss": 1.4277,
+      "step": 2260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999047708400247e-06,
+      "loss": 1.4175,
+      "step": 2270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.999009393139103e-06,
+      "loss": 1.3974,
+      "step": 2280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998970322354492e-06,
+      "loss": 1.4188,
+      "step": 2290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998930496058229e-06,
+      "loss": 1.5091,
+      "step": 2300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998889914262353e-06,
+      "loss": 1.5589,
+      "step": 2310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998848576979135e-06,
+      "loss": 1.5005,
+      "step": 2320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99880648422107e-06,
+      "loss": 1.5777,
+      "step": 2330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998763636000885e-06,
+      "loss": 1.3951,
+      "step": 2340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998720032331535e-06,
+      "loss": 1.4171,
+      "step": 2350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998675673226202e-06,
+      "loss": 1.5328,
+      "step": 2360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998630558698297e-06,
+      "loss": 1.4626,
+      "step": 2370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99858468876146e-06,
+      "loss": 1.5178,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998538063429558e-06,
+      "loss": 1.4214,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998490682716687e-06,
+      "loss": 1.422,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998442546637172e-06,
+      "loss": 1.4081,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998393655205567e-06,
+      "loss": 1.3941,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998344008436652e-06,
+      "loss": 1.5361,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998293606345437e-06,
+      "loss": 1.3893,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9982424489471595e-06,
+      "loss": 1.491,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9981905362572865e-06,
+      "loss": 1.5052,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998137868291514e-06,
+      "loss": 1.4204,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998084445065763e-06,
+      "loss": 1.5995,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.998030266596185e-06,
+      "loss": 1.4069,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997975332899161e-06,
+      "loss": 1.4696,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997919643991298e-06,
+      "loss": 1.4874,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997863199889433e-06,
+      "loss": 1.3864,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99780600061063e-06,
+      "loss": 1.5414,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997748046172183e-06,
+      "loss": 1.4299,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997689336591613e-06,
+      "loss": 1.4298,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997629871886669e-06,
+      "loss": 1.5485,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997569652075329e-06,
+      "loss": 1.4114,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9975086771758e-06,
+      "loss": 1.5619,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997446947206517e-06,
+      "loss": 1.4948,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.99738446218614e-06,
+      "loss": 1.4656,
+      "step": 2600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997321222133563e-06,
+      "loss": 1.4826,
+      "step": 2610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9972572270679045e-06,
+      "loss": 1.4005,
+      "step": 2620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997192477008511e-06,
+      "loss": 1.5676,
+      "step": 2630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997126971974959e-06,
+      "loss": 1.4399,
+      "step": 2640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.997060711987053e-06,
+      "loss": 1.3826,
+      "step": 2650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996993697064824e-06,
+      "loss": 1.4241,
+      "step": 2660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9969259272285355e-06,
+      "loss": 1.4572,
+      "step": 2670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9968574024986735e-06,
+      "loss": 1.3838,
+      "step": 2680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996788122895956e-06,
+      "loss": 1.3924,
+      "step": 2690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996718088441327e-06,
+      "loss": 1.4446,
+      "step": 2700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996647299155962e-06,
+      "loss": 1.4941,
+      "step": 2710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996575755061262e-06,
+      "loss": 1.3804,
+      "step": 2720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996503456178856e-06,
+      "loss": 1.6013,
+      "step": 2730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9964304025306036e-06,
+      "loss": 1.4525,
+      "step": 2740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996356594138589e-06,
+      "loss": 1.5245,
+      "step": 2750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9962820310251286e-06,
+      "loss": 1.4705,
+      "step": 2760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996206713212764e-06,
+      "loss": 1.4208,
+      "step": 2770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.996130640724266e-06,
+      "loss": 1.4592,
+      "step": 2780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9960538135826345e-06,
+      "loss": 1.3764,
+      "step": 2790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995976231811094e-06,
+      "loss": 1.5068,
+      "step": 2800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995897895433103e-06,
+      "loss": 1.5637,
+      "step": 2810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9958188044723425e-06,
+      "loss": 1.403,
+      "step": 2820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995738958952725e-06,
+      "loss": 1.5894,
+      "step": 2830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995658358898391e-06,
+      "loss": 1.4451,
+      "step": 2840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995577004333706e-06,
+      "loss": 1.4545,
+      "step": 2850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995494895283268e-06,
+      "loss": 1.5041,
+      "step": 2860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995412031771899e-06,
+      "loss": 1.4141,
+      "step": 2870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995328413824653e-06,
+      "loss": 1.5243,
+      "step": 2880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995244041466809e-06,
+      "loss": 1.4508,
+      "step": 2890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995158914723876e-06,
+      "loss": 1.4926,
+      "step": 2900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.995073033621589e-06,
+      "loss": 1.5325,
+      "step": 2910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994986398185914e-06,
+      "loss": 1.4195,
+      "step": 2920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994899008443042e-06,
+      "loss": 1.5451,
+      "step": 2930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9948108644193944e-06,
+      "loss": 1.5161,
+      "step": 2940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994721966141619e-06,
+      "loss": 1.4686,
+      "step": 2950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994632313636593e-06,
+      "loss": 1.5876,
+      "step": 2960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994541906931421e-06,
+      "loss": 1.4038,
+      "step": 2970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994450746053435e-06,
+      "loss": 1.5334,
+      "step": 2980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994358831030195e-06,
+      "loss": 1.4081,
+      "step": 2990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9942661618894915e-06,
+      "loss": 1.4902,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994172738659339e-06,
+      "loss": 1.4899,
+      "step": 3010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.994078561367984e-06,
+      "loss": 1.4165,
+      "step": 3020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993983630043898e-06,
+      "loss": 1.5057,
+      "step": 3030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993887944715783e-06,
+      "loss": 1.4232,
+      "step": 3040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993791505412564e-06,
+      "loss": 1.295,
+      "step": 3050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9936943121634e-06,
+      "loss": 1.5294,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993596364997675e-06,
+      "loss": 1.4918,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993497663945002e-06,
+      "loss": 1.478,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99339820903522e-06,
+      "loss": 1.4342,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993298000298397e-06,
+      "loss": 1.4342,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99319703776483e-06,
+      "loss": 1.5075,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.993095321465042e-06,
+      "loss": 1.4219,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992992851429785e-06,
+      "loss": 1.5178,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.99288962769004e-06,
+      "loss": 1.4205,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992785650277013e-06,
+      "loss": 1.4201,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992680919222141e-06,
+      "loss": 1.5238,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992575434557084e-06,
+      "loss": 1.4272,
+      "step": 3170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9924691963137365e-06,
+      "loss": 1.5143,
+      "step": 3180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992362204524216e-06,
+      "loss": 1.4496,
+      "step": 3190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9922544592208695e-06,
+      "loss": 1.4056,
+      "step": 3200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992145960436271e-06,
+      "loss": 1.6066,
+      "step": 3210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.992036708203224e-06,
+      "loss": 1.4332,
+      "step": 3220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9919267025547576e-06,
+      "loss": 1.5139,
+      "step": 3230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.99181594352413e-06,
+      "loss": 1.3822,
+      "step": 3240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991704431144827e-06,
+      "loss": 1.4085,
+      "step": 3250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9915921654505614e-06,
+      "loss": 1.5929,
+      "step": 3260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991479146475277e-06,
+      "loss": 1.4264,
+      "step": 3270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9913653742531385e-06,
+      "loss": 1.5663,
+      "step": 3280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991250848818546e-06,
+      "loss": 1.3858,
+      "step": 3290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.991135570206122e-06,
+      "loss": 1.4508,
+      "step": 3300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9910195384507195e-06,
+      "loss": 1.5474,
+      "step": 3310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990902753587418e-06,
+      "loss": 1.3627,
+      "step": 3320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990785215651526e-06,
+      "loss": 1.4989,
+      "step": 3330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990666924678577e-06,
+      "loss": 1.4702,
+      "step": 3340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990547880704334e-06,
+      "loss": 1.4736,
+      "step": 3350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990428083764788e-06,
+      "loss": 1.4802,
+      "step": 3360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990307533896157e-06,
+      "loss": 1.4159,
+      "step": 3370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.990186231134887e-06,
+      "loss": 1.5414,
+      "step": 3380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9900641755176515e-06,
+      "loss": 1.4176,
+      "step": 3390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989941367081351e-06,
+      "loss": 1.475,
+      "step": 3400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989817805863114e-06,
+      "loss": 1.5012,
+      "step": 3410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989693491900297e-06,
+      "loss": 1.4614,
+      "step": 3420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989568425230483e-06,
+      "loss": 1.4872,
+      "step": 3430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9894426058914856e-06,
+      "loss": 1.4351,
+      "step": 3440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9893160339213405e-06,
+      "loss": 1.4578,
+      "step": 3450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989188709358316e-06,
+      "loss": 1.5161,
+      "step": 3460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.989060632240905e-06,
+      "loss": 1.4736,
+      "step": 3470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988931802607831e-06,
+      "loss": 1.4561,
+      "step": 3480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9888022204980405e-06,
+      "loss": 1.492,
+      "step": 3490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988671885950711e-06,
+      "loss": 1.4588,
+      "step": 3500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9885407990052466e-06,
+      "loss": 1.5282,
+      "step": 3510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9884089597012784e-06,
+      "loss": 1.3947,
+      "step": 3520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988276368078666e-06,
+      "loss": 1.5018,
+      "step": 3530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988143024177494e-06,
+      "loss": 1.3993,
+      "step": 3540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.988008928038078e-06,
+      "loss": 1.445,
+      "step": 3550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9878740797009585e-06,
+      "loss": 1.5647,
+      "step": 3560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987738479206903e-06,
+      "loss": 1.5842,
+      "step": 3570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98760212659691e-06,
+      "loss": 1.5348,
+      "step": 3580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9874650219122004e-06,
+      "loss": 1.4912,
+      "step": 3590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987327165194225e-06,
+      "loss": 1.4188,
+      "step": 3600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9871885564846645e-06,
+      "loss": 1.4747,
+      "step": 3610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.987049195825422e-06,
+      "loss": 1.4021,
+      "step": 3620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986909083258632e-06,
+      "loss": 1.4351,
+      "step": 3630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9867682188266524e-06,
+      "loss": 1.4167,
+      "step": 3640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986626602572072e-06,
+      "loss": 1.4173,
+      "step": 3650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986484234537706e-06,
+      "loss": 1.4513,
+      "step": 3660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986341114766595e-06,
+      "loss": 1.4415,
+      "step": 3670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98619724330201e-06,
+      "loss": 1.4643,
+      "step": 3680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.986052620187445e-06,
+      "loss": 1.3739,
+      "step": 3690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985907245466627e-06,
+      "loss": 1.5043,
+      "step": 3700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9857611191835045e-06,
+      "loss": 1.5082,
+      "step": 3710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985614241382257e-06,
+      "loss": 1.4664,
+      "step": 3720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985466612107289e-06,
+      "loss": 1.4956,
+      "step": 3730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985318231403234e-06,
+      "loss": 1.3663,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.985169099314952e-06,
+      "loss": 1.3684,
+      "step": 3750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98501921588753e-06,
+      "loss": 1.4453,
+      "step": 3760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.98486858116628e-06,
+      "loss": 1.4815,
+      "step": 3770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984717195196747e-06,
+      "loss": 1.5231,
+      "step": 3780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984565058024695e-06,
+      "loss": 1.3922,
+      "step": 3790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.9844121696961235e-06,
+      "loss": 1.4669,
+      "step": 3800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984258530257253e-06,
+      "loss": 1.5285,
+      "step": 3810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.984104139754533e-06,
+      "loss": 1.452,
+      "step": 3820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983948998234642e-06,
+      "loss": 1.4347,
+      "step": 3830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983793105744482e-06,
+      "loss": 1.4697,
+      "step": 3840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983636462331185e-06,
+      "loss": 1.4757,
+      "step": 3850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983479068042107e-06,
+      "loss": 1.5572,
+      "step": 3860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983320922924836e-06,
+      "loss": 1.4198,
+      "step": 3870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983162027027182e-06,
+      "loss": 1.56,
+      "step": 3880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.983002380397183e-06,
+      "loss": 1.4603,
+      "step": 3890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982841983083106e-06,
+      "loss": 1.5076,
+      "step": 3900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982680835133445e-06,
+      "loss": 1.5286,
+      "step": 3910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982518936596917e-06,
+      "loss": 1.3947,
+      "step": 3920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9823562875224705e-06,
+      "loss": 1.4712,
+      "step": 3930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982192887959279e-06,
+      "loss": 1.4569,
+      "step": 3940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.982028737956743e-06,
+      "loss": 1.3805,
+      "step": 3950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981863837564488e-06,
+      "loss": 1.505,
+      "step": 3960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.98169818683237e-06,
+      "loss": 1.4096,
+      "step": 3970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981531785810471e-06,
+      "loss": 1.5323,
+      "step": 3980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9813646345490956e-06,
+      "loss": 1.4851,
+      "step": 3990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9811967330987816e-06,
+      "loss": 1.4155,
+      "step": 4000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.981028081510288e-06,
+      "loss": 1.5397,
+      "step": 4010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980858679834607e-06,
+      "loss": 1.4625,
+      "step": 4020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980688528122949e-06,
+      "loss": 1.5063,
+      "step": 4030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980517626426758e-06,
+      "loss": 1.4523,
+      "step": 4040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.980345974797703e-06,
+      "loss": 1.3793,
+      "step": 4050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.98017357328768e-06,
+      "loss": 1.478,
+      "step": 4060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9800004219488085e-06,
+      "loss": 1.4085,
+      "step": 4070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979826520833439e-06,
+      "loss": 1.4852,
+      "step": 4080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979651869994145e-06,
+      "loss": 1.4153,
+      "step": 4090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979476469483732e-06,
+      "loss": 1.4517,
+      "step": 4100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979300319355226e-06,
+      "loss": 1.5164,
+      "step": 4110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.979123419661882e-06,
+      "loss": 1.4095,
+      "step": 4120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978945770457184e-06,
+      "loss": 1.4913,
+      "step": 4130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97876737179484e-06,
+      "loss": 1.4144,
+      "step": 4140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978588223728783e-06,
+      "loss": 1.4062,
+      "step": 4150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9784083263131764e-06,
+      "loss": 1.5903,
+      "step": 4160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978227679602408e-06,
+      "loss": 1.4546,
+      "step": 4170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.978046283651095e-06,
+      "loss": 1.5481,
+      "step": 4180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977864138514075e-06,
+      "loss": 1.3926,
+      "step": 4190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977681244246417e-06,
+      "loss": 1.5306,
+      "step": 4200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9774976009034165e-06,
+      "loss": 1.4754,
+      "step": 4210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977313208540593e-06,
+      "loss": 1.3965,
+      "step": 4220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.977128067213694e-06,
+      "loss": 1.538,
+      "step": 4230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9769421769786935e-06,
+      "loss": 1.4035,
+      "step": 4240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9767555378917905e-06,
+      "loss": 1.4671,
+      "step": 4250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976568150009413e-06,
+      "loss": 1.5457,
+      "step": 4260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976380013388213e-06,
+      "loss": 1.4047,
+      "step": 4270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976191128085069e-06,
+      "loss": 1.4349,
+      "step": 4280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.976001494157088e-06,
+      "loss": 1.4366,
+      "step": 4290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9758111116616005e-06,
+      "loss": 1.4432,
+      "step": 4300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975619980656165e-06,
+      "loss": 1.5248,
+      "step": 4310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.9754281011985666e-06,
+      "loss": 1.5187,
+      "step": 4320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975235473346816e-06,
+      "loss": 1.478,
+      "step": 4330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.975042097159149e-06,
+      "loss": 1.5115,
+      "step": 4340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974847972694031e-06,
+      "loss": 1.416,
+      "step": 4350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.97465310001015e-06,
+      "loss": 1.447,
+      "step": 4360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974457479166422e-06,
+      "loss": 1.4575,
+      "step": 4370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974261110221988e-06,
+      "loss": 1.4721,
+      "step": 4380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.974063993236218e-06,
+      "loss": 1.4474,
+      "step": 4390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.973866128268705e-06,
+      "loss": 1.4474,
+      "step": 4400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.973667515379269e-06,
+      "loss": 1.5032,
+      "step": 4410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.973468154627957e-06,
+      "loss": 1.3976,
+      "step": 4420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.973268046075043e-06,
+      "loss": 1.5152,
+      "step": 4430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.973067189781022e-06,
+      "loss": 1.4357,
+      "step": 4440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972865585806622e-06,
+      "loss": 1.5042,
+      "step": 4450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9726632342127926e-06,
+      "loss": 1.4581,
+      "step": 4460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972460135060711e-06,
+      "loss": 1.4322,
+      "step": 4470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.97225628841178e-06,
+      "loss": 1.5122,
+      "step": 4480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.972051694327627e-06,
+      "loss": 1.4024,
+      "step": 4490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971846352870109e-06,
+      "loss": 1.3898,
+      "step": 4500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971640264101305e-06,
+      "loss": 1.5312,
+      "step": 4510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971433428083522e-06,
+      "loss": 1.4878,
+      "step": 4520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971225844879293e-06,
+      "loss": 1.5128,
+      "step": 4530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.971017514551375e-06,
+      "loss": 1.4232,
+      "step": 4540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970808437162756e-06,
+      "loss": 1.4027,
+      "step": 4550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970598612776641e-06,
+      "loss": 1.4991,
+      "step": 4560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970388041456471e-06,
+      "loss": 1.3082,
+      "step": 4570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.970176723265905e-06,
+      "loss": 1.5382,
+      "step": 4580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969964658268831e-06,
+      "loss": 1.4533,
+      "step": 4590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969751846529363e-06,
+      "loss": 1.4065,
+      "step": 4600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96953828811184e-06,
+      "loss": 1.4569,
+      "step": 4610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969323983080827e-06,
+      "loss": 1.3586,
+      "step": 4620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.969108931501114e-06,
+      "loss": 1.4907,
+      "step": 4630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968893133437718e-06,
+      "loss": 1.3845,
+      "step": 4640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9686765889558815e-06,
+      "loss": 1.5598,
+      "step": 4650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968459298121072e-06,
+      "loss": 1.4792,
+      "step": 4660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.968241260998982e-06,
+      "loss": 1.3578,
+      "step": 4670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9680224776555305e-06,
+      "loss": 1.457,
+      "step": 4680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.967802948156863e-06,
+      "loss": 1.4253,
+      "step": 4690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96758267256935e-06,
+      "loss": 1.4427,
+      "step": 4700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9673616509595855e-06,
+      "loss": 1.5018,
+      "step": 4710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.967139883394392e-06,
+      "loss": 1.4773,
+      "step": 4720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966917369940817e-06,
+      "loss": 1.4882,
+      "step": 4730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966694110666131e-06,
+      "loss": 1.4579,
+      "step": 4740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966470105637833e-06,
+      "loss": 1.5047,
+      "step": 4750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966245354923645e-06,
+      "loss": 1.5952,
+      "step": 4760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.966019858591519e-06,
+      "loss": 1.4054,
+      "step": 4770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965793616709625e-06,
+      "loss": 1.5132,
+      "step": 4780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965566629346365e-06,
+      "loss": 1.4037,
+      "step": 4790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.965338896570363e-06,
+      "loss": 1.4161,
+      "step": 4800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9651104184504695e-06,
+      "loss": 1.512,
+      "step": 4810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96488119505576e-06,
+      "loss": 1.3951,
+      "step": 4820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.964651226455536e-06,
+      "loss": 1.4651,
+      "step": 4830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9644205127193225e-06,
+      "loss": 1.3567,
+      "step": 4840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.964189053916872e-06,
+      "loss": 1.3462,
+      "step": 4850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963956850118162e-06,
+      "loss": 1.5105,
+      "step": 4860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963723901393392e-06,
+      "loss": 1.4354,
+      "step": 4870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.963490207812991e-06,
+      "loss": 1.489,
+      "step": 4880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9632557694476115e-06,
+      "loss": 1.4402,
+      "step": 4890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.96302058636813e-06,
+      "loss": 1.4668,
+      "step": 4900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962784658645651e-06,
+      "loss": 1.4677,
+      "step": 4910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962547986351501e-06,
+      "loss": 1.5023,
+      "step": 4920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962310569557232e-06,
+      "loss": 1.503,
+      "step": 4930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.962072408334624e-06,
+      "loss": 1.4746,
+      "step": 4940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961833502755678e-06,
+      "loss": 1.4409,
+      "step": 4950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961593852892625e-06,
+      "loss": 1.5168,
+      "step": 4960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961353458817915e-06,
+      "loss": 1.4077,
+      "step": 4970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.961112320604229e-06,
+      "loss": 1.4972,
+      "step": 4980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.960870438324468e-06,
+      "loss": 1.287,
+      "step": 4990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.9606278120517606e-06,
+      "loss": 1.4007,
+      "step": 5000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.96038444185946e-06,
+      "loss": 1.4672,
+      "step": 5010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.960140327821145e-06,
+      "loss": 1.4173,
+      "step": 5020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959895470010616e-06,
+      "loss": 1.551,
+      "step": 5030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959649868501903e-06,
+      "loss": 1.444,
+      "step": 5040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959403523369259e-06,
+      "loss": 1.4112,
+      "step": 5050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.959156434687158e-06,
+      "loss": 1.5271,
+      "step": 5060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958908602530306e-06,
+      "loss": 1.4548,
+      "step": 5070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958660026973626e-06,
+      "loss": 1.5683,
+      "step": 5080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958410708092273e-06,
+      "loss": 1.4577,
+      "step": 5090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.958160645961622e-06,
+      "loss": 1.388,
+      "step": 5100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957909840657274e-06,
+      "loss": 1.5217,
+      "step": 5110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957658292255055e-06,
+      "loss": 1.4436,
+      "step": 5120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957406000831015e-06,
+      "loss": 1.4131,
+      "step": 5130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.957152966461429e-06,
+      "loss": 1.416,
+      "step": 5140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956899189222798e-06,
+      "loss": 1.3651,
+      "step": 5150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956644669191845e-06,
+      "loss": 1.4931,
+      "step": 5160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9563894064455186e-06,
+      "loss": 1.4367,
+      "step": 5170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.956133401060993e-06,
+      "loss": 1.547,
+      "step": 5180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955876653115666e-06,
+      "loss": 1.4535,
+      "step": 5190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.95561916268716e-06,
+      "loss": 1.4241,
+      "step": 5200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955360929853321e-06,
+      "loss": 1.5641,
+      "step": 5210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.955101954692222e-06,
+      "loss": 1.3793,
+      "step": 5220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954842237282157e-06,
+      "loss": 1.5718,
+      "step": 5230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954581777701647e-06,
+      "loss": 1.4324,
+      "step": 5240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954320576029436e-06,
+      "loss": 1.4595,
+      "step": 5250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.954058632344494e-06,
+      "loss": 1.5296,
+      "step": 5260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953795946726013e-06,
+      "loss": 1.4267,
+      "step": 5270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953532519253412e-06,
+      "loss": 1.533,
+      "step": 5280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953268350006332e-06,
+      "loss": 1.4179,
+      "step": 5290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.953003439064637e-06,
+      "loss": 1.398,
+      "step": 5300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952737786508421e-06,
+      "loss": 1.5046,
+      "step": 5310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.952471392417997e-06,
+      "loss": 1.4068,
+      "step": 5320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9522042568739025e-06,
+      "loss": 1.5226,
+      "step": 5330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951936379956902e-06,
+      "loss": 1.3876,
+      "step": 5340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9516677617479816e-06,
+      "loss": 1.4449,
+      "step": 5350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951398402328353e-06,
+      "loss": 1.574,
+      "step": 5360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.951128301779451e-06,
+      "loss": 1.4011,
+      "step": 5370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.950857460182934e-06,
+      "loss": 1.5441,
+      "step": 5380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.950585877620688e-06,
+      "loss": 1.4581,
+      "step": 5390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.950313554174817e-06,
+      "loss": 1.3958,
+      "step": 5400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.950040489927656e-06,
+      "loss": 1.5189,
+      "step": 5410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949766684961756e-06,
+      "loss": 1.3556,
+      "step": 5420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949492139359899e-06,
+      "loss": 1.5353,
+      "step": 5430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.949216853205089e-06,
+      "loss": 1.4634,
+      "step": 5440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94894082658055e-06,
+      "loss": 1.4505,
+      "step": 5450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.948664059569736e-06,
+      "loss": 1.4459,
+      "step": 5460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.94838655225632e-06,
+      "loss": 1.4651,
+      "step": 5470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.948108304724201e-06,
+      "loss": 1.4019,
+      "step": 5480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.947829317057501e-06,
+      "loss": 1.3286,
+      "step": 5490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.9475495893405655e-06,
+      "loss": 1.432,
+      "step": 5500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.947269121657966e-06,
+      "loss": 1.4925,
+      "step": 5510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.946987914094495e-06,
+      "loss": 1.4754,
+      "step": 5520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.946705966735171e-06,
+      "loss": 1.4702,
+      "step": 5530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.946423279665233e-06,
+      "loss": 1.4101,
+      "step": 5540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.946139852970145e-06,
+      "loss": 1.5099,
+      "step": 5550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945855686735599e-06,
+      "loss": 1.4531,
+      "step": 5560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945570781047504e-06,
+      "loss": 1.4489,
+      "step": 5570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.945285135991995e-06,
+      "loss": 1.5326,
+      "step": 5580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.944998751655432e-06,
+      "loss": 1.4092,
+      "step": 5590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944711628124396e-06,
+      "loss": 1.4042,
+      "step": 5600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944423765485695e-06,
+      "loss": 1.4863,
+      "step": 5610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.944135163826356e-06,
+      "loss": 1.4768,
+      "step": 5620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.943845823233633e-06,
+      "loss": 1.5316,
+      "step": 5630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.943555743795002e-06,
+      "loss": 1.4212,
+      "step": 5640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9432649255981614e-06,
+      "loss": 1.4582,
+      "step": 5650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942973368731036e-06,
+      "loss": 1.5213,
+      "step": 5660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.94268107328177e-06,
+      "loss": 1.4514,
+      "step": 5670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.942388039338733e-06,
+      "loss": 1.4642,
+      "step": 5680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.94209426699052e-06,
+      "loss": 1.4084,
+      "step": 5690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9417997563259444e-06,
+      "loss": 1.4224,
+      "step": 5700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.941504507434047e-06,
+      "loss": 1.4966,
+      "step": 5710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9412085204040896e-06,
+      "loss": 1.4337,
+      "step": 5720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940911795325557e-06,
+      "loss": 1.4267,
+      "step": 5730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9406143322881586e-06,
+      "loss": 1.5011,
+      "step": 5740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940316131381825e-06,
+      "loss": 1.4268,
+      "step": 5750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.940017192696713e-06,
+      "loss": 1.4369,
+      "step": 5760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.939717516323199e-06,
+      "loss": 1.5087,
+      "step": 5770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9394171023518855e-06,
+      "loss": 1.5486,
+      "step": 5780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9391159508735945e-06,
+      "loss": 1.3977,
+      "step": 5790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9388140619793735e-06,
+      "loss": 1.5377,
+      "step": 5800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938511435760494e-06,
+      "loss": 1.5364,
+      "step": 5810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.938208072308445e-06,
+      "loss": 1.3716,
+      "step": 5820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937903971714946e-06,
+      "loss": 1.58,
+      "step": 5830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937599134071933e-06,
+      "loss": 1.4059,
+      "step": 5840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.937293559471569e-06,
+      "loss": 1.5141,
+      "step": 5850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9369872480062365e-06,
+      "loss": 1.5064,
+      "step": 5860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936680199768545e-06,
+      "loss": 1.4567,
+      "step": 5870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936372414851321e-06,
+      "loss": 1.4904,
+      "step": 5880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.936063893347617e-06,
+      "loss": 1.4206,
+      "step": 5890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935754635350711e-06,
+      "loss": 1.4501,
+      "step": 5900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935444640954098e-06,
+      "loss": 1.4794,
+      "step": 5910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.935133910251499e-06,
+      "loss": 1.3777,
+      "step": 5920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.934822443336858e-06,
+      "loss": 1.5063,
+      "step": 5930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.934510240304341e-06,
+      "loss": 1.5084,
+      "step": 5940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.9341973012483335e-06,
+      "loss": 1.4633,
+      "step": 5950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933883626263448e-06,
+      "loss": 1.5299,
+      "step": 5960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933569215444517e-06,
+      "loss": 1.4182,
+      "step": 5970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.933254068886596e-06,
+      "loss": 1.4824,
+      "step": 5980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932938186684963e-06,
+      "loss": 1.3725,
+      "step": 5990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.93262156893512e-06,
+      "loss": 1.4103,
+      "step": 6000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.932304215732787e-06,
+      "loss": 1.5565,
+      "step": 6010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931986127173911e-06,
+      "loss": 1.4205,
+      "step": 6020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931667303354659e-06,
+      "loss": 1.658,
+      "step": 6030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931347744371421e-06,
+      "loss": 1.4694,
+      "step": 6040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.931027450320808e-06,
+      "loss": 1.4775,
+      "step": 6050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.930706421299655e-06,
+      "loss": 1.5513,
+      "step": 6060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.930384657405018e-06,
+      "loss": 1.3282,
+      "step": 6070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.930062158734178e-06,
+      "loss": 1.4546,
+      "step": 6080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929738925384632e-06,
+      "loss": 1.4197,
+      "step": 6090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929414957454105e-06,
+      "loss": 1.4306,
+      "step": 6100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.929090255040541e-06,
+      "loss": 1.5378,
+      "step": 6110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.928764818242109e-06,
+      "loss": 1.3612,
+      "step": 6120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.928438647157197e-06,
+      "loss": 1.5351,
+      "step": 6130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.928111741884415e-06,
+      "loss": 1.4257,
+      "step": 6140
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.927784102522596e-06,
+      "loss": 1.4778,
+      "step": 6150
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.927455729170796e-06,
+      "loss": 1.511,
+      "step": 6160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.927126621928292e-06,
+      "loss": 1.4569,
+      "step": 6170
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.926796780894582e-06,
+      "loss": 1.549,
+      "step": 6180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.926466206169386e-06,
+      "loss": 1.4422,
+      "step": 6190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.926134897852648e-06,
+      "loss": 1.4503,
+      "step": 6200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925802856044532e-06,
+      "loss": 1.4996,
+      "step": 6210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925470080845422e-06,
+      "loss": 1.5266,
+      "step": 6220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.925136572355927e-06,
+      "loss": 1.4945,
+      "step": 6230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.924802330676877e-06,
+      "loss": 1.4019,
+      "step": 6240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.924467355909322e-06,
+      "loss": 1.3886,
+      "step": 6250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.924131648154536e-06,
+      "loss": 1.4926,
+      "step": 6260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92379520751401e-06,
+      "loss": 1.4609,
+      "step": 6270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.923458034089463e-06,
+      "loss": 1.5187,
+      "step": 6280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.923120127982832e-06,
+      "loss": 1.38,
+      "step": 6290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922781489296274e-06,
+      "loss": 1.4699,
+      "step": 6300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922442118132172e-06,
+      "loss": 1.6033,
+      "step": 6310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.922102014593126e-06,
+      "loss": 1.4099,
+      "step": 6320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.92176117878196e-06,
+      "loss": 1.5635,
+      "step": 6330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.921419610801717e-06,
+      "loss": 1.4405,
+      "step": 6340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9210773107556655e-06,
+      "loss": 1.4319,
+      "step": 6350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.920734278747291e-06,
+      "loss": 1.5077,
+      "step": 6360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.920390514880302e-06,
+      "loss": 1.45,
+      "step": 6370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.920046019258629e-06,
+      "loss": 1.5265,
+      "step": 6380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.919700791986425e-06,
+      "loss": 1.4225,
+      "step": 6390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.919354833168059e-06,
+      "loss": 1.4696,
+      "step": 6400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.919008142908125e-06,
+      "loss": 1.4802,
+      "step": 6410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.918660721311439e-06,
+      "loss": 1.4518,
+      "step": 6420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.918312568483036e-06,
+      "loss": 1.5918,
+      "step": 6430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9179636845281735e-06,
+      "loss": 1.5012,
+      "step": 6440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9176140695523276e-06,
+      "loss": 1.4512,
+      "step": 6450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.917263723661198e-06,
+      "loss": 1.5069,
+      "step": 6460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916912646960705e-06,
+      "loss": 1.3988,
+      "step": 6470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916560839556989e-06,
+      "loss": 1.469,
+      "step": 6480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.916208301556411e-06,
+      "loss": 1.4563,
+      "step": 6490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915855033065555e-06,
+      "loss": 1.4556,
+      "step": 6500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915501034191223e-06,
+      "loss": 1.4758,
+      "step": 6510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.915146305040439e-06,
+      "loss": 1.4358,
+      "step": 6520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9147908457204495e-06,
+      "loss": 1.4403,
+      "step": 6530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.914434656338719e-06,
+      "loss": 1.4285,
+      "step": 6540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.914077737002934e-06,
+      "loss": 1.5014,
+      "step": 6550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913720087821003e-06,
+      "loss": 1.5255,
+      "step": 6560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913361708901052e-06,
+      "loss": 1.4708,
+      "step": 6570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.913002600351431e-06,
+      "loss": 1.4998,
+      "step": 6580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.912642762280707e-06,
+      "loss": 1.4158,
+      "step": 6590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.91228219479767e-06,
+      "loss": 1.5156,
+      "step": 6600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911920898011333e-06,
+      "loss": 1.5561,
+      "step": 6610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.911558872030922e-06,
+      "loss": 1.457,
+      "step": 6620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9111961169658916e-06,
+      "loss": 1.5409,
+      "step": 6630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910832632925912e-06,
+      "loss": 1.4722,
+      "step": 6640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910468420020874e-06,
+      "loss": 1.4112,
+      "step": 6650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.910103478360893e-06,
+      "loss": 1.479,
+      "step": 6660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.909737808056297e-06,
+      "loss": 1.5052,
+      "step": 6670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.909371409217642e-06,
+      "loss": 1.4316,
+      "step": 6680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.909004281955699e-06,
+      "loss": 1.4393,
+      "step": 6690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.908636426381464e-06,
+      "loss": 1.3998,
+      "step": 6700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.908267842606148e-06,
+      "loss": 1.5134,
+      "step": 6710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9078985307411865e-06,
+      "loss": 1.5032,
+      "step": 6720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.907528490898231e-06,
+      "loss": 1.5423,
+      "step": 6730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.907157723189158e-06,
+      "loss": 1.3855,
+      "step": 6740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.906786227726059e-06,
+      "loss": 1.4392,
+      "step": 6750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.906414004621249e-06,
+      "loss": 1.4457,
+      "step": 6760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.9060410539872625e-06,
+      "loss": 1.4663,
+      "step": 6770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9056673759368526e-06,
+      "loss": 1.4329,
+      "step": 6780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.905292970582993e-06,
+      "loss": 1.4488,
+      "step": 6790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.904917838038877e-06,
+      "loss": 1.4108,
+      "step": 6800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.904541978417919e-06,
+      "loss": 1.5219,
+      "step": 6810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.904165391833753e-06,
+      "loss": 1.5253,
+      "step": 6820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.90378807840023e-06,
+      "loss": 1.5024,
+      "step": 6830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.903410038231425e-06,
+      "loss": 1.3931,
+      "step": 6840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9030312714416285e-06,
+      "loss": 1.4106,
+      "step": 6850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.902651778145354e-06,
+      "loss": 1.4441,
+      "step": 6860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.902271558457334e-06,
+      "loss": 1.4366,
+      "step": 6870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.901890612492519e-06,
+      "loss": 1.5343,
+      "step": 6880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.901508940366079e-06,
+      "loss": 1.4303,
+      "step": 6890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.901126542193408e-06,
+      "loss": 1.4826,
+      "step": 6900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.9007434180901125e-06,
+      "loss": 1.5317,
+      "step": 6910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.900359568172024e-06,
+      "loss": 1.3986,
+      "step": 6920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899974992555192e-06,
+      "loss": 1.5309,
+      "step": 6930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899589691355884e-06,
+      "loss": 1.412,
+      "step": 6940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.899203664690587e-06,
+      "loss": 1.386,
+      "step": 6950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8988169126760085e-06,
+      "loss": 1.4784,
+      "step": 6960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.898429435429077e-06,
+      "loss": 1.387,
+      "step": 6970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.898041233066935e-06,
+      "loss": 1.5162,
+      "step": 6980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.897652305706951e-06,
+      "loss": 1.4889,
+      "step": 6990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.897262653466706e-06,
+      "loss": 1.3065,
+      "step": 7000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.896872276464005e-06,
+      "loss": 1.5205,
+      "step": 7010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.896481174816869e-06,
+      "loss": 1.4492,
+      "step": 7020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.89608934864354e-06,
+      "loss": 1.532,
+      "step": 7030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8956967980624806e-06,
+      "loss": 1.4294,
+      "step": 7040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.895303523192366e-06,
+      "loss": 1.4815,
+      "step": 7050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.894909524152098e-06,
+      "loss": 1.5256,
+      "step": 7060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.894514801060793e-06,
+      "loss": 1.3841,
+      "step": 7070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.894119354037787e-06,
+      "loss": 1.4862,
+      "step": 7080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.893723183202637e-06,
+      "loss": 1.3666,
+      "step": 7090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.893326288675113e-06,
+      "loss": 1.3959,
+      "step": 7100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892928670575212e-06,
+      "loss": 1.4215,
+      "step": 7110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892530329023144e-06,
+      "loss": 1.4322,
+      "step": 7120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.892131264139338e-06,
+      "loss": 1.44,
+      "step": 7130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.891731476044446e-06,
+      "loss": 1.4482,
+      "step": 7140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.891330964859332e-06,
+      "loss": 1.4613,
+      "step": 7150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.890929730705084e-06,
+      "loss": 1.5197,
+      "step": 7160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8905277737030076e-06,
+      "loss": 1.4349,
+      "step": 7170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.890125093974625e-06,
+      "loss": 1.4933,
+      "step": 7180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.889721691641678e-06,
+      "loss": 1.3831,
+      "step": 7190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.889317566826128e-06,
+      "loss": 1.3665,
+      "step": 7200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.888912719650152e-06,
+      "loss": 1.4973,
+      "step": 7210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.888507150236148e-06,
+      "loss": 1.4781,
+      "step": 7220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8881008587067315e-06,
+      "loss": 1.4407,
+      "step": 7230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.8876938451847364e-06,
+      "loss": 1.4793,
+      "step": 7240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.887286109793215e-06,
+      "loss": 1.4418,
+      "step": 7250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886877652655437e-06,
+      "loss": 1.5221,
+      "step": 7260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886468473894891e-06,
+      "loss": 1.418,
+      "step": 7270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.886058573635284e-06,
+      "loss": 1.5194,
+      "step": 7280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.885647952000541e-06,
+      "loss": 1.4484,
+      "step": 7290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.885236609114805e-06,
+      "loss": 1.4004,
+      "step": 7300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.884824545102436e-06,
+      "loss": 1.5709,
+      "step": 7310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.884411760088012e-06,
+      "loss": 1.434,
+      "step": 7320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883998254196334e-06,
+      "loss": 1.5022,
+      "step": 7330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883584027552413e-06,
+      "loss": 1.4681,
+      "step": 7340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.883169080281483e-06,
+      "loss": 1.3556,
+      "step": 7350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.882753412508996e-06,
+      "loss": 1.4233,
+      "step": 7360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.882337024360618e-06,
+      "loss": 1.4186,
+      "step": 7370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.881919915962236e-06,
+      "loss": 1.4828,
+      "step": 7380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.881502087439957e-06,
+      "loss": 1.4188,
+      "step": 7390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.881083538920098e-06,
+      "loss": 1.4104,
+      "step": 7400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.880664270529203e-06,
+      "loss": 1.4973,
+      "step": 7410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.880244282394025e-06,
+      "loss": 1.4466,
+      "step": 7420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.879823574641542e-06,
+      "loss": 1.4127,
+      "step": 7430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.879402147398944e-06,
+      "loss": 1.4078,
+      "step": 7440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878980000793642e-06,
+      "loss": 1.4075,
+      "step": 7450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878557134953263e-06,
+      "loss": 1.5393,
+      "step": 7460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.878133550005651e-06,
+      "loss": 1.4298,
+      "step": 7470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.877709246078869e-06,
+      "loss": 1.5512,
+      "step": 7480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8772842233011955e-06,
+      "loss": 1.3967,
+      "step": 7490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8768584818011275e-06,
+      "loss": 1.3832,
+      "step": 7500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8764320217073805e-06,
+      "loss": 1.4303,
+      "step": 7510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.876004843148884e-06,
+      "loss": 1.4341,
+      "step": 7520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.875576946254788e-06,
+      "loss": 1.4153,
+      "step": 7530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.875148331154456e-06,
+      "loss": 1.4153,
+      "step": 7540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.874718997977474e-06,
+      "loss": 1.3684,
+      "step": 7550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.874288946853639e-06,
+      "loss": 1.4749,
+      "step": 7560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8738581779129705e-06,
+      "loss": 1.405,
+      "step": 7570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8734266912857e-06,
+      "loss": 1.4875,
+      "step": 7580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.872994487102281e-06,
+      "loss": 1.4493,
+      "step": 7590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.872561565493379e-06,
+      "loss": 1.4639,
+      "step": 7600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8721279265898816e-06,
+      "loss": 1.5277,
+      "step": 7610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8716935705228875e-06,
+      "loss": 1.4207,
+      "step": 7620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.871258497423717e-06,
+      "loss": 1.4021,
+      "step": 7630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8708227074239064e-06,
+      "loss": 1.4183,
+      "step": 7640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.870386200655205e-06,
+      "loss": 1.4135,
+      "step": 7650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869948977249583e-06,
+      "loss": 1.5467,
+      "step": 7660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869511037339227e-06,
+      "loss": 1.3883,
+      "step": 7670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.869072381056537e-06,
+      "loss": 1.5283,
+      "step": 7680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.868633008534133e-06,
+      "loss": 1.4464,
+      "step": 7690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8681929199048485e-06,
+      "loss": 1.4178,
+      "step": 7700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.867752115301736e-06,
+      "loss": 1.5876,
+      "step": 7710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.867310594858065e-06,
+      "loss": 1.4084,
+      "step": 7720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.866868358707317e-06,
+      "loss": 1.5118,
+      "step": 7730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.866425406983197e-06,
+      "loss": 1.4276,
+      "step": 7740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8659817398196175e-06,
+      "loss": 1.5102,
+      "step": 7750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.865537357350716e-06,
+      "loss": 1.5511,
+      "step": 7760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.8650922597108396e-06,
+      "loss": 1.4089,
+      "step": 7770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.864646447034556e-06,
+      "loss": 1.4478,
+      "step": 7780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.864199919456647e-06,
+      "loss": 1.399,
+      "step": 7790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.86375267711211e-06,
+      "loss": 1.4264,
+      "step": 7800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.86330472013616e-06,
+      "loss": 1.5155,
+      "step": 7810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.862856048664229e-06,
+      "loss": 1.409,
+      "step": 7820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.862406662831961e-06,
+      "loss": 1.5901,
+      "step": 7830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.86195656277522e-06,
+      "loss": 1.4158,
+      "step": 7840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.861505748630083e-06,
+      "loss": 1.4191,
+      "step": 7850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.861054220532846e-06,
+      "loss": 1.4134,
+      "step": 7860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.860601978620018e-06,
+      "loss": 1.412,
+      "step": 7870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.860149023028325e-06,
+      "loss": 1.5435,
+      "step": 7880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.859695353894709e-06,
+      "loss": 1.4137,
+      "step": 7890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.859240971356327e-06,
+      "loss": 1.4319,
+      "step": 7900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.858785875550553e-06,
+      "loss": 1.5885,
+      "step": 7910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.858330066614975e-06,
+      "loss": 1.4568,
+      "step": 7920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.857873544687397e-06,
+      "loss": 1.5608,
+      "step": 7930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.85741630990584e-06,
+      "loss": 1.4378,
+      "step": 7940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.856958362408539e-06,
+      "loss": 1.4962,
+      "step": 7950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.856499702333944e-06,
+      "loss": 1.5053,
+      "step": 7960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.856040329820722e-06,
+      "loss": 1.4232,
+      "step": 7970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.855580245007756e-06,
+      "loss": 1.4842,
+      "step": 7980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.855119448034141e-06,
+      "loss": 1.4563,
+      "step": 7990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.85465793903919e-06,
+      "loss": 1.3475,
+      "step": 8000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.854195718162431e-06,
+      "loss": 1.5157,
+      "step": 8010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.853732785543607e-06,
+      "loss": 1.417,
+      "step": 8020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.853269141322675e-06,
+      "loss": 1.539,
+      "step": 8030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.85280478563981e-06,
+      "loss": 1.375,
+      "step": 8040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.852339718635398e-06,
+      "loss": 1.4577,
+      "step": 8050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.851873940450045e-06,
+      "loss": 1.5016,
+      "step": 8060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.851407451224566e-06,
+      "loss": 1.38,
+      "step": 8070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850940251099997e-06,
+      "loss": 1.5147,
+      "step": 8080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850472340217585e-06,
+      "loss": 1.4792,
+      "step": 8090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.850003718718794e-06,
+      "loss": 1.3599,
+      "step": 8100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.849534386745302e-06,
+      "loss": 1.4484,
+      "step": 8110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.849064344439e-06,
+      "loss": 1.5297,
+      "step": 8120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.848593591941998e-06,
+      "loss": 1.4806,
+      "step": 8130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.848122129396616e-06,
+      "loss": 1.4703,
+      "step": 8140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.847649956945392e-06,
+      "loss": 1.4077,
+      "step": 8150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.847177074731078e-06,
+      "loss": 1.4861,
+      "step": 8160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.846703482896638e-06,
+      "loss": 1.3933,
+      "step": 8170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.846229181585255e-06,
+      "loss": 1.5176,
+      "step": 8180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8457541709403224e-06,
+      "loss": 1.4208,
+      "step": 8190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.845278451105451e-06,
+      "loss": 1.4831,
+      "step": 8200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.844802022224465e-06,
+      "loss": 1.4716,
+      "step": 8210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.844324884441401e-06,
+      "loss": 1.5009,
+      "step": 8220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.843847037900513e-06,
+      "loss": 1.4764,
+      "step": 8230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.843368482746269e-06,
+      "loss": 1.3762,
+      "step": 8240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.842889219123348e-06,
+      "loss": 1.457,
+      "step": 8250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.842409247176646e-06,
+      "loss": 1.4331,
+      "step": 8260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.841928567051273e-06,
+      "loss": 1.4361,
+      "step": 8270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.841447178892552e-06,
+      "loss": 1.5784,
+      "step": 8280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.840965082846022e-06,
+      "loss": 1.4153,
+      "step": 8290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.840482279057433e-06,
+      "loss": 1.3272,
+      "step": 8300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.839998767672752e-06,
+      "loss": 1.4929,
+      "step": 8310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.839514548838158e-06,
+      "loss": 1.4922,
+      "step": 8320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8390296227000455e-06,
+      "loss": 1.5458,
+      "step": 8330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.83854398940502e-06,
+      "loss": 1.4677,
+      "step": 8340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.838057649099903e-06,
+      "loss": 1.4613,
+      "step": 8350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.837570601931731e-06,
+      "loss": 1.4793,
+      "step": 8360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.837082848047751e-06,
+      "loss": 1.4002,
+      "step": 8370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.836594387595426e-06,
+      "loss": 1.5195,
+      "step": 8380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.836105220722431e-06,
+      "loss": 1.4741,
+      "step": 8390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8356153475766564e-06,
+      "loss": 1.4661,
+      "step": 8400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.835124768306204e-06,
+      "loss": 1.5061,
+      "step": 8410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.834633483059392e-06,
+      "loss": 1.4569,
+      "step": 8420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8341414919847485e-06,
+      "loss": 1.4887,
+      "step": 8430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.833648795231018e-06,
+      "loss": 1.4068,
+      "step": 8440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.833155392947156e-06,
+      "loss": 1.4485,
+      "step": 8450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.832661285282334e-06,
+      "loss": 1.4727,
+      "step": 8460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.832166472385933e-06,
+      "loss": 1.4777,
+      "step": 8470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.831670954407551e-06,
+      "loss": 1.4672,
+      "step": 8480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.831174731496997e-06,
+      "loss": 1.4943,
+      "step": 8490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8306778038042944e-06,
+      "loss": 1.4021,
+      "step": 8500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.830180171479677e-06,
+      "loss": 1.5211,
+      "step": 8510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.829681834673597e-06,
+      "loss": 1.4146,
+      "step": 8520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.829182793536712e-06,
+      "loss": 1.5522,
+      "step": 8530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.8286830482198994e-06,
+      "loss": 1.4262,
+      "step": 8540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.828182598874246e-06,
+      "loss": 1.3862,
+      "step": 8550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.827681445651052e-06,
+      "loss": 1.5371,
+      "step": 8560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.827179588701832e-06,
+      "loss": 1.4512,
+      "step": 8570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.826677028178311e-06,
+      "loss": 1.5351,
+      "step": 8580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8261737642324264e-06,
+      "loss": 1.5501,
+      "step": 8590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8256697970163315e-06,
+      "loss": 1.4538,
+      "step": 8600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.82516512668239e-06,
+      "loss": 1.5008,
+      "step": 8610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.824659753383178e-06,
+      "loss": 1.4418,
+      "step": 8620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.824153677271485e-06,
+      "loss": 1.5636,
+      "step": 8630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.823646898500311e-06,
+      "loss": 1.3915,
+      "step": 8640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.823139417222874e-06,
+      "loss": 1.4586,
+      "step": 8650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.822631233592596e-06,
+      "loss": 1.5635,
+      "step": 8660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.822122347763117e-06,
+      "loss": 1.4216,
+      "step": 8670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.821612759888289e-06,
+      "loss": 1.5581,
+      "step": 8680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.821102470122175e-06,
+      "loss": 1.3925,
+      "step": 8690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8205914786190505e-06,
+      "loss": 1.4552,
+      "step": 8700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8200797855334025e-06,
+      "loss": 1.5122,
+      "step": 8710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.819567391019931e-06,
+      "loss": 1.4218,
+      "step": 8720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.81905429523355e-06,
+      "loss": 1.4486,
+      "step": 8730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.818540498329379e-06,
+      "loss": 1.3122,
+      "step": 8740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.818026000462757e-06,
+      "loss": 1.5279,
+      "step": 8750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.81751080178923e-06,
+      "loss": 1.5975,
+      "step": 8760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.81699490246456e-06,
+      "loss": 1.434,
+      "step": 8770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8164783026447156e-06,
+      "loss": 1.4636,
+      "step": 8780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.815961002485882e-06,
+      "loss": 1.4064,
+      "step": 8790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.815443002144453e-06,
+      "loss": 1.4814,
+      "step": 8800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.814924301777035e-06,
+      "loss": 1.5615,
+      "step": 8810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8144049015404484e-06,
+      "loss": 1.4448,
+      "step": 8820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.81388480159172e-06,
+      "loss": 1.4893,
+      "step": 8830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.813364002088093e-06,
+      "loss": 1.486,
+      "step": 8840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.812842503187021e-06,
+      "loss": 1.3684,
+      "step": 8850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.812320305046167e-06,
+      "loss": 1.5195,
+      "step": 8860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8117974078234065e-06,
+      "loss": 1.4849,
+      "step": 8870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8112738116768276e-06,
+      "loss": 1.567,
+      "step": 8880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.810749516764728e-06,
+      "loss": 1.3721,
+      "step": 8890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.810224523245618e-06,
+      "loss": 1.4407,
+      "step": 8900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.809698831278217e-06,
+      "loss": 1.4836,
+      "step": 8910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8091724410214595e-06,
+      "loss": 1.4218,
+      "step": 8920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.808645352634486e-06,
+      "loss": 1.5074,
+      "step": 8930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.808117566276653e-06,
+      "loss": 1.41,
+      "step": 8940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.807589082107524e-06,
+      "loss": 1.4514,
+      "step": 8950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8070599002868756e-06,
+      "loss": 1.4874,
+      "step": 8960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.806530020974696e-06,
+      "loss": 1.4342,
+      "step": 8970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.805999444331182e-06,
+      "loss": 1.507,
+      "step": 8980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.805468170516743e-06,
+      "loss": 1.3928,
+      "step": 8990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.804936199691997e-06,
+      "loss": 1.4909,
+      "step": 9000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8044035320177765e-06,
+      "loss": 1.4462,
+      "step": 9010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.803870167655122e-06,
+      "loss": 1.421,
+      "step": 9020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.803336106765284e-06,
+      "loss": 1.4755,
+      "step": 9030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8028013495097255e-06,
+      "loss": 1.4347,
+      "step": 9040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.80226589605012e-06,
+      "loss": 1.4311,
+      "step": 9050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.801729746548349e-06,
+      "loss": 1.4955,
+      "step": 9060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8011929011665075e-06,
+      "loss": 1.4878,
+      "step": 9070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.800655360066898e-06,
+      "loss": 1.4728,
+      "step": 9080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.800117123412036e-06,
+      "loss": 1.4124,
+      "step": 9090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.799578191364647e-06,
+      "loss": 1.433,
+      "step": 9100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.7990385640876635e-06,
+      "loss": 1.5979,
+      "step": 9110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.7984982417442325e-06,
+      "loss": 1.4203,
+      "step": 9120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.797957224497709e-06,
+      "loss": 1.5148,
+      "step": 9130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.797415512511658e-06,
+      "loss": 1.3833,
+      "step": 9140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.796873105949855e-06,
+      "loss": 1.4415,
+      "step": 9150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.796330004976285e-06,
+      "loss": 1.4852,
+      "step": 9160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.795786209755144e-06,
+      "loss": 1.3575,
+      "step": 9170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.795241720450835e-06,
+      "loss": 1.5049,
+      "step": 9180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.794696537227975e-06,
+      "loss": 1.3535,
+      "step": 9190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.794150660251389e-06,
+      "loss": 1.5128,
+      "step": 9200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.79360408968611e-06,
+      "loss": 1.509,
+      "step": 9210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7930568256973845e-06,
+      "loss": 1.3792,
+      "step": 9220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.792508868450664e-06,
+      "loss": 1.515,
+      "step": 9230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.791960218111613e-06,
+      "loss": 1.4143,
+      "step": 9240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.791410874846105e-06,
+      "loss": 1.4244,
+      "step": 9250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.790860838820222e-06,
+      "loss": 1.5038,
+      "step": 9260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.790310110200256e-06,
+      "loss": 1.3975,
+      "step": 9270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.789758689152708e-06,
+      "loss": 1.5432,
+      "step": 9280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.789206575844288e-06,
+      "loss": 1.4596,
+      "step": 9290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.788653770441919e-06,
+      "loss": 1.4248,
+      "step": 9300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.788100273112727e-06,
+      "loss": 1.6034,
+      "step": 9310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.787546084024051e-06,
+      "loss": 1.4865,
+      "step": 9320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.786991203343439e-06,
+      "loss": 1.5232,
+      "step": 9330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.786435631238649e-06,
+      "loss": 1.3967,
+      "step": 9340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.785879367877644e-06,
+      "loss": 1.5062,
+      "step": 9350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.785322413428601e-06,
+      "loss": 1.5307,
+      "step": 9360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.784764768059902e-06,
+      "loss": 1.4076,
+      "step": 9370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7842064319401405e-06,
+      "loss": 1.5345,
+      "step": 9380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.783647405238116e-06,
+      "loss": 1.452,
+      "step": 9390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.783087688122842e-06,
+      "loss": 1.4228,
+      "step": 9400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7825272807635335e-06,
+      "loss": 1.5329,
+      "step": 9410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.78196618332962e-06,
+      "loss": 1.4879,
+      "step": 9420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.781404395990738e-06,
+      "loss": 1.5342,
+      "step": 9430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.780841918916731e-06,
+      "loss": 1.3457,
+      "step": 9440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.780278752277653e-06,
+      "loss": 1.4573,
+      "step": 9450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.779714896243766e-06,
+      "loss": 1.523,
+      "step": 9460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7791503509855385e-06,
+      "loss": 1.3973,
+      "step": 9470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7785851166736505e-06,
+      "loss": 1.4844,
+      "step": 9480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.778019193478989e-06,
+      "loss": 1.4181,
+      "step": 9490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.777452581572649e-06,
+      "loss": 1.364,
+      "step": 9500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.776885281125933e-06,
+      "loss": 1.4752,
+      "step": 9510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.776317292310352e-06,
+      "loss": 1.4313,
+      "step": 9520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.775748615297628e-06,
+      "loss": 1.4929,
+      "step": 9530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7751792502596846e-06,
+      "loss": 1.4038,
+      "step": 9540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.774609197368662e-06,
+      "loss": 1.4285,
+      "step": 9550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.774038456796901e-06,
+      "loss": 1.4323,
+      "step": 9560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7734670287169545e-06,
+      "loss": 1.4111,
+      "step": 9570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.772894913301581e-06,
+      "loss": 1.4646,
+      "step": 9580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.772322110723748e-06,
+      "loss": 1.4598,
+      "step": 9590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.771748621156631e-06,
+      "loss": 1.4899,
+      "step": 9600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7711744447736115e-06,
+      "loss": 1.5134,
+      "step": 9610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7705995817482805e-06,
+      "loss": 1.4929,
+      "step": 9620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.770024032254435e-06,
+      "loss": 1.5202,
+      "step": 9630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.769447796466082e-06,
+      "loss": 1.4392,
+      "step": 9640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.768870874557434e-06,
+      "loss": 1.429,
+      "step": 9650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.76829326670291e-06,
+      "loss": 1.4761,
+      "step": 9660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.7677149730771386e-06,
+      "loss": 1.4322,
+      "step": 9670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.767135993854955e-06,
+      "loss": 1.479,
+      "step": 9680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.766556329211401e-06,
+      "loss": 1.4132,
+      "step": 9690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.765975979321727e-06,
+      "loss": 1.4413,
+      "step": 9700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.765394944361389e-06,
+      "loss": 1.5509,
+      "step": 9710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.76481322450605e-06,
+      "loss": 1.3931,
+      "step": 9720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.764230819931583e-06,
+      "loss": 1.4367,
+      "step": 9730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7636477308140645e-06,
+      "loss": 1.4078,
+      "step": 9740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.763063957329779e-06,
+      "loss": 1.4419,
+      "step": 9750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.76247949965522e-06,
+      "loss": 1.6089,
+      "step": 9760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.761894357967084e-06,
+      "loss": 1.4693,
+      "step": 9770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.761308532442278e-06,
+      "loss": 1.5091,
+      "step": 9780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.760722023257913e-06,
+      "loss": 1.5197,
+      "step": 9790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7601348305913085e-06,
+      "loss": 1.4495,
+      "step": 9800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75954695461999e-06,
+      "loss": 1.5167,
+      "step": 9810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75895839552169e-06,
+      "loss": 1.4794,
+      "step": 9820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7583691534743455e-06,
+      "loss": 1.4987,
+      "step": 9830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.757779228656103e-06,
+      "loss": 1.4688,
+      "step": 9840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.757188621245314e-06,
+      "loss": 1.3934,
+      "step": 9850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.756597331420536e-06,
+      "loss": 1.5041,
+      "step": 9860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.756005359360532e-06,
+      "loss": 1.4246,
+      "step": 9870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.755412705244274e-06,
+      "loss": 1.4242,
+      "step": 9880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7548193692509395e-06,
+      "loss": 1.405,
+      "step": 9890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.754225351559909e-06,
+      "loss": 1.3972,
+      "step": 9900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.753630652350771e-06,
+      "loss": 1.5097,
+      "step": 9910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.753035271803323e-06,
+      "loss": 1.4094,
+      "step": 9920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.752439210097566e-06,
+      "loss": 1.5039,
+      "step": 9930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7518424674137036e-06,
+      "loss": 1.4983,
+      "step": 9940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.751245043932151e-06,
+      "loss": 1.4538,
+      "step": 9950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.750646939833526e-06,
+      "loss": 1.4959,
+      "step": 9960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.750048155298653e-06,
+      "loss": 1.4386,
+      "step": 9970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.749448690508562e-06,
+      "loss": 1.5585,
+      "step": 9980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74884854564449e-06,
+      "loss": 1.442,
+      "step": 9990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.748247720887876e-06,
+      "loss": 1.4516,
+      "step": 10000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.747646216420368e-06,
+      "loss": 1.5292,
+      "step": 10010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74704403242382e-06,
+      "loss": 1.4254,
+      "step": 10020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.746441169080287e-06,
+      "loss": 1.5433,
+      "step": 10030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.745837626572033e-06,
+      "loss": 1.4437,
+      "step": 10040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.745233405081528e-06,
+      "loss": 1.4165,
+      "step": 10050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.744628504791443e-06,
+      "loss": 1.4137,
+      "step": 10060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74402292588466e-06,
+      "loss": 1.4314,
+      "step": 10070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74341666854426e-06,
+      "loss": 1.4754,
+      "step": 10080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.742809732953536e-06,
+      "loss": 1.4749,
+      "step": 10090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.74220211929598e-06,
+      "loss": 1.4107,
+      "step": 10100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.741593827755292e-06,
+      "loss": 1.495,
+      "step": 10110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.740984858515376e-06,
+      "loss": 1.4801,
+      "step": 10120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.740375211760341e-06,
+      "loss": 1.4785,
+      "step": 10130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.739764887674501e-06,
+      "loss": 1.3998,
+      "step": 10140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.739153886442376e-06,
+      "loss": 1.4052,
+      "step": 10150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7385422082486874e-06,
+      "loss": 1.4653,
+      "step": 10160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.737929853278365e-06,
+      "loss": 1.4913,
+      "step": 10170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.737316821716541e-06,
+      "loss": 1.4994,
+      "step": 10180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.736703113748553e-06,
+      "loss": 1.428,
+      "step": 10190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.736088729559943e-06,
+      "loss": 1.4326,
+      "step": 10200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.735473669336457e-06,
+      "loss": 1.516,
+      "step": 10210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.734857933264045e-06,
+      "loss": 1.2812,
+      "step": 10220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.7342415215288625e-06,
+      "loss": 1.5292,
+      "step": 10230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.733624434317269e-06,
+      "loss": 1.4038,
+      "step": 10240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.733006671815828e-06,
+      "loss": 1.4555,
+      "step": 10250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.732388234211306e-06,
+      "loss": 1.4831,
+      "step": 10260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.731769121690677e-06,
+      "loss": 1.4112,
+      "step": 10270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.731149334441114e-06,
+      "loss": 1.4546,
+      "step": 10280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.730528872649999e-06,
+      "loss": 1.441,
+      "step": 10290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.729907736504915e-06,
+      "loss": 1.4369,
+      "step": 10300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.729285926193649e-06,
+      "loss": 1.4337,
+      "step": 10310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.728663441904194e-06,
+      "loss": 1.3928,
+      "step": 10320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.728040283824743e-06,
+      "loss": 1.4929,
+      "step": 10330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.727416452143696e-06,
+      "loss": 1.3554,
+      "step": 10340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7267919470496555e-06,
+      "loss": 1.4696,
+      "step": 10350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7261667687314285e-06,
+      "loss": 1.5638,
+      "step": 10360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.725540917378023e-06,
+      "loss": 1.4949,
+      "step": 10370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.724914393178652e-06,
+      "loss": 1.4108,
+      "step": 10380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.724287196322734e-06,
+      "loss": 1.4031,
+      "step": 10390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.723659326999888e-06,
+      "loss": 1.3597,
+      "step": 10400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7230307853999365e-06,
+      "loss": 1.4776,
+      "step": 10410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.722401571712907e-06,
+      "loss": 1.4592,
+      "step": 10420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.721771686129027e-06,
+      "loss": 1.473,
+      "step": 10430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.721141128838733e-06,
+      "loss": 1.4534,
+      "step": 10440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.720509900032658e-06,
+      "loss": 1.4185,
+      "step": 10450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.719877999901642e-06,
+      "loss": 1.4631,
+      "step": 10460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.719245428636725e-06,
+      "loss": 1.3607,
+      "step": 10470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.718612186429156e-06,
+      "loss": 1.4643,
+      "step": 10480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7179782734703776e-06,
+      "loss": 1.464,
+      "step": 10490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.717343689952043e-06,
+      "loss": 1.4216,
+      "step": 10500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.716708436066006e-06,
+      "loss": 1.4537,
+      "step": 10510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.71607251200432e-06,
+      "loss": 1.3918,
+      "step": 10520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.715435917959245e-06,
+      "loss": 1.5242,
+      "step": 10530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.714798654123241e-06,
+      "loss": 1.4405,
+      "step": 10540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.714160720688973e-06,
+      "loss": 1.481,
+      "step": 10550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.713522117849305e-06,
+      "loss": 1.5623,
+      "step": 10560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.712882845797307e-06,
+      "loss": 1.3354,
+      "step": 10570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.712242904726248e-06,
+      "loss": 1.4864,
+      "step": 10580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.711602294829602e-06,
+      "loss": 1.4281,
+      "step": 10590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7109610163010435e-06,
+      "loss": 1.4084,
+      "step": 10600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.71031906933445e-06,
+      "loss": 1.5513,
+      "step": 10610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.709676454123901e-06,
+      "loss": 1.4858,
+      "step": 10620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.709033170863677e-06,
+      "loss": 1.4371,
+      "step": 10630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.708389219748263e-06,
+      "loss": 1.5354,
+      "step": 10640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.707744600972341e-06,
+      "loss": 1.4388,
+      "step": 10650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.707099314730802e-06,
+      "loss": 1.5422,
+      "step": 10660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.706453361218732e-06,
+      "loss": 1.3845,
+      "step": 10670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.7058067406314235e-06,
+      "loss": 1.5511,
+      "step": 10680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.705159453164367e-06,
+      "loss": 1.4634,
+      "step": 10690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.704511499013258e-06,
+      "loss": 1.4281,
+      "step": 10700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.703862878373991e-06,
+      "loss": 1.4984,
+      "step": 10710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.703213591442663e-06,
+      "loss": 1.4267,
+      "step": 10720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.702563638415571e-06,
+      "loss": 1.4941,
+      "step": 10730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.701913019489218e-06,
+      "loss": 1.4001,
+      "step": 10740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.701261734860303e-06,
+      "loss": 1.4955,
+      "step": 10750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.70060978472573e-06,
+      "loss": 1.4492,
+      "step": 10760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.699957169282599e-06,
+      "loss": 1.4908,
+      "step": 10770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6993038887282185e-06,
+      "loss": 1.4929,
+      "step": 10780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6986499432600914e-06,
+      "loss": 1.4487,
+      "step": 10790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.697995333075927e-06,
+      "loss": 1.4442,
+      "step": 10800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6973400583736315e-06,
+      "loss": 1.4703,
+      "step": 10810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6966841193513145e-06,
+      "loss": 1.3521,
+      "step": 10820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.696027516207284e-06,
+      "loss": 1.4954,
+      "step": 10830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.695370249140052e-06,
+      "loss": 1.4025,
+      "step": 10840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.694712318348329e-06,
+      "loss": 1.4744,
+      "step": 10850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.694053724031027e-06,
+      "loss": 1.5408,
+      "step": 10860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.693394466387256e-06,
+      "loss": 1.427,
+      "step": 10870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.692734545616331e-06,
+      "loss": 1.4471,
+      "step": 10880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.6920739619177655e-06,
+      "loss": 1.4722,
+      "step": 10890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6914127154912725e-06,
+      "loss": 1.3492,
+      "step": 10900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.690750806536766e-06,
+      "loss": 1.4061,
+      "step": 10910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6900882352543614e-06,
+      "loss": 1.514,
+      "step": 10920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6894250018443715e-06,
+      "loss": 1.3756,
+      "step": 10930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.688761106507313e-06,
+      "loss": 1.3968,
+      "step": 10940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.688096549443899e-06,
+      "loss": 1.4047,
+      "step": 10950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.687431330855046e-06,
+      "loss": 1.5932,
+      "step": 10960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.686765450941869e-06,
+      "loss": 1.4495,
+      "step": 10970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.686098909905682e-06,
+      "loss": 1.4887,
+      "step": 10980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.685431707947999e-06,
+      "loss": 1.3135,
+      "step": 10990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.684763845270537e-06,
+      "loss": 1.4363,
+      "step": 11000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.684095322075209e-06,
+      "loss": 1.5227,
+      "step": 11010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6834261385641295e-06,
+      "loss": 1.388,
+      "step": 11020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6827562949396115e-06,
+      "loss": 1.4697,
+      "step": 11030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.682085791404169e-06,
+      "loss": 1.4344,
+      "step": 11040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.681414628160514e-06,
+      "loss": 1.3837,
+      "step": 11050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.68074280541156e-06,
+      "loss": 1.484,
+      "step": 11060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.680070323360418e-06,
+      "loss": 1.3928,
+      "step": 11070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6793971822103985e-06,
+      "loss": 1.4346,
+      "step": 11080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.678723382165012e-06,
+      "loss": 1.529,
+      "step": 11090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.678048923427968e-06,
+      "loss": 1.4555,
+      "step": 11100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.677373806203175e-06,
+      "loss": 1.5154,
+      "step": 11110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.676698030694741e-06,
+      "loss": 1.4629,
+      "step": 11120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.676021597106971e-06,
+      "loss": 1.4681,
+      "step": 11130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6753445056443735e-06,
+      "loss": 1.4179,
+      "step": 11140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.674666756511651e-06,
+      "loss": 1.3941,
+      "step": 11150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.673988349913707e-06,
+      "loss": 1.4219,
+      "step": 11160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.673309286055644e-06,
+      "loss": 1.4328,
+      "step": 11170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.672629565142762e-06,
+      "loss": 1.4539,
+      "step": 11180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6719491873805625e-06,
+      "loss": 1.419,
+      "step": 11190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.671268152974742e-06,
+      "loss": 1.4178,
+      "step": 11200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.670586462131196e-06,
+      "loss": 1.533,
+      "step": 11210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.669904115056022e-06,
+      "loss": 1.4374,
+      "step": 11220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.669221111955512e-06,
+      "loss": 1.5171,
+      "step": 11230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.668537453036159e-06,
+      "loss": 1.4236,
+      "step": 11240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.66785313850465e-06,
+      "loss": 1.4209,
+      "step": 11250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.667168168567877e-06,
+      "loss": 1.5589,
+      "step": 11260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6664825434329245e-06,
+      "loss": 1.473,
+      "step": 11270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6657962633070765e-06,
+      "loss": 1.5097,
+      "step": 11280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.665109328397816e-06,
+      "loss": 1.3739,
+      "step": 11290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.664421738912825e-06,
+      "loss": 1.3294,
+      "step": 11300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.66373349505998e-06,
+      "loss": 1.533,
+      "step": 11310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6630445970473576e-06,
+      "loss": 1.3542,
+      "step": 11320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.662355045083232e-06,
+      "loss": 1.4368,
+      "step": 11330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.661664839376075e-06,
+      "loss": 1.4149,
+      "step": 11340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.660973980134555e-06,
+      "loss": 1.4503,
+      "step": 11350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.660282467567541e-06,
+      "loss": 1.4588,
+      "step": 11360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6595903018840946e-06,
+      "loss": 1.4454,
+      "step": 11370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.65889748329348e-06,
+      "loss": 1.5926,
+      "step": 11380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.658204012005156e-06,
+      "loss": 1.466,
+      "step": 11390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.657509888228779e-06,
+      "loss": 1.3763,
+      "step": 11400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6568151121742014e-06,
+      "loss": 1.4527,
+      "step": 11410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.656119684051477e-06,
+      "loss": 1.3513,
+      "step": 11420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.655423604070851e-06,
+      "loss": 1.482,
+      "step": 11430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.654726872442771e-06,
+      "loss": 1.3483,
+      "step": 11440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6540294893778796e-06,
+      "loss": 1.4597,
+      "step": 11450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.653331455087014e-06,
+      "loss": 1.4656,
+      "step": 11460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.652632769781212e-06,
+      "loss": 1.4198,
+      "step": 11470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.6519334336717055e-06,
+      "loss": 1.4881,
+      "step": 11480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.651233446969925e-06,
+      "loss": 1.3137,
+      "step": 11490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.650532809887496e-06,
+      "loss": 1.5084,
+      "step": 11500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.649831522636243e-06,
+      "loss": 1.5879,
+      "step": 11510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.649129585428184e-06,
+      "loss": 1.3676,
+      "step": 11520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.648426998475536e-06,
+      "loss": 1.4059,
+      "step": 11530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6477237619907114e-06,
+      "loss": 1.4702,
+      "step": 11540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6470198761863195e-06,
+      "loss": 1.401,
+      "step": 11550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.646315341275164e-06,
+      "loss": 1.4313,
+      "step": 11560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6456101574702485e-06,
+      "loss": 1.4919,
+      "step": 11570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.644904324984769e-06,
+      "loss": 1.4394,
+      "step": 11580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.64419784403212e-06,
+      "loss": 1.4812,
+      "step": 11590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6434907148258915e-06,
+      "loss": 1.477,
+      "step": 11600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.642782937579869e-06,
+      "loss": 1.5019,
+      "step": 11610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.642074512508033e-06,
+      "loss": 1.4189,
+      "step": 11620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6413654398245635e-06,
+      "loss": 1.5266,
+      "step": 11630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6406557197438324e-06,
+      "loss": 1.4062,
+      "step": 11640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.639945352480409e-06,
+      "loss": 1.4235,
+      "step": 11650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6392343382490576e-06,
+      "loss": 1.4699,
+      "step": 11660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.638522677264739e-06,
+      "loss": 1.3462,
+      "step": 11670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.637810369742609e-06,
+      "loss": 1.4667,
+      "step": 11680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.63709741589802e-06,
+      "loss": 1.488,
+      "step": 11690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.636383815946517e-06,
+      "loss": 1.5291,
+      "step": 11700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.635669570103843e-06,
+      "loss": 1.5018,
+      "step": 11710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.634954678585935e-06,
+      "loss": 1.449,
+      "step": 11720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.634239141608926e-06,
+      "loss": 1.5008,
+      "step": 11730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.633522959389144e-06,
+      "loss": 1.4289,
+      "step": 11740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.63280613214311e-06,
+      "loss": 1.4265,
+      "step": 11750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.632088660087545e-06,
+      "loss": 1.4982,
+      "step": 11760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.631370543439358e-06,
+      "loss": 1.4815,
+      "step": 11770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.63065178241566e-06,
+      "loss": 1.4499,
+      "step": 11780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.629932377233751e-06,
+      "loss": 1.4305,
+      "step": 11790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.629212328111129e-06,
+      "loss": 1.3837,
+      "step": 11800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.628491635265486e-06,
+      "loss": 1.4622,
+      "step": 11810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.627770298914709e-06,
+      "loss": 1.3913,
+      "step": 11820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.627048319276878e-06,
+      "loss": 1.4981,
+      "step": 11830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6263256965702684e-06,
+      "loss": 1.3994,
+      "step": 11840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.62560243101335e-06,
+      "loss": 1.4614,
+      "step": 11850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.62487852282479e-06,
+      "loss": 1.5018,
+      "step": 11860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.624153972223444e-06,
+      "loss": 1.4005,
+      "step": 11870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.623428779428365e-06,
+      "loss": 1.5353,
+      "step": 11880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6227029446588e-06,
+      "loss": 1.4428,
+      "step": 11890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.621976468134192e-06,
+      "loss": 1.449,
+      "step": 11900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6212493500741725e-06,
+      "loss": 1.4651,
+      "step": 11910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.620521590698574e-06,
+      "loss": 1.4757,
+      "step": 11920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6197931902274185e-06,
+      "loss": 1.5054,
+      "step": 11930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6190641488809205e-06,
+      "loss": 1.4491,
+      "step": 11940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6183344668794936e-06,
+      "loss": 1.4626,
+      "step": 11950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.61760414444374e-06,
+      "loss": 1.5537,
+      "step": 11960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.616873181794458e-06,
+      "loss": 1.4706,
+      "step": 11970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.616141579152639e-06,
+      "loss": 1.4632,
+      "step": 11980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.615409336739467e-06,
+      "loss": 1.5207,
+      "step": 11990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.614676454776321e-06,
+      "loss": 1.4862,
+      "step": 12000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.613942933484774e-06,
+      "loss": 1.6014,
+      "step": 12010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6132087730865885e-06,
+      "loss": 1.3666,
+      "step": 12020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.612473973803724e-06,
+      "loss": 1.4767,
+      "step": 12030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6117385358583315e-06,
+      "loss": 1.4115,
+      "step": 12040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.6110024594727555e-06,
+      "loss": 1.4312,
+      "step": 12050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.610265744869532e-06,
+      "loss": 1.5334,
+      "step": 12060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.609528392271393e-06,
+      "loss": 1.4075,
+      "step": 12070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.608790401901262e-06,
+      "loss": 1.4618,
+      "step": 12080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.608051773982253e-06,
+      "loss": 1.4179,
+      "step": 12090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.607312508737676e-06,
+      "loss": 1.3823,
+      "step": 12100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.606572606391033e-06,
+      "loss": 1.5312,
+      "step": 12110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.605832067166017e-06,
+      "loss": 1.3701,
+      "step": 12120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.605090891286515e-06,
+      "loss": 1.5462,
+      "step": 12130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.604349078976607e-06,
+      "loss": 1.4484,
+      "step": 12140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.603606630460562e-06,
+      "loss": 1.4404,
+      "step": 12150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.6028635459628476e-06,
+      "loss": 1.5111,
+      "step": 12160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.602119825708117e-06,
+      "loss": 1.3938,
+      "step": 12170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.601375469921219e-06,
+      "loss": 1.5212,
+      "step": 12180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.600630478827195e-06,
+      "loss": 1.4783,
+      "step": 12190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.599884852651277e-06,
+      "loss": 1.4754,
+      "step": 12200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.599138591618889e-06,
+      "loss": 1.4992,
+      "step": 12210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.598391695955649e-06,
+      "loss": 1.4227,
+      "step": 12220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.597644165887364e-06,
+      "loss": 1.5937,
+      "step": 12230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.596896001640034e-06,
+      "loss": 1.3707,
+      "step": 12240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5961472034398525e-06,
+      "loss": 1.4148,
+      "step": 12250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.595397771513202e-06,
+      "loss": 1.5245,
+      "step": 12260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.594647706086657e-06,
+      "loss": 1.3966,
+      "step": 12270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.593897007386985e-06,
+      "loss": 1.4597,
+      "step": 12280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.593145675641144e-06,
+      "loss": 1.3901,
+      "step": 12290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.592393711076283e-06,
+      "loss": 1.3784,
+      "step": 12300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.591641113919744e-06,
+      "loss": 1.5336,
+      "step": 12310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.590887884399058e-06,
+      "loss": 1.42,
+      "step": 12320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.590134022741949e-06,
+      "loss": 1.5017,
+      "step": 12330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.58937952917633e-06,
+      "loss": 1.4091,
+      "step": 12340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.588624403930308e-06,
+      "loss": 1.4397,
+      "step": 12350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.587868647232179e-06,
+      "loss": 1.5457,
+      "step": 12360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.58711225931043e-06,
+      "loss": 1.4078,
+      "step": 12370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.586355240393739e-06,
+      "loss": 1.4986,
+      "step": 12380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.585597590710976e-06,
+      "loss": 1.3852,
+      "step": 12390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.584839310491199e-06,
+      "loss": 1.4327,
+      "step": 12400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5840803999636605e-06,
+      "loss": 1.4665,
+      "step": 12410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.583320859357798e-06,
+      "loss": 1.4458,
+      "step": 12420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.582560688903246e-06,
+      "loss": 1.5489,
+      "step": 12430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.581799888829824e-06,
+      "loss": 1.3743,
+      "step": 12440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.581038459367545e-06,
+      "loss": 1.3847,
+      "step": 12450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.580276400746611e-06,
+      "loss": 1.5495,
+      "step": 12460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5795137131974155e-06,
+      "loss": 1.5343,
+      "step": 12470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5787503969505405e-06,
+      "loss": 1.5398,
+      "step": 12480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.577986452236758e-06,
+      "loss": 1.4465,
+      "step": 12490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.577221879287033e-06,
+      "loss": 1.3895,
+      "step": 12500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5764566783325145e-06,
+      "loss": 1.4864,
+      "step": 12510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.57569084960455e-06,
+      "loss": 1.4422,
+      "step": 12520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.574924393334667e-06,
+      "loss": 1.4358,
+      "step": 12530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.574157309754591e-06,
+      "loss": 1.4338,
+      "step": 12540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.573389599096232e-06,
+      "loss": 1.3257,
+      "step": 12550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.572621261591692e-06,
+      "loss": 1.4896,
+      "step": 12560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.571852297473263e-06,
+      "loss": 1.3736,
+      "step": 12570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.571082706973422e-06,
+      "loss": 1.5483,
+      "step": 12580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.570312490324842e-06,
+      "loss": 1.4751,
+      "step": 12590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.569541647760381e-06,
+      "loss": 1.4221,
+      "step": 12600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.5687701795130865e-06,
+      "loss": 1.4936,
+      "step": 12610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.567998085816196e-06,
+      "loss": 1.4779,
+      "step": 12620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.567225366903136e-06,
+      "loss": 1.5692,
+      "step": 12630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.566452023007523e-06,
+      "loss": 1.4899,
+      "step": 12640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.56567805436316e-06,
+      "loss": 1.5022,
+      "step": 12650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.564903461204041e-06,
+      "loss": 1.575,
+      "step": 12660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.564128243764347e-06,
+      "loss": 1.434,
+      "step": 12670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.563352402278451e-06,
+      "loss": 1.4817,
+      "step": 12680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.562575936980912e-06,
+      "loss": 1.304,
+      "step": 12690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5617988481064764e-06,
+      "loss": 1.3839,
+      "step": 12700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5610211358900836e-06,
+      "loss": 1.4677,
+      "step": 12710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.560242800566855e-06,
+      "loss": 1.4309,
+      "step": 12720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5594638423721076e-06,
+      "loss": 1.5099,
+      "step": 12730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.558684261541342e-06,
+      "loss": 1.4264,
+      "step": 12740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.557904058310247e-06,
+      "loss": 1.4191,
+      "step": 12750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.557123232914702e-06,
+      "loss": 1.4823,
+      "step": 12760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5563417855907735e-06,
+      "loss": 1.4763,
+      "step": 12770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5555597165747144e-06,
+      "loss": 1.4929,
+      "step": 12780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.554777026102969e-06,
+      "loss": 1.3592,
+      "step": 12790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.553993714412166e-06,
+      "loss": 1.3341,
+      "step": 12800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.553209781739124e-06,
+      "loss": 1.5586,
+      "step": 12810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.552425228320849e-06,
+      "loss": 1.4163,
+      "step": 12820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.551640054394533e-06,
+      "loss": 1.5203,
+      "step": 12830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.550854260197559e-06,
+      "loss": 1.3976,
+      "step": 12840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.550067845967494e-06,
+      "loss": 1.3848,
+      "step": 12850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.549280811942094e-06,
+      "loss": 1.5152,
+      "step": 12860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.548493158359304e-06,
+      "loss": 1.4605,
+      "step": 12870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5477048854572524e-06,
+      "loss": 1.476,
+      "step": 12880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.546915993474259e-06,
+      "loss": 1.351,
+      "step": 12890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.546126482648827e-06,
+      "loss": 1.4267,
+      "step": 12900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5453363532196504e-06,
+      "loss": 1.5225,
+      "step": 12910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.544545605425607e-06,
+      "loss": 1.3407,
+      "step": 12920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.543754239505764e-06,
+      "loss": 1.3568,
+      "step": 12930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.542962255699374e-06,
+      "loss": 1.4447,
+      "step": 12940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.542169654245877e-06,
+      "loss": 1.3843,
+      "step": 12950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.541376435384899e-06,
+      "loss": 1.4447,
+      "step": 12960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.540582599356254e-06,
+      "loss": 1.43,
+      "step": 12970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.53978814639994e-06,
+      "loss": 1.4385,
+      "step": 12980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.538993076756145e-06,
+      "loss": 1.4194,
+      "step": 12990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.538197390665242e-06,
+      "loss": 1.4155,
+      "step": 13000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.537401088367789e-06,
+      "loss": 1.532,
+      "step": 13010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.536604170104533e-06,
+      "loss": 1.4906,
+      "step": 13020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.535806636116403e-06,
+      "loss": 1.5545,
+      "step": 13030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5350084866445195e-06,
+      "loss": 1.4321,
+      "step": 13040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.534209721930184e-06,
+      "loss": 1.4312,
+      "step": 13050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.533410342214889e-06,
+      "loss": 1.5174,
+      "step": 13060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5326103477403085e-06,
+      "loss": 1.3833,
+      "step": 13070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.531809738748304e-06,
+      "loss": 1.4534,
+      "step": 13080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.531008515480925e-06,
+      "loss": 1.4193,
+      "step": 13090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.530206678180401e-06,
+      "loss": 1.4003,
+      "step": 13100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.529404227089155e-06,
+      "loss": 1.5472,
+      "step": 13110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.528601162449788e-06,
+      "loss": 1.4318,
+      "step": 13120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.527797484505093e-06,
+      "loss": 1.5826,
+      "step": 13130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.526993193498043e-06,
+      "loss": 1.3753,
+      "step": 13140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.526188289671798e-06,
+      "loss": 1.4201,
+      "step": 13150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.525382773269706e-06,
+      "loss": 1.4343,
+      "step": 13160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.524576644535298e-06,
+      "loss": 1.3204,
+      "step": 13170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.523769903712289e-06,
+      "loss": 1.4736,
+      "step": 13180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.522962551044581e-06,
+      "loss": 1.4821,
+      "step": 13190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.52215458677626e-06,
+      "loss": 1.3992,
+      "step": 13200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.521346011151597e-06,
+      "loss": 1.5604,
+      "step": 13210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.520536824415049e-06,
+      "loss": 1.4619,
+      "step": 13220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.519727026811255e-06,
+      "loss": 1.5872,
+      "step": 13230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.518916618585042e-06,
+      "loss": 1.4275,
+      "step": 13240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.5181055999814195e-06,
+      "loss": 1.4068,
+      "step": 13250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.517293971245582e-06,
+      "loss": 1.4736,
+      "step": 13260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.516481732622909e-06,
+      "loss": 1.3997,
+      "step": 13270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.515668884358963e-06,
+      "loss": 1.5272,
+      "step": 13280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.514855426699493e-06,
+      "loss": 1.4016,
+      "step": 13290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.51404135989043e-06,
+      "loss": 1.379,
+      "step": 13300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.513226684177891e-06,
+      "loss": 1.5764,
+      "step": 13310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.512411399808175e-06,
+      "loss": 1.4634,
+      "step": 13320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5115955070277675e-06,
+      "loss": 1.4637,
+      "step": 13330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5107790060833366e-06,
+      "loss": 1.4578,
+      "step": 13340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.509961897221733e-06,
+      "loss": 1.4785,
+      "step": 13350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.509144180689995e-06,
+      "loss": 1.5503,
+      "step": 13360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.508325856735339e-06,
+      "loss": 1.4013,
+      "step": 13370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.507506925605171e-06,
+      "loss": 1.5865,
+      "step": 13380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.506687387547076e-06,
+      "loss": 1.3347,
+      "step": 13390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.505867242808826e-06,
+      "loss": 1.4062,
+      "step": 13400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5050464916383726e-06,
+      "loss": 1.4763,
+      "step": 13410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.504225134283855e-06,
+      "loss": 1.3943,
+      "step": 13420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.503403170993591e-06,
+      "loss": 1.5227,
+      "step": 13430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.5025806020160865e-06,
+      "loss": 1.5048,
+      "step": 13440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.501757427600027e-06,
+      "loss": 1.346,
+      "step": 13450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.500933647994281e-06,
+      "loss": 1.4852,
+      "step": 13460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.500109263447904e-06,
+      "loss": 1.394,
+      "step": 13470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.499284274210128e-06,
+      "loss": 1.4958,
+      "step": 13480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.498458680530375e-06,
+      "loss": 1.3193,
+      "step": 13490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.497632482658243e-06,
+      "loss": 1.3997,
+      "step": 13500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.496805680843518e-06,
+      "loss": 1.4923,
+      "step": 13510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.495978275336164e-06,
+      "loss": 1.4663,
+      "step": 13520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.495150266386331e-06,
+      "loss": 1.4908,
+      "step": 13530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4943216542443515e-06,
+      "loss": 1.4604,
+      "step": 13540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4934924391607385e-06,
+      "loss": 1.4379,
+      "step": 13550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.492662621386186e-06,
+      "loss": 1.4633,
+      "step": 13560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.491832201171574e-06,
+      "loss": 1.3759,
+      "step": 13570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.491001178767962e-06,
+      "loss": 1.5167,
+      "step": 13580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.490169554426593e-06,
+      "loss": 1.4657,
+      "step": 13590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.489337328398891e-06,
+      "loss": 1.4307,
+      "step": 13600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.488504500936462e-06,
+      "loss": 1.478,
+      "step": 13610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.487671072291094e-06,
+      "loss": 1.3896,
+      "step": 13620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.486837042714757e-06,
+      "loss": 1.4809,
+      "step": 13630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.486002412459603e-06,
+      "loss": 1.4204,
+      "step": 13640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.485167181777964e-06,
+      "loss": 1.4974,
+      "step": 13650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.484331350922355e-06,
+      "loss": 1.5069,
+      "step": 13660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.483494920145473e-06,
+      "loss": 1.4762,
+      "step": 13670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.482657889700194e-06,
+      "loss": 1.5327,
+      "step": 13680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.481820259839577e-06,
+      "loss": 1.4943,
+      "step": 13690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4809820308168635e-06,
+      "loss": 1.4164,
+      "step": 13700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.480143202885473e-06,
+      "loss": 1.4815,
+      "step": 13710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.479303776299008e-06,
+      "loss": 1.3573,
+      "step": 13720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4784637513112515e-06,
+      "loss": 1.5231,
+      "step": 13730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.477623128176169e-06,
+      "loss": 1.3806,
+      "step": 13740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.476781907147903e-06,
+      "loss": 1.3836,
+      "step": 13750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4759400884807805e-06,
+      "loss": 1.4061,
+      "step": 13760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.475097672429307e-06,
+      "loss": 1.4315,
+      "step": 13770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.474254659248172e-06,
+      "loss": 1.4171,
+      "step": 13780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.47341104919224e-06,
+      "loss": 1.4359,
+      "step": 13790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4725668425165605e-06,
+      "loss": 1.4094,
+      "step": 13800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.47172203947636e-06,
+      "loss": 1.5126,
+      "step": 13810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.470876640327051e-06,
+      "loss": 1.4291,
+      "step": 13820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.4700306453242185e-06,
+      "loss": 1.5151,
+      "step": 13830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.469184054723632e-06,
+      "loss": 1.4476,
+      "step": 13840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4683368687812434e-06,
+      "loss": 1.4743,
+      "step": 13850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.467489087753178e-06,
+      "loss": 1.4806,
+      "step": 13860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.466640711895747e-06,
+      "loss": 1.3397,
+      "step": 13870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.465791741465439e-06,
+      "loss": 1.479,
+      "step": 13880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.464942176718922e-06,
+      "loss": 1.4427,
+      "step": 13890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.464092017913044e-06,
+      "loss": 1.2999,
+      "step": 13900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4632412653048344e-06,
+      "loss": 1.4645,
+      "step": 13910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.462389919151498e-06,
+      "loss": 1.3657,
+      "step": 13920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.461537979710424e-06,
+      "loss": 1.5118,
+      "step": 13930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4606854472391764e-06,
+      "loss": 1.3824,
+      "step": 13940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.459832321995502e-06,
+      "loss": 1.4039,
+      "step": 13950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.458978604237325e-06,
+      "loss": 1.5138,
+      "step": 13960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.45812429422275e-06,
+      "loss": 1.4406,
+      "step": 13970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4572693922100595e-06,
+      "loss": 1.4272,
+      "step": 13980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.456413898457714e-06,
+      "loss": 1.3803,
+      "step": 13990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.455557813224356e-06,
+      "loss": 1.4632,
+      "step": 14000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.454701136768804e-06,
+      "loss": 1.4623,
+      "step": 14010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.453843869350058e-06,
+      "loss": 1.3426,
+      "step": 14020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.452986011227292e-06,
+      "loss": 1.5597,
+      "step": 14030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.452127562659864e-06,
+      "loss": 1.5276,
+      "step": 14040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4512685239073064e-06,
+      "loss": 1.4508,
+      "step": 14050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.450408895229332e-06,
+      "loss": 1.5207,
+      "step": 14060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.449548676885833e-06,
+      "loss": 1.4547,
+      "step": 14070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.448687869136878e-06,
+      "loss": 1.5429,
+      "step": 14080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.447826472242713e-06,
+      "loss": 1.4236,
+      "step": 14090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.446964486463765e-06,
+      "loss": 1.4188,
+      "step": 14100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.446101912060636e-06,
+      "loss": 1.4515,
+      "step": 14110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.445238749294109e-06,
+      "loss": 1.4319,
+      "step": 14120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.444374998425141e-06,
+      "loss": 1.5133,
+      "step": 14130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.443510659714873e-06,
+      "loss": 1.4158,
+      "step": 14140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.442645733424616e-06,
+      "loss": 1.3736,
+      "step": 14150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.441780219815863e-06,
+      "loss": 1.4906,
+      "step": 14160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.440914119150287e-06,
+      "loss": 1.4137,
+      "step": 14170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.440047431689732e-06,
+      "loss": 1.5065,
+      "step": 14180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.439180157696224e-06,
+      "loss": 1.4045,
+      "step": 14190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.438312297431967e-06,
+      "loss": 1.4641,
+      "step": 14200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.437443851159338e-06,
+      "loss": 1.5064,
+      "step": 14210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.436574819140894e-06,
+      "loss": 1.3758,
+      "step": 14220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4357052016393695e-06,
+      "loss": 1.4754,
+      "step": 14230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.434834998917677e-06,
+      "loss": 1.4257,
+      "step": 14240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.433964211238901e-06,
+      "loss": 1.3709,
+      "step": 14250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.433092838866307e-06,
+      "loss": 1.5432,
+      "step": 14260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4322208820633376e-06,
+      "loss": 1.4285,
+      "step": 14270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.431348341093608e-06,
+      "loss": 1.4456,
+      "step": 14280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.4304752162209164e-06,
+      "loss": 1.4239,
+      "step": 14290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.429601507709231e-06,
+      "loss": 1.4438,
+      "step": 14300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.428727215822702e-06,
+      "loss": 1.5347,
+      "step": 14310
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.427852340825651e-06,
+      "loss": 1.4668,
+      "step": 14320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.426976882982579e-06,
+      "loss": 1.4662,
+      "step": 14330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.426100842558163e-06,
+      "loss": 1.3497,
+      "step": 14340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.425224219817255e-06,
+      "loss": 1.4549,
+      "step": 14350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.424347015024884e-06,
+      "loss": 1.455,
+      "step": 14360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.423469228446254e-06,
+      "loss": 1.3986,
+      "step": 14370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.422590860346746e-06,
+      "loss": 1.5209,
+      "step": 14380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.421711910991917e-06,
+      "loss": 1.3922,
+      "step": 14390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.420832380647498e-06,
+      "loss": 1.3661,
+      "step": 14400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.419952269579396e-06,
+      "loss": 1.4679,
+      "step": 14410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.419071578053695e-06,
+      "loss": 1.4153,
+      "step": 14420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.418190306336655e-06,
+      "loss": 1.5154,
+      "step": 14430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.417308454694709e-06,
+      "loss": 1.3818,
+      "step": 14440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.4164260233944665e-06,
+      "loss": 1.4226,
+      "step": 14450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.415543012702712e-06,
+      "loss": 1.5838,
+      "step": 14460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.414659422886406e-06,
+      "loss": 1.3815,
+      "step": 14470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.413775254212683e-06,
+      "loss": 1.4483,
+      "step": 14480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.412890506948854e-06,
+      "loss": 1.4283,
+      "step": 14490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.412005181362402e-06,
+      "loss": 1.3752,
+      "step": 14500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.411119277720989e-06,
+      "loss": 1.5239,
+      "step": 14510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.410232796292447e-06,
+      "loss": 1.4799,
+      "step": 14520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.409345737344789e-06,
+      "loss": 1.4997,
+      "step": 14530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.408458101146194e-06,
+      "loss": 1.3766,
+      "step": 14540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.407569887965024e-06,
+      "loss": 1.4738,
+      "step": 14550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.40668109806981e-06,
+      "loss": 1.5358,
+      "step": 14560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.40579173172926e-06,
+      "loss": 1.4763,
+      "step": 14570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.404901789212255e-06,
+      "loss": 1.5282,
+      "step": 14580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.40401127078785e-06,
+      "loss": 1.5202,
+      "step": 14590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.403120176725276e-06,
+      "loss": 1.3944,
+      "step": 14600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.402228507293935e-06,
+      "loss": 1.552,
+      "step": 14610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.401336262763407e-06,
+      "loss": 1.4729,
+      "step": 14620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.400443443403441e-06,
+      "loss": 1.4439,
+      "step": 14630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3995500494839656e-06,
+      "loss": 1.4535,
+      "step": 14640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.398656081275077e-06,
+      "loss": 1.4555,
+      "step": 14650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3977615390470484e-06,
+      "loss": 1.4632,
+      "step": 14660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.396866423070326e-06,
+      "loss": 1.3692,
+      "step": 14670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.395970733615531e-06,
+      "loss": 1.499,
+      "step": 14680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3950744709534554e-06,
+      "loss": 1.448,
+      "step": 14690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.394177635355065e-06,
+      "loss": 1.417,
+      "step": 14700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3932802270915e-06,
+      "loss": 1.4876,
+      "step": 14710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.392382246434073e-06,
+      "loss": 1.4716,
+      "step": 14720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.39148369365427e-06,
+      "loss": 1.5054,
+      "step": 14730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3905845690237485e-06,
+      "loss": 1.3932,
+      "step": 14740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3896848728143416e-06,
+      "loss": 1.4428,
+      "step": 14750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.388784605298053e-06,
+      "loss": 1.5697,
+      "step": 14760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.387883766747061e-06,
+      "loss": 1.3691,
+      "step": 14770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.386982357433713e-06,
+      "loss": 1.5161,
+      "step": 14780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3860803776305335e-06,
+      "loss": 1.3671,
+      "step": 14790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.385177827610218e-06,
+      "loss": 1.3859,
+      "step": 14800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.38427470764563e-06,
+      "loss": 1.4594,
+      "step": 14810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.383371018009812e-06,
+      "loss": 1.3854,
+      "step": 14820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.382466758975976e-06,
+      "loss": 1.5057,
+      "step": 14830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.381561930817504e-06,
+      "loss": 1.4597,
+      "step": 14840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.380656533807953e-06,
+      "loss": 1.3907,
+      "step": 14850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.379750568221052e-06,
+      "loss": 1.4355,
+      "step": 14860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.378844034330698e-06,
+      "loss": 1.5124,
+      "step": 14870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.377936932410966e-06,
+      "loss": 1.5122,
+      "step": 14880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.3770292627360964e-06,
+      "loss": 1.3221,
+      "step": 14890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.376121025580505e-06,
+      "loss": 1.4,
+      "step": 14900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.375212221218781e-06,
+      "loss": 1.4233,
+      "step": 14910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.374302849925679e-06,
+      "loss": 1.3461,
+      "step": 14920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.37339291197613e-06,
+      "loss": 1.4451,
+      "step": 14930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.372482407645235e-06,
+      "loss": 1.3922,
+      "step": 14940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.371571337208265e-06,
+      "loss": 1.5358,
+      "step": 14950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.370659700940665e-06,
+      "loss": 1.4707,
+      "step": 14960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.369747499118048e-06,
+      "loss": 1.4407,
+      "step": 14970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.368834732016199e-06,
+      "loss": 1.5582,
+      "step": 14980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.367921399911075e-06,
+      "loss": 1.4048,
+      "step": 14990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.367007503078802e-06,
+      "loss": 1.4246,
+      "step": 15000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.366093041795678e-06,
+      "loss": 1.5192,
+      "step": 15010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.365178016338172e-06,
+      "loss": 1.3598,
+      "step": 15020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.364262426982922e-06,
+      "loss": 1.5618,
+      "step": 15030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3633462740067375e-06,
+      "loss": 1.5388,
+      "step": 15040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.362429557686599e-06,
+      "loss": 1.3783,
+      "step": 15050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.361512278299656e-06,
+      "loss": 1.488,
+      "step": 15060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.360594436123229e-06,
+      "loss": 1.418,
+      "step": 15070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.359676031434808e-06,
+      "loss": 1.488,
+      "step": 15080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.358757064512055e-06,
+      "loss": 1.4875,
+      "step": 15090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.357837535632798e-06,
+      "loss": 1.3969,
+      "step": 15100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.356917445075041e-06,
+      "loss": 1.5398,
+      "step": 15110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.35599679311695e-06,
+      "loss": 1.412,
+      "step": 15120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.355075580036868e-06,
+      "loss": 1.4538,
+      "step": 15130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3541538061133035e-06,
+      "loss": 1.4917,
+      "step": 15140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.353231471624936e-06,
+      "loss": 1.4273,
+      "step": 15150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.352308576850615e-06,
+      "loss": 1.4634,
+      "step": 15160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.351385122069356e-06,
+      "loss": 1.4701,
+      "step": 15170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.350461107560348e-06,
+      "loss": 1.4993,
+      "step": 15180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.349536533602947e-06,
+      "loss": 1.3728,
+      "step": 15190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.34861140047668e-06,
+      "loss": 1.4315,
+      "step": 15200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.347685708461241e-06,
+      "loss": 1.5747,
+      "step": 15210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.346759457836493e-06,
+      "loss": 1.4576,
+      "step": 15220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.345832648882468e-06,
+      "loss": 1.4758,
+      "step": 15230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.34490528187937e-06,
+      "loss": 1.4628,
+      "step": 15240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3439773571075674e-06,
+      "loss": 1.3801,
+      "step": 15250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3430488748475986e-06,
+      "loss": 1.4293,
+      "step": 15260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.342119835380172e-06,
+      "loss": 1.4258,
+      "step": 15270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.341190238986162e-06,
+      "loss": 1.5175,
+      "step": 15280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.340260085946614e-06,
+      "loss": 1.4562,
+      "step": 15290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3393293765427405e-06,
+      "loss": 1.4097,
+      "step": 15300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3383981110559206e-06,
+      "loss": 1.5418,
+      "step": 15310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.337466289767704e-06,
+      "loss": 1.4001,
+      "step": 15320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.336533912959807e-06,
+      "loss": 1.5128,
+      "step": 15330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.335600980914116e-06,
+      "loss": 1.4042,
+      "step": 15340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.334667493912681e-06,
+      "loss": 1.4309,
+      "step": 15350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3337334522377235e-06,
+      "loss": 1.4142,
+      "step": 15360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.332798856171633e-06,
+      "loss": 1.5113,
+      "step": 15370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.331863705996963e-06,
+      "loss": 1.5452,
+      "step": 15380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.330928001996438e-06,
+      "loss": 1.4285,
+      "step": 15390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.329991744452948e-06,
+      "loss": 1.4299,
+      "step": 15400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.329054933649551e-06,
+      "loss": 1.4986,
+      "step": 15410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.328117569869472e-06,
+      "loss": 1.3827,
+      "step": 15420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.327179653396105e-06,
+      "loss": 1.4912,
+      "step": 15430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.326241184513008e-06,
+      "loss": 1.424,
+      "step": 15440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.325302163503907e-06,
+      "loss": 1.4378,
+      "step": 15450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.324362590652698e-06,
+      "loss": 1.4885,
+      "step": 15460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.323422466243438e-06,
+      "loss": 1.4938,
+      "step": 15470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.322481790560356e-06,
+      "loss": 1.5134,
+      "step": 15480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.321540563887846e-06,
+      "loss": 1.4821,
+      "step": 15490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.320598786510468e-06,
+      "loss": 1.2971,
+      "step": 15500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.319656458712947e-06,
+      "loss": 1.5016,
+      "step": 15510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.318713580780179e-06,
+      "loss": 1.4259,
+      "step": 15520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.317770152997222e-06,
+      "loss": 1.5084,
+      "step": 15530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.316826175649301e-06,
+      "loss": 1.4076,
+      "step": 15540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.315881649021809e-06,
+      "loss": 1.4006,
+      "step": 15550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.314936573400305e-06,
+      "loss": 1.4504,
+      "step": 15560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.31399094907051e-06,
+      "loss": 1.4519,
+      "step": 15570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3130447763183165e-06,
+      "loss": 1.4521,
+      "step": 15580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.3120980554297794e-06,
+      "loss": 1.4199,
+      "step": 15590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.311150786691119e-06,
+      "loss": 1.4205,
+      "step": 15600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.310202970388724e-06,
+      "loss": 1.4979,
+      "step": 15610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3092546068091454e-06,
+      "loss": 1.4194,
+      "step": 15620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3083056962391015e-06,
+      "loss": 1.4558,
+      "step": 15630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3073562389654764e-06,
+      "loss": 1.392,
+      "step": 15640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3064062352753186e-06,
+      "loss": 1.3816,
+      "step": 15650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.3054556854558416e-06,
+      "loss": 1.4242,
+      "step": 15660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.304504589794425e-06,
+      "loss": 1.4038,
+      "step": 15670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.303552948578611e-06,
+      "loss": 1.4916,
+      "step": 15680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.30260076209611e-06,
+      "loss": 1.3499,
+      "step": 15690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.301648030634796e-06,
+      "loss": 1.4589,
+      "step": 15700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.300694754482706e-06,
+      "loss": 1.4917,
+      "step": 15710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.299740933928045e-06,
+      "loss": 1.5231,
+      "step": 15720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2987865692591795e-06,
+      "loss": 1.4697,
+      "step": 15730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.297831660764643e-06,
+      "loss": 1.3726,
+      "step": 15740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.29687620873313e-06,
+      "loss": 1.5016,
+      "step": 15750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.295920213453504e-06,
+      "loss": 1.4923,
+      "step": 15760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.29496367521479e-06,
+      "loss": 1.4349,
+      "step": 15770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.294006594306175e-06,
+      "loss": 1.5391,
+      "step": 15780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.293048971017015e-06,
+      "loss": 1.4341,
+      "step": 15790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.292090805636827e-06,
+      "loss": 1.3617,
+      "step": 15800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.291132098455292e-06,
+      "loss": 1.4813,
+      "step": 15810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.290172849762254e-06,
+      "loss": 1.4778,
+      "step": 15820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2892130598477245e-06,
+      "loss": 1.5558,
+      "step": 15830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.288252729001874e-06,
+      "loss": 1.4679,
+      "step": 15840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.287291857515038e-06,
+      "loss": 1.4731,
+      "step": 15850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.286330445677719e-06,
+      "loss": 1.4641,
+      "step": 15860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.285368493780576e-06,
+      "loss": 1.4509,
+      "step": 15870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.284406002114438e-06,
+      "loss": 1.5213,
+      "step": 15880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.283442970970293e-06,
+      "loss": 1.4451,
+      "step": 15890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.282479400639295e-06,
+      "loss": 1.4627,
+      "step": 15900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.281515291412757e-06,
+      "loss": 1.4633,
+      "step": 15910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.280550643582158e-06,
+      "loss": 1.3828,
+      "step": 15920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2795854574391406e-06,
+      "loss": 1.5376,
+      "step": 15930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.278619733275507e-06,
+      "loss": 1.4286,
+      "step": 15940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2776534713832265e-06,
+      "loss": 1.4295,
+      "step": 15950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.276686672054424e-06,
+      "loss": 1.5106,
+      "step": 15960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.275719335581395e-06,
+      "loss": 1.426,
+      "step": 15970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.274751462256593e-06,
+      "loss": 1.5052,
+      "step": 15980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.273783052372632e-06,
+      "loss": 1.4228,
+      "step": 15990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.272814106222293e-06,
+      "loss": 1.361,
+      "step": 16000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2718446240985154e-06,
+      "loss": 1.503,
+      "step": 16010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.270874606294402e-06,
+      "loss": 1.4144,
+      "step": 16020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2699040531032175e-06,
+      "loss": 1.4975,
+      "step": 16030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.268932964818389e-06,
+      "loss": 1.363,
+      "step": 16040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.267961341733505e-06,
+      "loss": 1.4349,
+      "step": 16050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.266989184142315e-06,
+      "loss": 1.3758,
+      "step": 16060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.266016492338731e-06,
+      "loss": 1.3788,
+      "step": 16070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.265043266616825e-06,
+      "loss": 1.4492,
+      "step": 16080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.264069507270833e-06,
+      "loss": 1.4405,
+      "step": 16090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.26309521459515e-06,
+      "loss": 1.4417,
+      "step": 16100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.262120388884333e-06,
+      "loss": 1.4546,
+      "step": 16110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.261145030433101e-06,
+      "loss": 1.3892,
+      "step": 16120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.260169139536332e-06,
+      "loss": 1.5393,
+      "step": 16130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2591927164890676e-06,
+      "loss": 1.4386,
+      "step": 16140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.258215761586509e-06,
+      "loss": 1.434,
+      "step": 16150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.257238275124017e-06,
+      "loss": 1.4207,
+      "step": 16160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.2562602573971165e-06,
+      "loss": 1.3979,
+      "step": 16170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.255281708701488e-06,
+      "loss": 1.4963,
+      "step": 16180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.254302629332977e-06,
+      "loss": 1.4498,
+      "step": 16190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.253323019587588e-06,
+      "loss": 1.4172,
+      "step": 16200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2523428797614845e-06,
+      "loss": 1.4803,
+      "step": 16210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.251362210150991e-06,
+      "loss": 1.5141,
+      "step": 16220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.250381011052595e-06,
+      "loss": 1.542,
+      "step": 16230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.249399282762938e-06,
+      "loss": 1.3981,
+      "step": 16240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.248417025578829e-06,
+      "loss": 1.4433,
+      "step": 16250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.247434239797231e-06,
+      "loss": 1.505,
+      "step": 16260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.246450925715269e-06,
+      "loss": 1.3103,
+      "step": 16270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.245467083630227e-06,
+      "loss": 1.5006,
+      "step": 16280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2444827138395495e-06,
+      "loss": 1.4143,
+      "step": 16290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.243497816640841e-06,
+      "loss": 1.3795,
+      "step": 16300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.242512392331864e-06,
+      "loss": 1.4962,
+      "step": 16310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.241526441210541e-06,
+      "loss": 1.488,
+      "step": 16320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.240539963574954e-06,
+      "loss": 1.5039,
+      "step": 16330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2395529597233434e-06,
+      "loss": 1.4932,
+      "step": 16340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.238565429954111e-06,
+      "loss": 1.4787,
+      "step": 16350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.237577374565813e-06,
+      "loss": 1.4272,
+      "step": 16360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.236588793857171e-06,
+      "loss": 1.459,
+      "step": 16370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.235599688127058e-06,
+      "loss": 1.5462,
+      "step": 16380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.234610057674512e-06,
+      "loss": 1.348,
+      "step": 16390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.233619902798727e-06,
+      "loss": 1.406,
+      "step": 16400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.232629223799054e-06,
+      "loss": 1.473,
+      "step": 16410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.231638020975007e-06,
+      "loss": 1.3754,
+      "step": 16420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2306462946262515e-06,
+      "loss": 1.5406,
+      "step": 16430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.22965404505262e-06,
+      "loss": 1.4127,
+      "step": 16440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.228661272554095e-06,
+      "loss": 1.4088,
+      "step": 16450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2276679774308225e-06,
+      "loss": 1.5249,
+      "step": 16460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2266741599831025e-06,
+      "loss": 1.4349,
+      "step": 16470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.225679820511398e-06,
+      "loss": 1.4465,
+      "step": 16480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.224684959316325e-06,
+      "loss": 1.4159,
+      "step": 16490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.223689576698659e-06,
+      "loss": 1.4123,
+      "step": 16500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.222693672959335e-06,
+      "loss": 1.4947,
+      "step": 16510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.22169724839944e-06,
+      "loss": 1.398,
+      "step": 16520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.220700303320227e-06,
+      "loss": 1.4457,
+      "step": 16530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.219702838023098e-06,
+      "loss": 1.4546,
+      "step": 16540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.218704852809616e-06,
+      "loss": 1.4295,
+      "step": 16550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2177063479815025e-06,
+      "loss": 1.5738,
+      "step": 16560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.216707323840635e-06,
+      "loss": 1.4745,
+      "step": 16570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.215707780689047e-06,
+      "loss": 1.55,
+      "step": 16580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.214707718828928e-06,
+      "loss": 1.43,
+      "step": 16590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.213707138562628e-06,
+      "loss": 1.444,
+      "step": 16600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.212706040192652e-06,
+      "loss": 1.4915,
+      "step": 16610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.211704424021659e-06,
+      "loss": 1.3715,
+      "step": 16620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.210702290352469e-06,
+      "loss": 1.5155,
+      "step": 16630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.209699639488054e-06,
+      "loss": 1.4701,
+      "step": 16640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2086964717315456e-06,
+      "loss": 1.4197,
+      "step": 16650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.207692787386232e-06,
+      "loss": 1.5376,
+      "step": 16660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.206688586755555e-06,
+      "loss": 1.4333,
+      "step": 16670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.205683870143114e-06,
+      "loss": 1.5255,
+      "step": 16680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.204678637852664e-06,
+      "loss": 1.4544,
+      "step": 16690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.203672890188116e-06,
+      "loss": 1.355,
+      "step": 16700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.202666627453537e-06,
+      "loss": 1.4977,
+      "step": 16710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.2016598499531495e-06,
+      "loss": 1.4642,
+      "step": 16720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.200652557991332e-06,
+      "loss": 1.4812,
+      "step": 16730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.199644751872618e-06,
+      "loss": 1.4789,
+      "step": 16740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.198636431901697e-06,
+      "loss": 1.3453,
+      "step": 16750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1976275983834135e-06,
+      "loss": 1.4636,
+      "step": 16760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1966182516227674e-06,
+      "loss": 1.3924,
+      "step": 16770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.195608391924912e-06,
+      "loss": 1.5222,
+      "step": 16780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.194598019595159e-06,
+      "loss": 1.3917,
+      "step": 16790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.193587134938975e-06,
+      "loss": 1.5153,
+      "step": 16800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.192575738261977e-06,
+      "loss": 1.467,
+      "step": 16810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.191563829869942e-06,
+      "loss": 1.4359,
+      "step": 16820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.190551410068796e-06,
+      "loss": 1.4992,
+      "step": 16830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.189538479164627e-06,
+      "loss": 1.4615,
+      "step": 16840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.188525037463672e-06,
+      "loss": 1.4366,
+      "step": 16850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.187511085272323e-06,
+      "loss": 1.4706,
+      "step": 16860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1864966228971285e-06,
+      "loss": 1.4873,
+      "step": 16870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.18548165064479e-06,
+      "loss": 1.5893,
+      "step": 16880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.184466168822164e-06,
+      "loss": 1.3938,
+      "step": 16890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.183450177736259e-06,
+      "loss": 1.4207,
+      "step": 16900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.182433677694238e-06,
+      "loss": 1.443,
+      "step": 16910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.181416669003421e-06,
+      "loss": 1.3734,
+      "step": 16920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.180399151971277e-06,
+      "loss": 1.4526,
+      "step": 16930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.179381126905434e-06,
+      "loss": 1.4346,
+      "step": 16940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.178362594113668e-06,
+      "loss": 1.3957,
+      "step": 16950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.177343553903912e-06,
+      "loss": 1.4296,
+      "step": 16960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.176324006584251e-06,
+      "loss": 1.3909,
+      "step": 16970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.175303952462926e-06,
+      "loss": 1.4716,
+      "step": 16980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.174283391848327e-06,
+      "loss": 1.4298,
+      "step": 16990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.173262325049e-06,
+      "loss": 1.413,
+      "step": 17000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.172240752373643e-06,
+      "loss": 1.4861,
+      "step": 17010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.171218674131109e-06,
+      "loss": 1.321,
+      "step": 17020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.170196090630399e-06,
+      "loss": 1.4874,
+      "step": 17030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.169173002180672e-06,
+      "loss": 1.4583,
+      "step": 17040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.168149409091236e-06,
+      "loss": 1.4596,
+      "step": 17050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.167125311671556e-06,
+      "loss": 1.4362,
+      "step": 17060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1661007102312434e-06,
+      "loss": 1.4189,
+      "step": 17070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.165075605080066e-06,
+      "loss": 1.4773,
+      "step": 17080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.164049996527944e-06,
+      "loss": 1.3459,
+      "step": 17090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1630238848849474e-06,
+      "loss": 1.4204,
+      "step": 17100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.161997270461301e-06,
+      "loss": 1.4868,
+      "step": 17110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.160970153567379e-06,
+      "loss": 1.4429,
+      "step": 17120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1599425345137095e-06,
+      "loss": 1.4869,
+      "step": 17130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.158914413610971e-06,
+      "loss": 1.4216,
+      "step": 17140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.157885791169996e-06,
+      "loss": 1.4434,
+      "step": 17150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1568566675017675e-06,
+      "loss": 1.537,
+      "step": 17160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.155827042917416e-06,
+      "loss": 1.3612,
+      "step": 17170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.154796917728231e-06,
+      "loss": 1.4955,
+      "step": 17180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.153766292245646e-06,
+      "loss": 1.4724,
+      "step": 17190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.152735166781253e-06,
+      "loss": 1.3954,
+      "step": 17200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.151703541646789e-06,
+      "loss": 1.5051,
+      "step": 17210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.150671417154144e-06,
+      "loss": 1.4092,
+      "step": 17220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.149638793615359e-06,
+      "loss": 1.4472,
+      "step": 17230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.148605671342629e-06,
+      "loss": 1.4086,
+      "step": 17240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.147572050648294e-06,
+      "loss": 1.4201,
+      "step": 17250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.14653793184485e-06,
+      "loss": 1.4792,
+      "step": 17260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.14550331524494e-06,
+      "loss": 1.4128,
+      "step": 17270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.144468201161358e-06,
+      "loss": 1.456,
+      "step": 17280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.143432589907051e-06,
+      "loss": 1.3647,
+      "step": 17290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.142396481795114e-06,
+      "loss": 1.4375,
+      "step": 17300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.141359877138792e-06,
+      "loss": 1.5086,
+      "step": 17310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.140322776251481e-06,
+      "loss": 1.4608,
+      "step": 17320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.139285179446728e-06,
+      "loss": 1.5129,
+      "step": 17330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.1382470870382264e-06,
+      "loss": 1.3582,
+      "step": 17340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.137208499339826e-06,
+      "loss": 1.4633,
+      "step": 17350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.136169416665517e-06,
+      "loss": 1.5195,
+      "step": 17360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.135129839329448e-06,
+      "loss": 1.487,
+      "step": 17370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.134089767645912e-06,
+      "loss": 1.5214,
+      "step": 17380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1330492019293545e-06,
+      "loss": 1.4336,
+      "step": 17390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.132008142494367e-06,
+      "loss": 1.2892,
+      "step": 17400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.130966589655693e-06,
+      "loss": 1.4819,
+      "step": 17410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.129924543728225e-06,
+      "loss": 1.3634,
+      "step": 17420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.128882005027003e-06,
+      "loss": 1.4098,
+      "step": 17430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.127838973867217e-06,
+      "loss": 1.3503,
+      "step": 17440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.126795450564206e-06,
+      "loss": 1.3994,
+      "step": 17450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.125751435433458e-06,
+      "loss": 1.5263,
+      "step": 17460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.124706928790608e-06,
+      "loss": 1.4774,
+      "step": 17470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.123661930951441e-06,
+      "loss": 1.4931,
+      "step": 17480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1226164422318925e-06,
+      "loss": 1.3747,
+      "step": 17490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.121570462948041e-06,
+      "loss": 1.4892,
+      "step": 17500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.120523993416119e-06,
+      "loss": 1.4467,
+      "step": 17510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.119477033952504e-06,
+      "loss": 1.4175,
+      "step": 17520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.118429584873722e-06,
+      "loss": 1.4176,
+      "step": 17530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.117381646496448e-06,
+      "loss": 1.4519,
+      "step": 17540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.116333219137504e-06,
+      "loss": 1.4502,
+      "step": 17550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.1152843031138604e-06,
+      "loss": 1.4735,
+      "step": 17560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.114234898742635e-06,
+      "loss": 1.4754,
+      "step": 17570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.113185006341094e-06,
+      "loss": 1.4508,
+      "step": 17580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.112134626226649e-06,
+      "loss": 1.4574,
+      "step": 17590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.111083758716862e-06,
+      "loss": 1.4521,
+      "step": 17600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.110032404129439e-06,
+      "loss": 1.4727,
+      "step": 17610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.108980562782239e-06,
+      "loss": 1.435,
+      "step": 17620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.10792823499326e-06,
+      "loss": 1.4467,
+      "step": 17630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.106875421080654e-06,
+      "loss": 1.4833,
+      "step": 17640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.105822121362717e-06,
+      "loss": 1.5013,
+      "step": 17650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.104768336157892e-06,
+      "loss": 1.5033,
+      "step": 17660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.103714065784767e-06,
+      "loss": 1.3619,
+      "step": 17670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.102659310562082e-06,
+      "loss": 1.4604,
+      "step": 17680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.101604070808719e-06,
+      "loss": 1.391,
+      "step": 17690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.100548346843708e-06,
+      "loss": 1.4973,
+      "step": 17700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0994921389862246e-06,
+      "loss": 1.5498,
+      "step": 17710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.098435447555591e-06,
+      "loss": 1.4096,
+      "step": 17720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.097378272871277e-06,
+      "loss": 1.5079,
+      "step": 17730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.096320615252896e-06,
+      "loss": 1.4332,
+      "step": 17740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.09526247502021e-06,
+      "loss": 1.4305,
+      "step": 17750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0942038524931245e-06,
+      "loss": 1.5021,
+      "step": 17760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.093144747991694e-06,
+      "loss": 1.4857,
+      "step": 17770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.092085161836114e-06,
+      "loss": 1.5118,
+      "step": 17780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.09102509434673e-06,
+      "loss": 1.4645,
+      "step": 17790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.089964545844032e-06,
+      "loss": 1.4886,
+      "step": 17800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.088903516648652e-06,
+      "loss": 1.5657,
+      "step": 17810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.087842007081373e-06,
+      "loss": 1.4218,
+      "step": 17820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.086780017463118e-06,
+      "loss": 1.4029,
+      "step": 17830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.08571754811496e-06,
+      "loss": 1.464,
+      "step": 17840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.084654599358112e-06,
+      "loss": 1.4148,
+      "step": 17850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.083591171513937e-06,
+      "loss": 1.4327,
+      "step": 17860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.082527264903937e-06,
+      "loss": 1.3322,
+      "step": 17870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.081462879849765e-06,
+      "loss": 1.4885,
+      "step": 17880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.080398016673214e-06,
+      "loss": 1.3978,
+      "step": 17890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.079332675696223e-06,
+      "loss": 1.4169,
+      "step": 17900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.078266857240877e-06,
+      "loss": 1.5029,
+      "step": 17910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0772005616294034e-06,
+      "loss": 1.4657,
+      "step": 17920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.076133789184175e-06,
+      "loss": 1.4232,
+      "step": 17930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0750665402277065e-06,
+      "loss": 1.4201,
+      "step": 17940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.07399881508266e-06,
+      "loss": 1.421,
+      "step": 17950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.07293061407184e-06,
+      "loss": 1.4542,
+      "step": 17960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.071861937518194e-06,
+      "loss": 1.2733,
+      "step": 17970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.070792785744815e-06,
+      "loss": 1.513,
+      "step": 17980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.069723159074938e-06,
+      "loss": 1.3775,
+      "step": 17990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.068653057831942e-06,
+      "loss": 1.3116,
+      "step": 18000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.067582482339351e-06,
+      "loss": 1.5333,
+      "step": 18010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0665114329208314e-06,
+      "loss": 1.3773,
+      "step": 18020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.065439909900191e-06,
+      "loss": 1.4512,
+      "step": 18030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.064367913601384e-06,
+      "loss": 1.4023,
+      "step": 18040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0632954443485055e-06,
+      "loss": 1.3246,
+      "step": 18050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.062222502465794e-06,
+      "loss": 1.5511,
+      "step": 18060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.061149088277631e-06,
+      "loss": 1.3866,
+      "step": 18070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.060075202108542e-06,
+      "loss": 1.4797,
+      "step": 18080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.059000844283194e-06,
+      "loss": 1.3863,
+      "step": 18090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.057926015126396e-06,
+      "loss": 1.4917,
+      "step": 18100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.056850714963101e-06,
+      "loss": 1.4839,
+      "step": 18110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0557749441184024e-06,
+      "loss": 1.3939,
+      "step": 18120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0546987029175375e-06,
+      "loss": 1.5535,
+      "step": 18130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.053621991685887e-06,
+      "loss": 1.4838,
+      "step": 18140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.05254481074897e-06,
+      "loss": 1.4803,
+      "step": 18150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0514671604324505e-06,
+      "loss": 1.4913,
+      "step": 18160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.050389041062135e-06,
+      "loss": 1.4093,
+      "step": 18170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.049310452963968e-06,
+      "loss": 1.4978,
+      "step": 18180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.04823139646404e-06,
+      "loss": 1.4046,
+      "step": 18190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.04715187188858e-06,
+      "loss": 1.422,
+      "step": 18200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.046071879563961e-06,
+      "loss": 1.5229,
+      "step": 18210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.044991419816695e-06,
+      "loss": 1.3454,
+      "step": 18220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.043910492973438e-06,
+      "loss": 1.5,
+      "step": 18230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0428290993609845e-06,
+      "loss": 1.4229,
+      "step": 18240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.041747239306272e-06,
+      "loss": 1.4332,
+      "step": 18250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.040664913136378e-06,
+      "loss": 1.4325,
+      "step": 18260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.03958212117852e-06,
+      "loss": 1.465,
+      "step": 18270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.038498863760061e-06,
+      "loss": 1.4609,
+      "step": 18280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0374151412084974e-06,
+      "loss": 1.401,
+      "step": 18290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.036330953851474e-06,
+      "loss": 1.4266,
+      "step": 18300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.035246302016769e-06,
+      "loss": 1.5267,
+      "step": 18310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.034161186032306e-06,
+      "loss": 1.3616,
+      "step": 18320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.033075606226146e-06,
+      "loss": 1.4889,
+      "step": 18330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.031989562926493e-06,
+      "loss": 1.4468,
+      "step": 18340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.030903056461689e-06,
+      "loss": 1.5374,
+      "step": 18350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.029816087160217e-06,
+      "loss": 1.5092,
+      "step": 18360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.028728655350697e-06,
+      "loss": 1.4554,
+      "step": 18370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.027640761361896e-06,
+      "loss": 1.5082,
+      "step": 18380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.026552405522712e-06,
+      "loss": 1.4067,
+      "step": 18390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.025463588162189e-06,
+      "loss": 1.3964,
+      "step": 18400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.024374309609507e-06,
+      "loss": 1.4397,
+      "step": 18410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.023284570193987e-06,
+      "loss": 1.3848,
+      "step": 18420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.02219437024509e-06,
+      "loss": 1.5401,
+      "step": 18430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0211037100924135e-06,
+      "loss": 1.3963,
+      "step": 18440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.020012590065698e-06,
+      "loss": 1.4653,
+      "step": 18450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.018921010494819e-06,
+      "loss": 1.5377,
+      "step": 18460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.017828971709793e-06,
+      "loss": 1.4158,
+      "step": 18470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.016736474040778e-06,
+      "loss": 1.5025,
+      "step": 18480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.015643517818064e-06,
+      "loss": 1.453,
+      "step": 18490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.014550103372086e-06,
+      "loss": 1.4176,
+      "step": 18500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.013456231033414e-06,
+      "loss": 1.3717,
+      "step": 18510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.012361901132758e-06,
+      "loss": 1.381,
+      "step": 18520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.011267114000966e-06,
+      "loss": 1.4638,
+      "step": 18530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.010171869969024e-06,
+      "loss": 1.4399,
+      "step": 18540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.009076169368056e-06,
+      "loss": 1.4335,
+      "step": 18550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.007980012529324e-06,
+      "loss": 1.4932,
+      "step": 18560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.006883399784229e-06,
+      "loss": 1.4555,
+      "step": 18570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.005786331464308e-06,
+      "loss": 1.5251,
+      "step": 18580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.004688807901239e-06,
+      "loss": 1.4566,
+      "step": 18590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.003590829426832e-06,
+      "loss": 1.456,
+      "step": 18600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.002492396373042e-06,
+      "loss": 1.5018,
+      "step": 18610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.001393509071953e-06,
+      "loss": 1.4797,
+      "step": 18620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.000294167855794e-06,
+      "loss": 1.5298,
+      "step": 18630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.999194373056926e-06,
+      "loss": 1.4629,
+      "step": 18640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99809412500785e-06,
+      "loss": 1.349,
+      "step": 18650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.996993424041204e-06,
+      "loss": 1.5538,
+      "step": 18660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99589227048976e-06,
+      "loss": 1.3976,
+      "step": 18670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.994790664686429e-06,
+      "loss": 1.4866,
+      "step": 18680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99368860696426e-06,
+      "loss": 1.4729,
+      "step": 18690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9925860976564365e-06,
+      "loss": 1.4264,
+      "step": 18700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.99148313709628e-06,
+      "loss": 1.4258,
+      "step": 18710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9903797256172464e-06,
+      "loss": 1.3701,
+      "step": 18720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.989275863552929e-06,
+      "loss": 1.4965,
+      "step": 18730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.98817155123706e-06,
+      "loss": 1.3953,
+      "step": 18740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9870667890035025e-06,
+      "loss": 1.437,
+      "step": 18750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9859615771862595e-06,
+      "loss": 1.4657,
+      "step": 18760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.984855916119468e-06,
+      "loss": 1.4914,
+      "step": 18770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.983749806137403e-06,
+      "loss": 1.4961,
+      "step": 18780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9826432475744735e-06,
+      "loss": 1.4152,
+      "step": 18790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.981536240765223e-06,
+      "loss": 1.4614,
+      "step": 18800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.980428786044334e-06,
+      "loss": 1.5091,
+      "step": 18810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.97932088374662e-06,
+      "loss": 1.4213,
+      "step": 18820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9782125342070346e-06,
+      "loss": 1.4304,
+      "step": 18830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9771037377606634e-06,
+      "loss": 1.4546,
+      "step": 18840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.975994494742727e-06,
+      "loss": 1.4035,
+      "step": 18850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.974884805488583e-06,
+      "loss": 1.4295,
+      "step": 18860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.973774670333722e-06,
+      "loss": 1.4076,
+      "step": 18870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.972664089613772e-06,
+      "loss": 1.4353,
+      "step": 18880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.971553063664491e-06,
+      "loss": 1.3928,
+      "step": 18890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.970441592821778e-06,
+      "loss": 1.4013,
+      "step": 18900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.96932967742166e-06,
+      "loss": 1.4951,
+      "step": 18910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.968217317800303e-06,
+      "loss": 1.4454,
+      "step": 18920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9671045142940066e-06,
+      "loss": 1.5549,
+      "step": 18930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.965991267239202e-06,
+      "loss": 1.3693,
+      "step": 18940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.964877576972456e-06,
+      "loss": 1.3863,
+      "step": 18950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9637634438304715e-06,
+      "loss": 1.5127,
+      "step": 18960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.962648868150084e-06,
+      "loss": 1.4586,
+      "step": 18970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9615338502682594e-06,
+      "loss": 1.4841,
+      "step": 18980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.960418390522102e-06,
+      "loss": 1.3558,
+      "step": 18990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.959302489248848e-06,
+      "loss": 1.4446,
+      "step": 19000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.958186146785865e-06,
+      "loss": 1.5308,
+      "step": 19010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.957069363470658e-06,
+      "loss": 1.4078,
+      "step": 19020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.955952139640863e-06,
+      "loss": 1.4694,
+      "step": 19030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9548344756342486e-06,
+      "loss": 1.4273,
+      "step": 19040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.953716371788717e-06,
+      "loss": 1.441,
+      "step": 19050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.952597828442305e-06,
+      "loss": 1.5436,
+      "step": 19060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.951478845933179e-06,
+      "loss": 1.398,
+      "step": 19070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.950359424599642e-06,
+      "loss": 1.5013,
+      "step": 19080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.949239564780126e-06,
+      "loss": 1.4028,
+      "step": 19090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.948119266813199e-06,
+      "loss": 1.4589,
+      "step": 19100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.946998531037559e-06,
+      "loss": 1.5012,
+      "step": 19110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9458773577920365e-06,
+      "loss": 1.4315,
+      "step": 19120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.944755747415595e-06,
+      "loss": 1.5392,
+      "step": 19130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.94363370024733e-06,
+      "loss": 1.3502,
+      "step": 19140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9425112166264715e-06,
+      "loss": 1.4322,
+      "step": 19150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9413882968923754e-06,
+      "loss": 1.4993,
+      "step": 19160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.940264941384535e-06,
+      "loss": 1.3762,
+      "step": 19170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.939141150442574e-06,
+      "loss": 1.4783,
+      "step": 19180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.938016924406246e-06,
+      "loss": 1.4307,
+      "step": 19190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.936892263615438e-06,
+      "loss": 1.4321,
+      "step": 19200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.935767168410168e-06,
+      "loss": 1.4771,
+      "step": 19210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.934641639130584e-06,
+      "loss": 1.399,
+      "step": 19220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.933515676116967e-06,
+      "loss": 1.4766,
+      "step": 19230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.932389279709729e-06,
+      "loss": 1.4484,
+      "step": 19240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.931262450249412e-06,
+      "loss": 1.4728,
+      "step": 19250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9301351880766895e-06,
+      "loss": 1.499,
+      "step": 19260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.929007493532365e-06,
+      "loss": 1.4482,
+      "step": 19270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.927879366957374e-06,
+      "loss": 1.4419,
+      "step": 19280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9267508086927845e-06,
+      "loss": 1.3746,
+      "step": 19290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.925621819079788e-06,
+      "loss": 1.4853,
+      "step": 19300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.924492398459714e-06,
+      "loss": 1.5132,
+      "step": 19310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.923362547174019e-06,
+      "loss": 1.3308,
+      "step": 19320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.92223226556429e-06,
+      "loss": 1.4828,
+      "step": 19330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.921101553972243e-06,
+      "loss": 1.3922,
+      "step": 19340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.919970412739725e-06,
+      "loss": 1.4572,
+      "step": 19350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.918838842208715e-06,
+      "loss": 1.5594,
+      "step": 19360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9177068427213195e-06,
+      "loss": 1.4694,
+      "step": 19370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.916574414619774e-06,
+      "loss": 1.5098,
+      "step": 19380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.915441558246443e-06,
+      "loss": 1.3629,
+      "step": 19390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.914308273943826e-06,
+      "loss": 1.4235,
+      "step": 19400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.913174562054544e-06,
+      "loss": 1.4606,
+      "step": 19410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.912040422921354e-06,
+      "loss": 1.4166,
+      "step": 19420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.910905856887138e-06,
+      "loss": 1.4357,
+      "step": 19430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9097708642949085e-06,
+      "loss": 1.3646,
+      "step": 19440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.908635445487806e-06,
+      "loss": 1.3922,
+      "step": 19450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.907499600809103e-06,
+      "loss": 1.4789,
+      "step": 19460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.906363330602197e-06,
+      "loss": 1.5169,
+      "step": 19470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.905226635210616e-06,
+      "loss": 1.4762,
+      "step": 19480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.904089514978017e-06,
+      "loss": 1.3453,
+      "step": 19490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.902951970248182e-06,
+      "loss": 1.4581,
+      "step": 19500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.901814001365027e-06,
+      "loss": 1.5657,
+      "step": 19510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.900675608672592e-06,
+      "loss": 1.3452,
+      "step": 19520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.899536792515045e-06,
+      "loss": 1.5333,
+      "step": 19530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.898397553236685e-06,
+      "loss": 1.4556,
+      "step": 19540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.897257891181937e-06,
+      "loss": 1.418,
+      "step": 19550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.896117806695355e-06,
+      "loss": 1.4822,
+      "step": 19560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.894977300121617e-06,
+      "loss": 1.3194,
+      "step": 19570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.893836371805534e-06,
+      "loss": 1.5289,
+      "step": 19580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.892695022092042e-06,
+      "loss": 1.3798,
+      "step": 19590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.891553251326203e-06,
+      "loss": 1.5026,
+      "step": 19600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8904110598532074e-06,
+      "loss": 1.5539,
+      "step": 19610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.889268448018374e-06,
+      "loss": 1.4227,
+      "step": 19620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.888125416167147e-06,
+      "loss": 1.57,
+      "step": 19630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.886981964645099e-06,
+      "loss": 1.4019,
+      "step": 19640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.885838093797929e-06,
+      "loss": 1.4518,
+      "step": 19650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.884693803971461e-06,
+      "loss": 1.5054,
+      "step": 19660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.883549095511649e-06,
+      "loss": 1.4756,
+      "step": 19670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.88240396876457e-06,
+      "loss": 1.5106,
+      "step": 19680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.8812584240764316e-06,
+      "loss": 1.3892,
+      "step": 19690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.880112461793563e-06,
+      "loss": 1.397,
+      "step": 19700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.878966082262423e-06,
+      "loss": 1.4857,
+      "step": 19710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.877819285829596e-06,
+      "loss": 1.3964,
+      "step": 19720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.876672072841791e-06,
+      "loss": 1.4969,
+      "step": 19730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.875524443645845e-06,
+      "loss": 1.4719,
+      "step": 19740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.874376398588719e-06,
+      "loss": 1.5006,
+      "step": 19750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.873227938017502e-06,
+      "loss": 1.5428,
+      "step": 19760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.872079062279405e-06,
+      "loss": 1.4066,
+      "step": 19770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.870929771721767e-06,
+      "loss": 1.4945,
+      "step": 19780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.869780066692055e-06,
+      "loss": 1.4276,
+      "step": 19790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.868629947537855e-06,
+      "loss": 1.3816,
+      "step": 19800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.867479414606882e-06,
+      "loss": 1.5141,
+      "step": 19810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.866328468246977e-06,
+      "loss": 1.3477,
+      "step": 19820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.865177108806105e-06,
+      "loss": 1.5406,
+      "step": 19830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8640253366323535e-06,
+      "loss": 1.4159,
+      "step": 19840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.862873152073938e-06,
+      "loss": 1.4056,
+      "step": 19850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.861720555479198e-06,
+      "loss": 1.4891,
+      "step": 19860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.860567547196596e-06,
+      "loss": 1.4059,
+      "step": 19870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.859414127574722e-06,
+      "loss": 1.4769,
+      "step": 19880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.858260296962286e-06,
+      "loss": 1.4672,
+      "step": 19890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.857106055708127e-06,
+      "loss": 1.5162,
+      "step": 19900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8559514041612026e-06,
+      "loss": 1.5464,
+      "step": 19910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.854796342670601e-06,
+      "loss": 1.3317,
+      "step": 19920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.853640871585529e-06,
+      "loss": 1.4705,
+      "step": 19930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.85248499125532e-06,
+      "loss": 1.4476,
+      "step": 19940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.851328702029429e-06,
+      "loss": 1.3487,
+      "step": 19950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.850172004257438e-06,
+      "loss": 1.505,
+      "step": 19960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.84901489828905e-06,
+      "loss": 1.4636,
+      "step": 19970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8478573844740916e-06,
+      "loss": 1.438,
+      "step": 19980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.846699463162513e-06,
+      "loss": 1.408,
+      "step": 19990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.845541134704387e-06,
+      "loss": 1.481,
+      "step": 20000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.844382399449911e-06,
+      "loss": 1.4408,
+      "step": 20010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.843223257749404e-06,
+      "loss": 1.4966,
+      "step": 20020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.842063709953309e-06,
+      "loss": 1.5775,
+      "step": 20030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8409037564121895e-06,
+      "loss": 1.375,
+      "step": 20040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.839743397476735e-06,
+      "loss": 1.4325,
+      "step": 20050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8385826334977555e-06,
+      "loss": 1.5686,
+      "step": 20060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.837421464826183e-06,
+      "loss": 1.3679,
+      "step": 20070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.836259891813075e-06,
+      "loss": 1.5594,
+      "step": 20080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.835097914809605e-06,
+      "loss": 1.4892,
+      "step": 20090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.833935534167077e-06,
+      "loss": 1.4152,
+      "step": 20100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.832772750236908e-06,
+      "loss": 1.5511,
+      "step": 20110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8316095633706455e-06,
+      "loss": 1.4859,
+      "step": 20120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.830445973919952e-06,
+      "loss": 1.4598,
+      "step": 20130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.829281982236617e-06,
+      "loss": 1.4816,
+      "step": 20140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.828117588672547e-06,
+      "loss": 1.4597,
+      "step": 20150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.826952793579774e-06,
+      "loss": 1.4917,
+      "step": 20160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.825787597310447e-06,
+      "loss": 1.4884,
+      "step": 20170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.824622000216841e-06,
+      "loss": 1.5154,
+      "step": 20180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8234560026513516e-06,
+      "loss": 1.4057,
+      "step": 20190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.822289604966489e-06,
+      "loss": 1.5054,
+      "step": 20200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8211228075148935e-06,
+      "loss": 1.5663,
+      "step": 20210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.81995561064932e-06,
+      "loss": 1.3562,
+      "step": 20220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.818788014722647e-06,
+      "loss": 1.5526,
+      "step": 20230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.817620020087872e-06,
+      "loss": 1.4606,
+      "step": 20240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.816451627098114e-06,
+      "loss": 1.3198,
+      "step": 20250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.815282836106614e-06,
+      "loss": 1.4511,
+      "step": 20260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.814113647466729e-06,
+      "loss": 1.4217,
+      "step": 20270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8129440615319414e-06,
+      "loss": 1.5214,
+      "step": 20280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.811774078655849e-06,
+      "loss": 1.4312,
+      "step": 20290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.810603699192173e-06,
+      "loss": 1.4696,
+      "step": 20300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8094329234947524e-06,
+      "loss": 1.4754,
+      "step": 20310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.808261751917548e-06,
+      "loss": 1.4117,
+      "step": 20320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.807090184814639e-06,
+      "loss": 1.3637,
+      "step": 20330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.805918222540223e-06,
+      "loss": 1.3694,
+      "step": 20340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.804745865448619e-06,
+      "loss": 1.4992,
+      "step": 20350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.803573113894266e-06,
+      "loss": 1.5238,
+      "step": 20360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.802399968231719e-06,
+      "loss": 1.5092,
+      "step": 20370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8012264288156566e-06,
+      "loss": 1.4622,
+      "step": 20380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8000524960008715e-06,
+      "loss": 1.4379,
+      "step": 20390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7988781701422795e-06,
+      "loss": 1.4653,
+      "step": 20400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7977034515949127e-06,
+      "loss": 1.4859,
+      "step": 20410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.796528340713923e-06,
+      "loss": 1.3513,
+      "step": 20420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7953528378545813e-06,
+      "loss": 1.4884,
+      "step": 20430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.794176943372275e-06,
+      "loss": 1.4336,
+      "step": 20440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.793000657622514e-06,
+      "loss": 1.3443,
+      "step": 20450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7918239809609216e-06,
+      "loss": 1.4801,
+      "step": 20460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7906469137432405e-06,
+      "loss": 1.419,
+      "step": 20470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7894694563253352e-06,
+      "loss": 1.505,
+      "step": 20480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7882916090631837e-06,
+      "loss": 1.3647,
+      "step": 20490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7871133723128846e-06,
+      "loss": 1.4428,
+      "step": 20500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.785934746430653e-06,
+      "loss": 1.5038,
+      "step": 20510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.78475573177282e-06,
+      "loss": 1.4527,
+      "step": 20520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7835763286958394e-06,
+      "loss": 1.5522,
+      "step": 20530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.782396537556277e-06,
+      "loss": 1.3931,
+      "step": 20540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.781216358710819e-06,
+      "loss": 1.4424,
+      "step": 20550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7800357925162675e-06,
+      "loss": 1.5177,
+      "step": 20560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.778854839329541e-06,
+      "loss": 1.4084,
+      "step": 20570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.777673499507678e-06,
+      "loss": 1.5393,
+      "step": 20580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7764917734078314e-06,
+      "loss": 1.4011,
+      "step": 20590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.775309661387272e-06,
+      "loss": 1.3854,
+      "step": 20600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7741271638033853e-06,
+      "loss": 1.523,
+      "step": 20610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.772944281013676e-06,
+      "loss": 1.4824,
+      "step": 20620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7717610133757633e-06,
+      "loss": 1.4418,
+      "step": 20630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.770577361247384e-06,
+      "loss": 1.4463,
+      "step": 20640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7693933249863912e-06,
+      "loss": 1.448,
+      "step": 20650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7682089049507524e-06,
+      "loss": 1.3978,
+      "step": 20660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.767024101498553e-06,
+      "loss": 1.4439,
+      "step": 20670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7658389149879936e-06,
+      "loss": 1.4631,
+      "step": 20680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7646533457773914e-06,
+      "loss": 1.4638,
+      "step": 20690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7634673942251775e-06,
+      "loss": 1.4515,
+      "step": 20700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7622810606899e-06,
+      "loss": 1.4213,
+      "step": 20710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7610943455302223e-06,
+      "loss": 1.4735,
+      "step": 20720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7599072491049225e-06,
+      "loss": 1.4968,
+      "step": 20730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.758719771772896e-06,
+      "loss": 1.4414,
+      "step": 20740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7575319138931497e-06,
+      "loss": 1.4072,
+      "step": 20750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7563436758248094e-06,
+      "loss": 1.52,
+      "step": 20760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7551550579271127e-06,
+      "loss": 1.5114,
+      "step": 20770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7539660605594154e-06,
+      "loss": 1.4588,
+      "step": 20780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7527766840811843e-06,
+      "loss": 1.3992,
+      "step": 20790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7515869288520027e-06,
+      "loss": 1.4052,
+      "step": 20800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.75039679523157e-06,
+      "loss": 1.5203,
+      "step": 20810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7492062835796967e-06,
+      "loss": 1.4692,
+      "step": 20820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.74801539425631e-06,
+      "loss": 1.5562,
+      "step": 20830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.74682412762145e-06,
+      "loss": 1.396,
+      "step": 20840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.745632484035272e-06,
+      "loss": 1.4125,
+      "step": 20850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.744440463858044e-06,
+      "loss": 1.514,
+      "step": 20860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.743248067450149e-06,
+      "loss": 1.4972,
+      "step": 20870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7420552951720822e-06,
+      "loss": 1.4457,
+      "step": 20880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7408621473844548e-06,
+      "loss": 1.3572,
+      "step": 20890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.7396686244479894e-06,
+      "loss": 1.3314,
+      "step": 20900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.738474726723523e-06,
+      "loss": 1.4785,
+      "step": 20910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7372804545720067e-06,
+      "loss": 1.3667,
+      "step": 20920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7360858083545022e-06,
+      "loss": 1.4633,
+      "step": 20930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7348907884321868e-06,
+      "loss": 1.4003,
+      "step": 20940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.733695395166349e-06,
+      "loss": 1.3888,
+      "step": 20950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7324996289183927e-06,
+      "loss": 1.4908,
+      "step": 20960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7313034900498317e-06,
+      "loss": 1.4519,
+      "step": 20970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7301069789222933e-06,
+      "loss": 1.4937,
+      "step": 20980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7289100958975194e-06,
+      "loss": 1.3747,
+      "step": 20990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7277128413373607e-06,
+      "loss": 1.5033,
+      "step": 21000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.726515215603784e-06,
+      "loss": 1.5795,
+      "step": 21010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.725317219058866e-06,
+      "loss": 1.3633,
+      "step": 21020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.724118852064795e-06,
+      "loss": 1.4581,
+      "step": 21030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7229201149838727e-06,
+      "loss": 1.4255,
+      "step": 21040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7217210081785137e-06,
+      "loss": 1.4274,
+      "step": 21050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.720521532011241e-06,
+      "loss": 1.4841,
+      "step": 21060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7193216868446924e-06,
+      "loss": 1.4272,
+      "step": 21070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.718121473041615e-06,
+      "loss": 1.457,
+      "step": 21080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.71692089096487e-06,
+      "loss": 1.3874,
+      "step": 21090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7157199409774288e-06,
+      "loss": 1.4023,
+      "step": 21100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7145186234423715e-06,
+      "loss": 1.4674,
+      "step": 21110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.713316938722892e-06,
+      "loss": 1.414,
+      "step": 21120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7121148871822954e-06,
+      "loss": 1.5469,
+      "step": 21130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.710912469183997e-06,
+      "loss": 1.4315,
+      "step": 21140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7097096850915225e-06,
+      "loss": 1.419,
+      "step": 21150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.708506535268509e-06,
+      "loss": 1.5408,
+      "step": 21160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.707303020078703e-06,
+      "loss": 1.4201,
+      "step": 21170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7060991398859625e-06,
+      "loss": 1.5469,
+      "step": 21180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7048948950542564e-06,
+      "loss": 1.3964,
+      "step": 21190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.703690285947662e-06,
+      "loss": 1.3997,
+      "step": 21200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.7024853129303685e-06,
+      "loss": 1.5538,
+      "step": 21210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.701279976366674e-06,
+      "loss": 1.3711,
+      "step": 21220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.700074276620987e-06,
+      "loss": 1.4149,
+      "step": 21230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.698868214057827e-06,
+      "loss": 1.4328,
+      "step": 21240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6976617890418197e-06,
+      "loss": 1.4419,
+      "step": 21250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.696455001937704e-06,
+      "loss": 1.5173,
+      "step": 21260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6952478531103265e-06,
+      "loss": 1.4171,
+      "step": 21270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.694040342924643e-06,
+      "loss": 1.6024,
+      "step": 21280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.69283247174572e-06,
+      "loss": 1.3839,
+      "step": 21290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6916242399387313e-06,
+      "loss": 1.3977,
+      "step": 21300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6904156478689613e-06,
+      "loss": 1.384,
+      "step": 21310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6892066959018034e-06,
+      "loss": 1.4517,
+      "step": 21320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6879973844027565e-06,
+      "loss": 1.5518,
+      "step": 21330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.686787713737433e-06,
+      "loss": 1.5355,
+      "step": 21340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.685577684271552e-06,
+      "loss": 1.409,
+      "step": 21350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.684367296370939e-06,
+      "loss": 1.5566,
+      "step": 21360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6831565504015303e-06,
+      "loss": 1.4315,
+      "step": 21370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6819454467293692e-06,
+      "loss": 1.4719,
+      "step": 21380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6807339857206087e-06,
+      "loss": 1.3695,
+      "step": 21390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6795221677415085e-06,
+      "loss": 1.3866,
+      "step": 21400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6783099931584364e-06,
+      "loss": 1.4672,
+      "step": 21410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6770974623378685e-06,
+      "loss": 1.3475,
+      "step": 21420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.675884575646387e-06,
+      "loss": 1.4314,
+      "step": 21430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.674671333450684e-06,
+      "loss": 1.3649,
+      "step": 21440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6734577361175585e-06,
+      "loss": 1.4168,
+      "step": 21450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6722437840139152e-06,
+      "loss": 1.4977,
+      "step": 21460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6710294775067677e-06,
+      "loss": 1.3701,
+      "step": 21470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.669814816963237e-06,
+      "loss": 1.5035,
+      "step": 21480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.6685998027505485e-06,
+      "loss": 1.4187,
+      "step": 21490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6673844352360395e-06,
+      "loss": 1.3987,
+      "step": 21500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6661687147871485e-06,
+      "loss": 1.4887,
+      "step": 21510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6649526417714244e-06,
+      "loss": 1.4923,
+      "step": 21520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6637362165565204e-06,
+      "loss": 1.4163,
+      "step": 21530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.662519439510198e-06,
+      "loss": 1.3959,
+      "step": 21540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6613023110003265e-06,
+      "loss": 1.4407,
+      "step": 21550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6600848313948762e-06,
+      "loss": 1.5259,
+      "step": 21560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.658867001061928e-06,
+      "loss": 1.4803,
+      "step": 21570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6576488203696676e-06,
+      "loss": 1.5599,
+      "step": 21580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6564302896863867e-06,
+      "loss": 1.3596,
+      "step": 21590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6552114093804825e-06,
+      "loss": 1.4139,
+      "step": 21600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6539921798204575e-06,
+      "loss": 1.5136,
+      "step": 21610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6527726013749214e-06,
+      "loss": 1.4646,
+      "step": 21620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.651552674412588e-06,
+      "loss": 1.5295,
+      "step": 21630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6503323993022767e-06,
+      "loss": 1.4125,
+      "step": 21640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.649111776412912e-06,
+      "loss": 1.4038,
+      "step": 21650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.647890806113524e-06,
+      "loss": 1.4882,
+      "step": 21660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6466694887732484e-06,
+      "loss": 1.3603,
+      "step": 21670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.645447824761324e-06,
+      "loss": 1.5396,
+      "step": 21680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.644225814447096e-06,
+      "loss": 1.4844,
+      "step": 21690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.643003458200014e-06,
+      "loss": 1.3496,
+      "step": 21700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6417807563896314e-06,
+      "loss": 1.46,
+      "step": 21710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6405577093856072e-06,
+      "loss": 1.4498,
+      "step": 21720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6393343175577033e-06,
+      "loss": 1.3928,
+      "step": 21730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6381105812757877e-06,
+      "loss": 1.4137,
+      "step": 21740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6368865009098307e-06,
+      "loss": 1.3798,
+      "step": 21750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6356620768299087e-06,
+      "loss": 1.5202,
+      "step": 21760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6344373094062e-06,
+      "loss": 1.3596,
+      "step": 21770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.633212199008988e-06,
+      "loss": 1.4319,
+      "step": 21780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.631986746008659e-06,
+      "loss": 1.5003,
+      "step": 21790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.630760950775703e-06,
+      "loss": 1.4096,
+      "step": 21800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.629534813680714e-06,
+      "loss": 1.4466,
+      "step": 21810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6283083350943894e-06,
+      "loss": 1.4673,
+      "step": 21820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.627081515387529e-06,
+      "loss": 1.4529,
+      "step": 21830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.625854354931035e-06,
+      "loss": 1.4381,
+      "step": 21840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6246268540959157e-06,
+      "loss": 1.3957,
+      "step": 21850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.62339901325328e-06,
+      "loss": 1.4986,
+      "step": 21860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.62217083277434e-06,
+      "loss": 1.4469,
+      "step": 21870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.62094231303041e-06,
+      "loss": 1.5017,
+      "step": 21880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.619713454392907e-06,
+      "loss": 1.2798,
+      "step": 21890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.618484257233352e-06,
+      "loss": 1.3001,
+      "step": 21900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.617254721923367e-06,
+      "loss": 1.5749,
+      "step": 21910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6160248488346757e-06,
+      "loss": 1.4021,
+      "step": 21920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6147946383391053e-06,
+      "loss": 1.4811,
+      "step": 21930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6135640908085827e-06,
+      "loss": 1.4129,
+      "step": 21940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.612333206615141e-06,
+      "loss": 1.4419,
+      "step": 21950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6111019861309105e-06,
+      "loss": 1.4908,
+      "step": 21960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.609870429728126e-06,
+      "loss": 1.5212,
+      "step": 21970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.608638537779122e-06,
+      "loss": 1.4759,
+      "step": 21980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6074063106563355e-06,
+      "loss": 1.4636,
+      "step": 21990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.606173748732307e-06,
+      "loss": 1.4255,
+      "step": 22000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6049408523796723e-06,
+      "loss": 1.5154,
+      "step": 22010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.603707621971175e-06,
+      "loss": 1.4045,
+      "step": 22020
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6024740578796548e-06,
+      "loss": 1.5077,
+      "step": 22030
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6012401604780543e-06,
+      "loss": 1.4555,
+      "step": 22040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.6000059301394173e-06,
+      "loss": 1.3618,
+      "step": 22050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5987713672368878e-06,
+      "loss": 1.5477,
+      "step": 22060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.5975364721437093e-06,
+      "loss": 1.2628,
+      "step": 22070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.596301245233228e-06,
+      "loss": 1.4643,
+      "step": 22080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5950656868788877e-06,
+      "loss": 1.4268,
+      "step": 22090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.593829797454234e-06,
+      "loss": 1.4157,
+      "step": 22100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5925935773329128e-06,
+      "loss": 1.4602,
+      "step": 22110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5913570268886695e-06,
+      "loss": 1.4086,
+      "step": 22120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.590120146495349e-06,
+      "loss": 1.4456,
+      "step": 22130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.588882936526897e-06,
+      "loss": 1.3958,
+      "step": 22140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5876453973573564e-06,
+      "loss": 1.4111,
+      "step": 22150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.586407529360874e-06,
+      "loss": 1.4294,
+      "step": 22160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.585169332911692e-06,
+      "loss": 1.493,
+      "step": 22170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5839308083841532e-06,
+      "loss": 1.4838,
+      "step": 22180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5826919561526998e-06,
+      "loss": 1.4299,
+      "step": 22190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5814527765918733e-06,
+      "loss": 1.5592,
+      "step": 22200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5802132700763135e-06,
+      "loss": 1.5191,
+      "step": 22210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5789734369807593e-06,
+      "loss": 1.3764,
+      "step": 22220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.57773327768005e-06,
+      "loss": 1.4613,
+      "step": 22230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.576492792549119e-06,
+      "loss": 1.3522,
+      "step": 22240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5752519819630037e-06,
+      "loss": 1.4249,
+      "step": 22250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.574010846296836e-06,
+      "loss": 1.4862,
+      "step": 22260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.572769385925849e-06,
+      "loss": 1.4258,
+      "step": 22270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.57152760122537e-06,
+      "loss": 1.4877,
+      "step": 22280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5702854925708285e-06,
+      "loss": 1.4347,
+      "step": 22290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.569043060337749e-06,
+      "loss": 1.4429,
+      "step": 22300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5678003049017566e-06,
+      "loss": 1.4187,
+      "step": 22310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.566557226638572e-06,
+      "loss": 1.4303,
+      "step": 22320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5653138259240135e-06,
+      "loss": 1.4596,
+      "step": 22330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5640701031339976e-06,
+      "loss": 1.3822,
+      "step": 22340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.562826058644537e-06,
+      "loss": 1.433,
+      "step": 22350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.561581692831745e-06,
+      "loss": 1.5684,
+      "step": 22360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5603370060718274e-06,
+      "loss": 1.4938,
+      "step": 22370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.559091998741091e-06,
+      "loss": 1.4782,
+      "step": 22380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5578466712159367e-06,
+      "loss": 1.3008,
+      "step": 22390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.556601023872863e-06,
+      "loss": 1.527,
+      "step": 22400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.555355057088468e-06,
+      "loss": 1.5137,
+      "step": 22410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.554108771239441e-06,
+      "loss": 1.3398,
+      "step": 22420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.552862166702572e-06,
+      "loss": 1.484,
+      "step": 22430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.551615243854746e-06,
+      "loss": 1.4749,
+      "step": 22440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5503680030729427e-06,
+      "loss": 1.4277,
+      "step": 22450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.549120444734241e-06,
+      "loss": 1.4905,
+      "step": 22460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5478725692158133e-06,
+      "loss": 1.3572,
+      "step": 22470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5466243768949292e-06,
+      "loss": 1.5193,
+      "step": 22480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5453758681489542e-06,
+      "loss": 1.3855,
+      "step": 22490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5441270433553477e-06,
+      "loss": 1.3749,
+      "step": 22500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5428779028916666e-06,
+      "loss": 1.4763,
+      "step": 22510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.541628447135562e-06,
+      "loss": 1.4503,
+      "step": 22520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.540378676464782e-06,
+      "loss": 1.3997,
+      "step": 22530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.539128591257167e-06,
+      "loss": 1.3805,
+      "step": 22540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.537878191890656e-06,
+      "loss": 1.5105,
+      "step": 22550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5366274787432804e-06,
+      "loss": 1.4508,
+      "step": 22560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5353764521931665e-06,
+      "loss": 1.4512,
+      "step": 22570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5341251126185376e-06,
+      "loss": 1.4589,
+      "step": 22580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5328734603977095e-06,
+      "loss": 1.4323,
+      "step": 22590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5316214959090933e-06,
+      "loss": 1.4399,
+      "step": 22600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.530369219531194e-06,
+      "loss": 1.4699,
+      "step": 22610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.529116631642612e-06,
+      "loss": 1.3819,
+      "step": 22620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5278637326220405e-06,
+      "loss": 1.4614,
+      "step": 22630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.526610522848267e-06,
+      "loss": 1.4398,
+      "step": 22640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.525357002700175e-06,
+      "loss": 1.4608,
+      "step": 22650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5241031725567386e-06,
+      "loss": 1.4707,
+      "step": 22660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.5228490327970283e-06,
+      "loss": 1.4741,
+      "step": 22670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5215945838002075e-06,
+      "loss": 1.4647,
+      "step": 22680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5203398259455314e-06,
+      "loss": 1.3717,
+      "step": 22690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.51908475961235e-06,
+      "loss": 1.4549,
+      "step": 22700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.517829385180107e-06,
+      "loss": 1.4968,
+      "step": 22710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5165737030283404e-06,
+      "loss": 1.4578,
+      "step": 22720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.515317713536677e-06,
+      "loss": 1.5201,
+      "step": 22730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5140614170848393e-06,
+      "loss": 1.4276,
+      "step": 22740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5128048140526426e-06,
+      "loss": 1.3396,
+      "step": 22750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5115479048199964e-06,
+      "loss": 1.4536,
+      "step": 22760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5102906897669e-06,
+      "loss": 1.3323,
+      "step": 22770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5090331692734452e-06,
+      "loss": 1.459,
+      "step": 22780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.507775343719818e-06,
+      "loss": 1.4218,
+      "step": 22790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.506517213486296e-06,
+      "loss": 1.4194,
+      "step": 22800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5052587789532482e-06,
+      "loss": 1.465,
+      "step": 22810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.504000040501136e-06,
+      "loss": 1.3962,
+      "step": 22820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5027409985105133e-06,
+      "loss": 1.4893,
+      "step": 22830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.501481653362025e-06,
+      "loss": 1.46,
+      "step": 22840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.5002220054364067e-06,
+      "loss": 1.3719,
+      "step": 22850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.49896205511449e-06,
+      "loss": 1.5078,
+      "step": 22860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.497701802777191e-06,
+      "loss": 1.3506,
+      "step": 22870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.496441248805522e-06,
+      "loss": 1.4629,
+      "step": 22880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.495180393580587e-06,
+      "loss": 1.3917,
+      "step": 22890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.493919237483576e-06,
+      "loss": 1.4311,
+      "step": 22900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.492657780895776e-06,
+      "loss": 1.4521,
+      "step": 22910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4913960241985616e-06,
+      "loss": 1.4265,
+      "step": 22920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.490133967773398e-06,
+      "loss": 1.5141,
+      "step": 22930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.488871612001843e-06,
+      "loss": 1.4101,
+      "step": 22940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4876089572655415e-06,
+      "loss": 1.4363,
+      "step": 22950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.486346003946232e-06,
+      "loss": 1.5131,
+      "step": 22960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4850827524257425e-06,
+      "loss": 1.3839,
+      "step": 22970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4838192030859902e-06,
+      "loss": 1.4736,
+      "step": 22980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.482555356308983e-06,
+      "loss": 1.3826,
+      "step": 22990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4812912124768193e-06,
+      "loss": 1.4678,
+      "step": 23000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.480026771971686e-06,
+      "loss": 1.4359,
+      "step": 23010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.478762035175861e-06,
+      "loss": 1.3982,
+      "step": 23020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.47749700247171e-06,
+      "loss": 1.4807,
+      "step": 23030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4762316742416913e-06,
+      "loss": 1.4267,
+      "step": 23040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4749660508683473e-06,
+      "loss": 1.4283,
+      "step": 23050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.473700132734316e-06,
+      "loss": 1.5828,
+      "step": 23060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.47243392022232e-06,
+      "loss": 1.4068,
+      "step": 23070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.471167413715173e-06,
+      "loss": 1.4887,
+      "step": 23080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4699006135957763e-06,
+      "loss": 1.4129,
+      "step": 23090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.46863352024712e-06,
+      "loss": 1.4255,
+      "step": 23100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.467366134052284e-06,
+      "loss": 1.4554,
+      "step": 23110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4660984553944367e-06,
+      "loss": 1.4454,
+      "step": 23120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.464830484656834e-06,
+      "loss": 1.4495,
+      "step": 23130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4635622222228206e-06,
+      "loss": 1.426,
+      "step": 23140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4622936684758273e-06,
+      "loss": 1.4362,
+      "step": 23150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4610248237993775e-06,
+      "loss": 1.4438,
+      "step": 23160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.459755688577079e-06,
+      "loss": 1.4104,
+      "step": 23170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.458486263192629e-06,
+      "loss": 1.534,
+      "step": 23180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.457216548029811e-06,
+      "loss": 1.4014,
+      "step": 23190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.455946543472497e-06,
+      "loss": 1.4264,
+      "step": 23200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.454676249904646e-06,
+      "loss": 1.482,
+      "step": 23210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4534056677103068e-06,
+      "loss": 1.4554,
+      "step": 23220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4521347972736114e-06,
+      "loss": 1.4627,
+      "step": 23230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4508636389787816e-06,
+      "loss": 1.4242,
+      "step": 23240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4495921932101264e-06,
+      "loss": 1.5095,
+      "step": 23250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.4483204603520385e-06,
+      "loss": 1.5174,
+      "step": 23260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.447048440789004e-06,
+      "loss": 1.4126,
+      "step": 23270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4457761349055885e-06,
+      "loss": 1.4592,
+      "step": 23280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.444503543086448e-06,
+      "loss": 1.4897,
+      "step": 23290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.443230665716324e-06,
+      "loss": 1.4288,
+      "step": 23300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4419575031800444e-06,
+      "loss": 1.4616,
+      "step": 23310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4406840558625237e-06,
+      "loss": 1.4173,
+      "step": 23320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.439410324148762e-06,
+      "loss": 1.4345,
+      "step": 23330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4381363084238457e-06,
+      "loss": 1.4424,
+      "step": 23340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4368620090729467e-06,
+      "loss": 1.382,
+      "step": 23350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4355874264813228e-06,
+      "loss": 1.4744,
+      "step": 23360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.434312561034318e-06,
+      "loss": 1.3598,
+      "step": 23370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4330374131173616e-06,
+      "loss": 1.5084,
+      "step": 23380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4317619831159666e-06,
+      "loss": 1.4409,
+      "step": 23390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.430486271415734e-06,
+      "loss": 1.4922,
+      "step": 23400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4292102784023486e-06,
+      "loss": 1.4758,
+      "step": 23410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4279340044615795e-06,
+      "loss": 1.4922,
+      "step": 23420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.426657449979282e-06,
+      "loss": 1.5022,
+      "step": 23430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.425380615341396e-06,
+      "loss": 1.3886,
+      "step": 23440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4241035009339457e-06,
+      "loss": 1.4486,
+      "step": 23450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.42282610714304e-06,
+      "loss": 1.4665,
+      "step": 23460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.421548434354872e-06,
+      "loss": 1.4379,
+      "step": 23470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.420270482955721e-06,
+      "loss": 1.43,
+      "step": 23480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4189922533319475e-06,
+      "loss": 1.4234,
+      "step": 23490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.417713745869997e-06,
+      "loss": 1.41,
+      "step": 23500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4164349609564008e-06,
+      "loss": 1.5137,
+      "step": 23510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.415155898977773e-06,
+      "loss": 1.4036,
+      "step": 23520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4138765603208113e-06,
+      "loss": 1.502,
+      "step": 23530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.412596945372297e-06,
+      "loss": 1.3837,
+      "step": 23540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4113170545190945e-06,
+      "loss": 1.5433,
+      "step": 23550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4100368881481515e-06,
+      "loss": 1.5732,
+      "step": 23560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4087564466465018e-06,
+      "loss": 1.4468,
+      "step": 23570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4074757304012597e-06,
+      "loss": 1.5368,
+      "step": 23580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4061947397996213e-06,
+      "loss": 1.402,
+      "step": 23590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4049134752288692e-06,
+      "loss": 1.3699,
+      "step": 23600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.403631937076366e-06,
+      "loss": 1.5089,
+      "step": 23610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.402350125729559e-06,
+      "loss": 1.4526,
+      "step": 23620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.4010680415759773e-06,
+      "loss": 1.5132,
+      "step": 23630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.399785685003231e-06,
+      "loss": 1.3311,
+      "step": 23640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.398503056399015e-06,
+      "loss": 1.418,
+      "step": 23650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.397220156151105e-06,
+      "loss": 1.4681,
+      "step": 23660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3959369846473596e-06,
+      "loss": 1.471,
+      "step": 23670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3946535422757187e-06,
+      "loss": 1.498,
+      "step": 23680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.393369829424204e-06,
+      "loss": 1.5507,
+      "step": 23690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3920858464809197e-06,
+      "loss": 1.4365,
+      "step": 23700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.390801593834051e-06,
+      "loss": 1.4981,
+      "step": 23710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3895170718718664e-06,
+      "loss": 1.5063,
+      "step": 23720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.388232280982713e-06,
+      "loss": 1.4763,
+      "step": 23730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.38694722155502e-06,
+      "loss": 1.4136,
+      "step": 23740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3856618939772996e-06,
+      "loss": 1.4672,
+      "step": 23750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3843762986381436e-06,
+      "loss": 1.4531,
+      "step": 23760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3830904359262246e-06,
+      "loss": 1.4713,
+      "step": 23770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3818043062302965e-06,
+      "loss": 1.5074,
+      "step": 23780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3805179099391937e-06,
+      "loss": 1.4515,
+      "step": 23790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3792312474418324e-06,
+      "loss": 1.3708,
+      "step": 23800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3779443191272072e-06,
+      "loss": 1.4412,
+      "step": 23810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3766571253843946e-06,
+      "loss": 1.3584,
+      "step": 23820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3753696666025498e-06,
+      "loss": 1.4995,
+      "step": 23830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3740819431709103e-06,
+      "loss": 1.4484,
+      "step": 23840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.3727939554787925e-06,
+      "loss": 1.4188,
+      "step": 23850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3715057039155912e-06,
+      "loss": 1.5133,
+      "step": 23860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.370217188870784e-06,
+      "loss": 1.3344,
+      "step": 23870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3689284107339265e-06,
+      "loss": 1.4492,
+      "step": 23880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.367639369894653e-06,
+      "loss": 1.3683,
+      "step": 23890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3663500667426798e-06,
+      "loss": 1.419,
+      "step": 23900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3650605016677978e-06,
+      "loss": 1.4508,
+      "step": 23910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.363770675059883e-06,
+      "loss": 1.3925,
+      "step": 23920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.362480587308887e-06,
+      "loss": 1.4468,
+      "step": 23930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.361190238804841e-06,
+      "loss": 1.3642,
+      "step": 23940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3598996299378544e-06,
+      "loss": 1.4291,
+      "step": 23950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3586087610981155e-06,
+      "loss": 1.4154,
+      "step": 23960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3573176326758934e-06,
+      "loss": 1.4009,
+      "step": 23970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3560262450615322e-06,
+      "loss": 1.4607,
+      "step": 23980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.354734598645457e-06,
+      "loss": 1.383,
+      "step": 23990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3534426938181704e-06,
+      "loss": 1.4524,
+      "step": 24000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.352150530970251e-06,
+      "loss": 1.455,
+      "step": 24010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3508581104923598e-06,
+      "loss": 1.4183,
+      "step": 24020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3495654327752326e-06,
+      "loss": 1.4736,
+      "step": 24030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.348272498209683e-06,
+      "loss": 1.4237,
+      "step": 24040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.346979307186603e-06,
+      "loss": 1.4929,
+      "step": 24050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3456858600969615e-06,
+      "loss": 1.4963,
+      "step": 24060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3443921573318056e-06,
+      "loss": 1.4429,
+      "step": 24070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3430981992822608e-06,
+      "loss": 1.4943,
+      "step": 24080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.341803986339527e-06,
+      "loss": 1.4878,
+      "step": 24090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3405095188948823e-06,
+      "loss": 1.3454,
+      "step": 24100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.339214797339682e-06,
+      "loss": 1.5474,
+      "step": 24110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3379198220653587e-06,
+      "loss": 1.4029,
+      "step": 24120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3366245934634224e-06,
+      "loss": 1.4441,
+      "step": 24130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3353291119254562e-06,
+      "loss": 1.4394,
+      "step": 24140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3340333778431235e-06,
+      "loss": 1.4517,
+      "step": 24150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3327373916081617e-06,
+      "loss": 1.4935,
+      "step": 24160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.331441153612386e-06,
+      "loss": 1.3882,
+      "step": 24170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3301446642476863e-06,
+      "loss": 1.4979,
+      "step": 24180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.32884792390603e-06,
+      "loss": 1.4271,
+      "step": 24190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3275509329794585e-06,
+      "loss": 1.3521,
+      "step": 24200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.326253691860091e-06,
+      "loss": 1.5124,
+      "step": 24210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.32495620094012e-06,
+      "loss": 1.3693,
+      "step": 24220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3236584606118163e-06,
+      "loss": 1.4407,
+      "step": 24230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.322360471267524e-06,
+      "loss": 1.3538,
+      "step": 24240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3210622332996633e-06,
+      "loss": 1.5233,
+      "step": 24250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3197637471007294e-06,
+      "loss": 1.4731,
+      "step": 24260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.318465013063292e-06,
+      "loss": 1.4556,
+      "step": 24270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.317166031579998e-06,
+      "loss": 1.4699,
+      "step": 24280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3158668030435657e-06,
+      "loss": 1.4676,
+      "step": 24290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3145673278467904e-06,
+      "loss": 1.3851,
+      "step": 24300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.313267606382541e-06,
+      "loss": 1.4924,
+      "step": 24310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3119676390437618e-06,
+      "loss": 1.5105,
+      "step": 24320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3106674262234707e-06,
+      "loss": 1.5132,
+      "step": 24330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.30936696831476e-06,
+      "loss": 1.3777,
+      "step": 24340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3080662657107966e-06,
+      "loss": 1.3826,
+      "step": 24350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.306765318804819e-06,
+      "loss": 1.538,
+      "step": 24360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.305464127990143e-06,
+      "loss": 1.4069,
+      "step": 24370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3041626936601556e-06,
+      "loss": 1.5348,
+      "step": 24380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3028610162083195e-06,
+      "loss": 1.4426,
+      "step": 24390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.3015590960281708e-06,
+      "loss": 1.3672,
+      "step": 24400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.300256933513315e-06,
+      "loss": 1.4553,
+      "step": 24410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.298954529057434e-06,
+      "loss": 1.4502,
+      "step": 24420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.2976518830542858e-06,
+      "loss": 1.5351,
+      "step": 24430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.296348995897696e-06,
+      "loss": 1.3687,
+      "step": 24440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.295045867981566e-06,
+      "loss": 1.4032,
+      "step": 24450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.293742499699869e-06,
+      "loss": 1.5024,
+      "step": 24460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.292438891446651e-06,
+      "loss": 1.4242,
+      "step": 24470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2911350436160317e-06,
+      "loss": 1.5027,
+      "step": 24480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2898309566022026e-06,
+      "loss": 1.3974,
+      "step": 24490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2885266307994257e-06,
+      "loss": 1.4378,
+      "step": 24500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.287222066602038e-06,
+      "loss": 1.4978,
+      "step": 24510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.285917264404446e-06,
+      "loss": 1.4774,
+      "step": 24520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.284612224601132e-06,
+      "loss": 1.539,
+      "step": 24530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.283306947586645e-06,
+      "loss": 1.436,
+      "step": 24540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2820014337556107e-06,
+      "loss": 1.494,
+      "step": 24550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2806956835027216e-06,
+      "loss": 1.4944,
+      "step": 24560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.279389697222745e-06,
+      "loss": 1.3551,
+      "step": 24570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2780834753105205e-06,
+      "loss": 1.5153,
+      "step": 24580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2767770181609544e-06,
+      "loss": 1.468,
+      "step": 24590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.275470326169028e-06,
+      "loss": 1.4483,
+      "step": 24600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.274163399729793e-06,
+      "loss": 1.4734,
+      "step": 24610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.27285623923837e-06,
+      "loss": 1.3293,
+      "step": 24620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2715488450899523e-06,
+      "loss": 1.5221,
+      "step": 24630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2702412176798045e-06,
+      "loss": 1.4231,
+      "step": 24640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2689333574032588e-06,
+      "loss": 1.3198,
+      "step": 24650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2676252646557204e-06,
+      "loss": 1.54,
+      "step": 24660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2663169398326644e-06,
+      "loss": 1.3895,
+      "step": 24670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2650083833296344e-06,
+      "loss": 1.4832,
+      "step": 24680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2636995955422456e-06,
+      "loss": 1.3789,
+      "step": 24690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2623905768661835e-06,
+      "loss": 1.4201,
+      "step": 24700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.261081327697202e-06,
+      "loss": 1.5113,
+      "step": 24710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2597718484311246e-06,
+      "loss": 1.3782,
+      "step": 24720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2584621394638465e-06,
+      "loss": 1.5333,
+      "step": 24730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2571522011913304e-06,
+      "loss": 1.3495,
+      "step": 24740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2558420340096088e-06,
+      "loss": 1.4627,
+      "step": 24750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.254531638314784e-06,
+      "loss": 1.4047,
+      "step": 24760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.253221014503025e-06,
+      "loss": 1.4198,
+      "step": 24770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.251910162970574e-06,
+      "loss": 1.4657,
+      "step": 24780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2505990841137382e-06,
+      "loss": 1.4287,
+      "step": 24790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.249287778328896e-06,
+      "loss": 1.4343,
+      "step": 24800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2479762460124926e-06,
+      "loss": 1.4996,
+      "step": 24810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2466644875610422e-06,
+      "loss": 1.4417,
+      "step": 24820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.245352503371128e-06,
+      "loss": 1.5249,
+      "step": 24830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2440402938394017e-06,
+      "loss": 1.3486,
+      "step": 24840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.242727859362583e-06,
+      "loss": 1.4112,
+      "step": 24850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.241415200337457e-06,
+      "loss": 1.4244,
+      "step": 24860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2401023171608803e-06,
+      "loss": 1.3455,
+      "step": 24870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2387892102297757e-06,
+      "loss": 1.4335,
+      "step": 24880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2374758799411343e-06,
+      "loss": 1.4041,
+      "step": 24890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2361623266920127e-06,
+      "loss": 1.4658,
+      "step": 24900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.234848550879538e-06,
+      "loss": 1.4309,
+      "step": 24910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2335345529009017e-06,
+      "loss": 1.4702,
+      "step": 24920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.232220333153363e-06,
+      "loss": 1.447,
+      "step": 24930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.230905892034252e-06,
+      "loss": 1.4112,
+      "step": 24940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.22959122994096e-06,
+      "loss": 1.3707,
+      "step": 24950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2282763472709484e-06,
+      "loss": 1.4601,
+      "step": 24960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2269612444217443e-06,
+      "loss": 1.4523,
+      "step": 24970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.225645921790942e-06,
+      "loss": 1.4643,
+      "step": 24980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2243303797762023e-06,
+      "loss": 1.4541,
+      "step": 24990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.223014618775251e-06,
+      "loss": 1.4813,
+      "step": 25000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2216986391858816e-06,
+      "loss": 1.4817,
+      "step": 25010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.220382441405953e-06,
+      "loss": 1.3466,
+      "step": 25020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.2190660258333908e-06,
+      "loss": 1.4687,
+      "step": 25030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2177493928661854e-06,
+      "loss": 1.389,
+      "step": 25040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2164325429023924e-06,
+      "loss": 1.4743,
+      "step": 25050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2151154763401356e-06,
+      "loss": 1.4578,
+      "step": 25060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.213798193577602e-06,
+      "loss": 1.3624,
+      "step": 25070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.212480695013044e-06,
+      "loss": 1.4776,
+      "step": 25080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2111629810447807e-06,
+      "loss": 1.3498,
+      "step": 25090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2098450520711954e-06,
+      "loss": 1.4656,
+      "step": 25100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2085269084907362e-06,
+      "loss": 1.5003,
+      "step": 25110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.207208550701916e-06,
+      "loss": 1.3827,
+      "step": 25120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2058899791033137e-06,
+      "loss": 1.4053,
+      "step": 25130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.204571194093572e-06,
+      "loss": 1.4178,
+      "step": 25140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.203252196071397e-06,
+      "loss": 1.3924,
+      "step": 25150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.201932985435561e-06,
+      "loss": 1.5561,
+      "step": 25160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.2006135625849e-06,
+      "loss": 1.3819,
+      "step": 25170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.199293927918314e-06,
+      "loss": 1.5353,
+      "step": 25180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.197974081834767e-06,
+      "loss": 1.4036,
+      "step": 25190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1966540247332865e-06,
+      "loss": 1.4698,
+      "step": 25200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.195333757012965e-06,
+      "loss": 1.4689,
+      "step": 25210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1940132790729577e-06,
+      "loss": 1.4893,
+      "step": 25220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.192692591312483e-06,
+      "loss": 1.5282,
+      "step": 25230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1913716941308243e-06,
+      "loss": 1.4438,
+      "step": 25240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1900505879273268e-06,
+      "loss": 1.4156,
+      "step": 25250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1887292731014007e-06,
+      "loss": 1.544,
+      "step": 25260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1874077500525157e-06,
+      "loss": 1.5061,
+      "step": 25270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.186086019180208e-06,
+      "loss": 1.435,
+      "step": 25280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.184764080884076e-06,
+      "loss": 1.3978,
+      "step": 25290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.18344193556378e-06,
+      "loss": 1.3965,
+      "step": 25300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.182119583619042e-06,
+      "loss": 1.537,
+      "step": 25310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.180797025449649e-06,
+      "loss": 1.4337,
+      "step": 25320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1794742614554474e-06,
+      "loss": 1.4782,
+      "step": 25330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1781512920363494e-06,
+      "loss": 1.3705,
+      "step": 25340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.176828117592326e-06,
+      "loss": 1.3994,
+      "step": 25350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1755047385234117e-06,
+      "loss": 1.4893,
+      "step": 25360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1741811552297017e-06,
+      "loss": 1.4209,
+      "step": 25370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1728573681113544e-06,
+      "loss": 1.4682,
+      "step": 25380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1715333775685907e-06,
+      "loss": 1.4419,
+      "step": 25390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.17020918400169e-06,
+      "loss": 1.3677,
+      "step": 25400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1688847878109945e-06,
+      "loss": 1.4574,
+      "step": 25410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.167560189396909e-06,
+      "loss": 1.3402,
+      "step": 25420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1662353891598967e-06,
+      "loss": 1.4333,
+      "step": 25430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1649103875004856e-06,
+      "loss": 1.3332,
+      "step": 25440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.16358518481926e-06,
+      "loss": 1.3837,
+      "step": 25450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1622597815168687e-06,
+      "loss": 1.4828,
+      "step": 25460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.16093417799402e-06,
+      "loss": 1.4023,
+      "step": 25470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1596083746514817e-06,
+      "loss": 1.4789,
+      "step": 25480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.158282371890083e-06,
+      "loss": 1.4033,
+      "step": 25490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.156956170110713e-06,
+      "loss": 1.457,
+      "step": 25500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.155629769714322e-06,
+      "loss": 1.5313,
+      "step": 25510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1543031711019203e-06,
+      "loss": 1.4528,
+      "step": 25520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1529763746745747e-06,
+      "loss": 1.4964,
+      "step": 25530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1516493808334165e-06,
+      "loss": 1.4148,
+      "step": 25540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1503221899796343e-06,
+      "loss": 1.3828,
+      "step": 25550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.148994802514476e-06,
+      "loss": 1.525,
+      "step": 25560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.147667218839251e-06,
+      "loss": 1.3933,
+      "step": 25570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1463394393553247e-06,
+      "loss": 1.5798,
+      "step": 25580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.145011464464125e-06,
+      "loss": 1.4245,
+      "step": 25590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1436832945671362e-06,
+      "loss": 1.4037,
+      "step": 25600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1423549300659046e-06,
+      "loss": 1.4645,
+      "step": 25610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.1410263713620326e-06,
+      "loss": 1.4222,
+      "step": 25620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1396976188571815e-06,
+      "loss": 1.5116,
+      "step": 25630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1383686729530727e-06,
+      "loss": 1.3572,
+      "step": 25640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.137039534051486e-06,
+      "loss": 1.3396,
+      "step": 25650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.135710202554258e-06,
+      "loss": 1.4981,
+      "step": 25660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1343806788632846e-06,
+      "loss": 1.4069,
+      "step": 25670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1330509633805183e-06,
+      "loss": 1.5656,
+      "step": 25680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1317210565079734e-06,
+      "loss": 1.4872,
+      "step": 25690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.130390958647718e-06,
+      "loss": 1.3655,
+      "step": 25700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.129060670201881e-06,
+      "loss": 1.5521,
+      "step": 25710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1277301915726456e-06,
+      "loss": 1.4528,
+      "step": 25720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.126399523162254e-06,
+      "loss": 1.5462,
+      "step": 25730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.125068665373008e-06,
+      "loss": 1.4107,
+      "step": 25740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.123737618607264e-06,
+      "loss": 1.3648,
+      "step": 25750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1224063832674358e-06,
+      "loss": 1.4708,
+      "step": 25760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1210749597559942e-06,
+      "loss": 1.478,
+      "step": 25770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1197433484754693e-06,
+      "loss": 1.4656,
+      "step": 25780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.118411549828444e-06,
+      "loss": 1.4147,
+      "step": 25790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1170795642175614e-06,
+      "loss": 1.353,
+      "step": 25800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1157473920455188e-06,
+      "loss": 1.46,
+      "step": 25810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1144150337150716e-06,
+      "loss": 1.3942,
+      "step": 25820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.113082489629029e-06,
+      "loss": 1.5101,
+      "step": 25830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1117497601902596e-06,
+      "loss": 1.378,
+      "step": 25840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1104168458016853e-06,
+      "loss": 1.4179,
+      "step": 25850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1090837468662853e-06,
+      "loss": 1.4178,
+      "step": 25860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.107750463787095e-06,
+      "loss": 1.4117,
+      "step": 25870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1064169969672043e-06,
+      "loss": 1.4712,
+      "step": 25880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1050833468097585e-06,
+      "loss": 1.4337,
+      "step": 25890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1037495137179595e-06,
+      "loss": 1.4234,
+      "step": 25900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1024154980950645e-06,
+      "loss": 1.5127,
+      "step": 25910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.1010813003443847e-06,
+      "loss": 1.3649,
+      "step": 25920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.099746920869287e-06,
+      "loss": 1.5283,
+      "step": 25930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0984123600731936e-06,
+      "loss": 1.4519,
+      "step": 25940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0970776183595806e-06,
+      "loss": 1.4163,
+      "step": 25950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.09574269613198e-06,
+      "loss": 1.4987,
+      "step": 25960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0944075937939776e-06,
+      "loss": 1.447,
+      "step": 25970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.093072311749214e-06,
+      "loss": 1.4866,
+      "step": 25980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0917368504013835e-06,
+      "loss": 1.4005,
+      "step": 25990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0904012101542346e-06,
+      "loss": 1.4061,
+      "step": 26000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0890653914115713e-06,
+      "loss": 1.5149,
+      "step": 26010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0877293945772506e-06,
+      "loss": 1.439,
+      "step": 26020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0863932200551814e-06,
+      "loss": 1.4999,
+      "step": 26030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0850568682493315e-06,
+      "loss": 1.4302,
+      "step": 26040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.083720339563716e-06,
+      "loss": 1.4196,
+      "step": 26050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.082383634402409e-06,
+      "loss": 1.475,
+      "step": 26060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0810467531695344e-06,
+      "loss": 1.4312,
+      "step": 26070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.079709696269269e-06,
+      "loss": 1.4493,
+      "step": 26080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.078372464105846e-06,
+      "loss": 1.339,
+      "step": 26090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0770350570835487e-06,
+      "loss": 1.4384,
+      "step": 26100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.075697475606716e-06,
+      "loss": 1.5258,
+      "step": 26110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.074359720079736e-06,
+      "loss": 1.393,
+      "step": 26120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0730217909070513e-06,
+      "loss": 1.5068,
+      "step": 26130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0716836884931573e-06,
+      "loss": 1.4504,
+      "step": 26140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.070345413242601e-06,
+      "loss": 1.4486,
+      "step": 26150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.069006965559985e-06,
+      "loss": 1.567,
+      "step": 26160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0676683458499566e-06,
+      "loss": 1.3861,
+      "step": 26170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.066329554517222e-06,
+      "loss": 1.4344,
+      "step": 26180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.064990591966536e-06,
+      "loss": 1.4173,
+      "step": 26190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.063651458602708e-06,
+      "loss": 1.3797,
+      "step": 26200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.0623121548305947e-06,
+      "loss": 1.538,
+      "step": 26210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.060972681055108e-06,
+      "loss": 1.3946,
+      "step": 26220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0596330376812095e-06,
+      "loss": 1.4884,
+      "step": 26230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.058293225113912e-06,
+      "loss": 1.412,
+      "step": 26240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0569532437582817e-06,
+      "loss": 1.5377,
+      "step": 26250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0556130940194325e-06,
+      "loss": 1.5174,
+      "step": 26260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.054272776302531e-06,
+      "loss": 1.4814,
+      "step": 26270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0529322910127944e-06,
+      "loss": 1.4531,
+      "step": 26280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.05159163855549e-06,
+      "loss": 1.5001,
+      "step": 26290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.050250819335938e-06,
+      "loss": 1.4335,
+      "step": 26300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0489098337595058e-06,
+      "loss": 1.4495,
+      "step": 26310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0475686822316126e-06,
+      "loss": 1.456,
+      "step": 26320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.046227365157728e-06,
+      "loss": 1.4239,
+      "step": 26330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0448858829433703e-06,
+      "loss": 1.4113,
+      "step": 26340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.04354423599411e-06,
+      "loss": 1.3821,
+      "step": 26350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.042202424715565e-06,
+      "loss": 1.5134,
+      "step": 26360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.040860449513405e-06,
+      "loss": 1.3947,
+      "step": 26370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.039518310793348e-06,
+      "loss": 1.5225,
+      "step": 26380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0381760089611607e-06,
+      "loss": 1.4167,
+      "step": 26390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0368335444226614e-06,
+      "loss": 1.4418,
+      "step": 26400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.035490917583716e-06,
+      "loss": 1.4713,
+      "step": 26410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0341481288502394e-06,
+      "loss": 1.351,
+      "step": 26420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0328051786281964e-06,
+      "loss": 1.5229,
+      "step": 26430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0314620673235995e-06,
+      "loss": 1.3953,
+      "step": 26440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0301187953425106e-06,
+      "loss": 1.346,
+      "step": 26450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0287753630910393e-06,
+      "loss": 1.3758,
+      "step": 26460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.027431770975346e-06,
+      "loss": 1.4829,
+      "step": 26470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.026088019401637e-06,
+      "loss": 1.5063,
+      "step": 26480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.024744108776167e-06,
+      "loss": 1.4513,
+      "step": 26490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.02340003950524e-06,
+      "loss": 1.4257,
+      "step": 26500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0220558119952075e-06,
+      "loss": 1.5235,
+      "step": 26510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.020711426652469e-06,
+      "loss": 1.3875,
+      "step": 26520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0193668838834707e-06,
+      "loss": 1.4696,
+      "step": 26530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.018022184094706e-06,
+      "loss": 1.4427,
+      "step": 26540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.016677327692719e-06,
+      "loss": 1.398,
+      "step": 26550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0153323150840987e-06,
+      "loss": 1.5533,
+      "step": 26560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0139871466754817e-06,
+      "loss": 1.3681,
+      "step": 26570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0126418228735506e-06,
+      "loss": 1.429,
+      "step": 26580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0112963440850355e-06,
+      "loss": 1.4147,
+      "step": 26590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0099507107167158e-06,
+      "loss": 1.4325,
+      "step": 26600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.008604923175416e-06,
+      "loss": 1.4961,
+      "step": 26610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0072589818680052e-06,
+      "loss": 1.3594,
+      "step": 26620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.005912887201401e-06,
+      "loss": 1.5081,
+      "step": 26630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.004566639582568e-06,
+      "loss": 1.4314,
+      "step": 26640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.003220239418515e-06,
+      "loss": 1.4001,
+      "step": 26650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.0018736871163e-06,
+      "loss": 1.5022,
+      "step": 26660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.000526983083023e-06,
+      "loss": 1.468,
+      "step": 26670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9991801277258327e-06,
+      "loss": 1.5449,
+      "step": 26680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.997833121451923e-06,
+      "loss": 1.3618,
+      "step": 26690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.996485964668533e-06,
+      "loss": 1.3946,
+      "step": 26700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.995138657782947e-06,
+      "loss": 1.4129,
+      "step": 26710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.993791201202496e-06,
+      "loss": 1.4489,
+      "step": 26720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9924435953345553e-06,
+      "loss": 1.4496,
+      "step": 26730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.991095840586546e-06,
+      "loss": 1.4117,
+      "step": 26740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9897479373659323e-06,
+      "loss": 1.3815,
+      "step": 26750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9883998860802253e-06,
+      "loss": 1.4586,
+      "step": 26760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9870516871369804e-06,
+      "loss": 1.408,
+      "step": 26770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9857033409437975e-06,
+      "loss": 1.5805,
+      "step": 26780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.984354847908321e-06,
+      "loss": 1.3888,
+      "step": 26790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2.9830062084382393e-06,
+      "loss": 1.4265,
+      "step": 26800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9816574229412866e-06,
+      "loss": 1.4391,
+      "step": 26810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9803084918252384e-06,
+      "loss": 1.4231,
+      "step": 26820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9789594154979173e-06,
+      "loss": 1.4473,
+      "step": 26830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9776101943671875e-06,
+      "loss": 1.3981,
+      "step": 26840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.976260828840959e-06,
+      "loss": 1.3973,
+      "step": 26850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.974911319327183e-06,
+      "loss": 1.5386,
+      "step": 26860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.973561666233857e-06,
+      "loss": 1.4648,
+      "step": 26870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9722118699690212e-06,
+      "loss": 1.4748,
+      "step": 26880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.970861930940755e-06,
+      "loss": 1.3008,
+      "step": 26890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9695118495571874e-06,
+      "loss": 1.4145,
+      "step": 26900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.968161626226486e-06,
+      "loss": 1.5258,
+      "step": 26910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9668112613568632e-06,
+      "loss": 1.4193,
+      "step": 26920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.965460755356575e-06,
+      "loss": 1.469,
+      "step": 26930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9641101086339164e-06,
+      "loss": 1.3601,
+      "step": 26940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9627593215972276e-06,
+      "loss": 1.4174,
+      "step": 26950
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9614083946548927e-06,
+      "loss": 1.4981,
+      "step": 26960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.960057328215336e-06,
+      "loss": 1.4029,
+      "step": 26970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.958706122687023e-06,
+      "loss": 1.4595,
+      "step": 26980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.957354778478463e-06,
+      "loss": 1.3447,
+      "step": 26990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9560032959982067e-06,
+      "loss": 1.3938,
+      "step": 27000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.954651675654847e-06,
+      "loss": 1.5145,
+      "step": 27010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9532999178570196e-06,
+      "loss": 1.4433,
+      "step": 27020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.951948023013398e-06,
+      "loss": 1.5196,
+      "step": 27030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9505959915327004e-06,
+      "loss": 1.5282,
+      "step": 27040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.949243823823685e-06,
+      "loss": 1.4287,
+      "step": 27050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9478915202951524e-06,
+      "loss": 1.565,
+      "step": 27060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9465390813559426e-06,
+      "loss": 1.4284,
+      "step": 27070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9451865074149378e-06,
+      "loss": 1.4006,
+      "step": 27080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9438337988810605e-06,
+      "loss": 1.4729,
+      "step": 27090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9424809561632734e-06,
+      "loss": 1.425,
+      "step": 27100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9411279796705816e-06,
+      "loss": 1.4543,
+      "step": 27110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.939774869812029e-06,
+      "loss": 1.4339,
+      "step": 27120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9384216269966986e-06,
+      "loss": 1.4098,
+      "step": 27130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9370682516337174e-06,
+      "loss": 1.4146,
+      "step": 27140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9357147441322488e-06,
+      "loss": 1.4229,
+      "step": 27150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.934361104901498e-06,
+      "loss": 1.4896,
+      "step": 27160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9330073343507104e-06,
+      "loss": 1.3886,
+      "step": 27170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.931653432889169e-06,
+      "loss": 1.4682,
+      "step": 27180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.930299400926199e-06,
+      "loss": 1.4332,
+      "step": 27190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9289452388711624e-06,
+      "loss": 1.3448,
+      "step": 27200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9275909471334628e-06,
+      "loss": 1.5305,
+      "step": 27210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.926236526122542e-06,
+      "loss": 1.5677,
+      "step": 27220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.924881976247882e-06,
+      "loss": 1.5161,
+      "step": 27230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9235272979190014e-06,
+      "loss": 1.3933,
+      "step": 27240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9221724915454587e-06,
+      "loss": 1.4179,
+      "step": 27250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.920817557536852e-06,
+      "loss": 1.4562,
+      "step": 27260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9194624963028177e-06,
+      "loss": 1.3993,
+      "step": 27270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9181073082530302e-06,
+      "loss": 1.44,
+      "step": 27280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.916751993797203e-06,
+      "loss": 1.4313,
+      "step": 27290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.915396553345085e-06,
+      "loss": 1.3983,
+      "step": 27300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.914040987306468e-06,
+      "loss": 1.4684,
+      "step": 27310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9126852960911773e-06,
+      "loss": 1.4478,
+      "step": 27320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.911329480109079e-06,
+      "loss": 1.4983,
+      "step": 27330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.909973539770076e-06,
+      "loss": 1.4242,
+      "step": 27340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.908617475484108e-06,
+      "loss": 1.4137,
+      "step": 27350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9072612876611524e-06,
+      "loss": 1.4258,
+      "step": 27360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.905904976711225e-06,
+      "loss": 1.4235,
+      "step": 27370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9045485430443793e-06,
+      "loss": 1.4887,
+      "step": 27380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2.9031919870707025e-06,
+      "loss": 1.4044,
+      "step": 27390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9018353092003216e-06,
+      "loss": 1.3704,
+      "step": 27400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.9004785098434007e-06,
+      "loss": 1.4671,
+      "step": 27410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8991215894101403e-06,
+      "loss": 1.4675,
+      "step": 27420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8977645483107766e-06,
+      "loss": 1.4368,
+      "step": 27430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.896407386955582e-06,
+      "loss": 1.3239,
+      "step": 27440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8950501057548654e-06,
+      "loss": 1.4091,
+      "step": 27450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8936927051189743e-06,
+      "loss": 1.4481,
+      "step": 27460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8923351854582904e-06,
+      "loss": 1.3784,
+      "step": 27470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.890977547183231e-06,
+      "loss": 1.5709,
+      "step": 27480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.889619790704249e-06,
+      "loss": 1.4652,
+      "step": 27490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.888261916431834e-06,
+      "loss": 1.4533,
+      "step": 27500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8869039247765116e-06,
+      "loss": 1.4872,
+      "step": 27510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.885545816148843e-06,
+      "loss": 1.3662,
+      "step": 27520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.884187590959423e-06,
+      "loss": 1.5041,
+      "step": 27530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8828292496188824e-06,
+      "loss": 1.4068,
+      "step": 27540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8814707925378886e-06,
+      "loss": 1.403,
+      "step": 27550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.880112220127142e-06,
+      "loss": 1.4084,
+      "step": 27560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.878753532797379e-06,
+      "loss": 1.4109,
+      "step": 27570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8773947309593698e-06,
+      "loss": 1.4933,
+      "step": 27580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.876035815023921e-06,
+      "loss": 1.371,
+      "step": 27590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8746767854018715e-06,
+      "loss": 1.411,
+      "step": 27600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.873317642504096e-06,
+      "loss": 1.4918,
+      "step": 27610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8719583867415035e-06,
+      "loss": 1.3905,
+      "step": 27620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.870599018525036e-06,
+      "loss": 1.4835,
+      "step": 27630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8692395382656702e-06,
+      "loss": 1.4399,
+      "step": 27640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8678799463744174e-06,
+      "loss": 1.3917,
+      "step": 27650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8665202432623213e-06,
+      "loss": 1.5572,
+      "step": 27660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8651604293404594e-06,
+      "loss": 1.5007,
+      "step": 27670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8638005050199443e-06,
+      "loss": 1.4877,
+      "step": 27680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8624404707119197e-06,
+      "loss": 1.5556,
+      "step": 27690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8610803268275643e-06,
+      "loss": 1.3931,
+      "step": 27700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8597200737780893e-06,
+      "loss": 1.5138,
+      "step": 27710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8583597119747385e-06,
+      "loss": 1.3579,
+      "step": 27720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8569992418287896e-06,
+      "loss": 1.5166,
+      "step": 27730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.855638663751552e-06,
+      "loss": 1.4269,
+      "step": 27740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8542779781543674e-06,
+      "loss": 1.3656,
+      "step": 27750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.852917185448612e-06,
+      "loss": 1.4519,
+      "step": 27760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8515562860456923e-06,
+      "loss": 1.3896,
+      "step": 27770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8501952803570486e-06,
+      "loss": 1.4758,
+      "step": 27780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8488341687941535e-06,
+      "loss": 1.3492,
+      "step": 27790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.847472951768508e-06,
+      "loss": 1.4773,
+      "step": 27800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8461116296916493e-06,
+      "loss": 1.5039,
+      "step": 27810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.844750202975145e-06,
+      "loss": 1.391,
+      "step": 27820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8433886720305946e-06,
+      "loss": 1.4653,
+      "step": 27830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8420270372696273e-06,
+      "loss": 1.3921,
+      "step": 27840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.840665299103906e-06,
+      "loss": 1.376,
+      "step": 27850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.839303457945123e-06,
+      "loss": 1.4167,
+      "step": 27860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.837941514205003e-06,
+      "loss": 1.3807,
+      "step": 27870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.836579468295303e-06,
+      "loss": 1.5572,
+      "step": 27880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8352173206278065e-06,
+      "loss": 1.3737,
+      "step": 27890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.833855071614332e-06,
+      "loss": 1.3831,
+      "step": 27900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8324927216667265e-06,
+      "loss": 1.4135,
+      "step": 27910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.831130271196869e-06,
+      "loss": 1.4236,
+      "step": 27920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8297677206166673e-06,
+      "loss": 1.4904,
+      "step": 27930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8284050703380604e-06,
+      "loss": 1.4499,
+      "step": 27940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8270423207730175e-06,
+      "loss": 1.4375,
+      "step": 27950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.825679472333537e-06,
+      "loss": 1.4541,
+      "step": 27960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2.8243165254316483e-06,
+      "loss": 1.4545,
+      "step": 27970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.82295348047941e-06,
+      "loss": 1.5003,
+      "step": 27980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8215903378889096e-06,
+      "loss": 1.3922,
+      "step": 27990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8202270980722663e-06,
+      "loss": 1.438,
+      "step": 28000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.818863761441626e-06,
+      "loss": 1.4623,
+      "step": 28010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8175003284091657e-06,
+      "loss": 1.4521,
+      "step": 28020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.816136799387091e-06,
+      "loss": 1.4827,
+      "step": 28030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.814773174787636e-06,
+      "loss": 1.3774,
+      "step": 28040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8134094550230644e-06,
+      "loss": 1.4636,
+      "step": 28050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8120456405056697e-06,
+      "loss": 1.4744,
+      "step": 28060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.810681731647771e-06,
+      "loss": 1.418,
+      "step": 28070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8093177288617187e-06,
+      "loss": 1.5243,
+      "step": 28080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8079536325598905e-06,
+      "loss": 1.3501,
+      "step": 28090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.806589443154693e-06,
+      "loss": 1.3944,
+      "step": 28100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.805225161058559e-06,
+      "loss": 1.4947,
+      "step": 28110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8038607866839522e-06,
+      "loss": 1.4087,
+      "step": 28120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8024963204433625e-06,
+      "loss": 1.4641,
+      "step": 28130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.8011317627493078e-06,
+      "loss": 1.4398,
+      "step": 28140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7997671140143345e-06,
+      "loss": 1.3468,
+      "step": 28150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7984023746510135e-06,
+      "loss": 1.4949,
+      "step": 28160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.797037545071947e-06,
+      "loss": 1.4034,
+      "step": 28170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7956726256897625e-06,
+      "loss": 1.5054,
+      "step": 28180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.794307616917116e-06,
+      "loss": 1.3798,
+      "step": 28190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.792942519166688e-06,
+      "loss": 1.4318,
+      "step": 28200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.791577332851187e-06,
+      "loss": 1.5125,
+      "step": 28210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7902120583833505e-06,
+      "loss": 1.4033,
+      "step": 28220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.788846696175939e-06,
+      "loss": 1.4788,
+      "step": 28230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7874812466417434e-06,
+      "loss": 1.3975,
+      "step": 28240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7861157101935766e-06,
+      "loss": 1.4547,
+      "step": 28250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7847500872442812e-06,
+      "loss": 1.5513,
+      "step": 28260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.783384378206725e-06,
+      "loss": 1.3773,
+      "step": 28270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7820185834938024e-06,
+      "loss": 1.4839,
+      "step": 28280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7806527035184316e-06,
+      "loss": 1.4252,
+      "step": 28290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7792867386935592e-06,
+      "loss": 1.4716,
+      "step": 28300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.777920689432154e-06,
+      "loss": 1.4978,
+      "step": 28310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7765545561472153e-06,
+      "loss": 1.4561,
+      "step": 28320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7751883392517643e-06,
+      "loss": 1.4582,
+      "step": 28330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.773822039158847e-06,
+      "loss": 1.3784,
+      "step": 28340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7724556562815374e-06,
+      "loss": 1.419,
+      "step": 28350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7710891910329317e-06,
+      "loss": 1.4473,
+      "step": 28360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7697226438261527e-06,
+      "loss": 1.413,
+      "step": 28370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7683560150743477e-06,
+      "loss": 1.4895,
+      "step": 28380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7669893051906874e-06,
+      "loss": 1.338,
+      "step": 28390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.76562251458837e-06,
+      "loss": 1.3552,
+      "step": 28400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7642556436806143e-06,
+      "loss": 1.5178,
+      "step": 28410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7628886928806664e-06,
+      "loss": 1.3924,
+      "step": 28420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7615216626017955e-06,
+      "loss": 1.4736,
+      "step": 28430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.760154553257294e-06,
+      "loss": 1.4147,
+      "step": 28440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7587873652604787e-06,
+      "loss": 1.3958,
+      "step": 28450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7574200990246912e-06,
+      "loss": 1.4633,
+      "step": 28460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7560527549632964e-06,
+      "loss": 1.3884,
+      "step": 28470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7546853334896813e-06,
+      "loss": 1.5459,
+      "step": 28480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7533178350172574e-06,
+      "loss": 1.3983,
+      "step": 28490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.75195025995946e-06,
+      "loss": 1.4439,
+      "step": 28500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7505826087297464e-06,
+      "loss": 1.4815,
+      "step": 28510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.749214881741598e-06,
+      "loss": 1.4183,
+      "step": 28520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.747847079408519e-06,
+      "loss": 1.4665,
+      "step": 28530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.746479202144034e-06,
+      "loss": 1.4652,
+      "step": 28540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.745111250361694e-06,
+      "loss": 1.3987,
+      "step": 28550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2.7437432244750705e-06,
+      "loss": 1.5246,
+      "step": 28560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.742375124897757e-06,
+      "loss": 1.494,
+      "step": 28570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.74100695204337e-06,
+      "loss": 1.5241,
+      "step": 28580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.739638706325549e-06,
+      "loss": 1.4631,
+      "step": 28590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.738270388157954e-06,
+      "loss": 1.4542,
+      "step": 28600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7369019979542673e-06,
+      "loss": 1.5046,
+      "step": 28610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7355335361281934e-06,
+      "loss": 1.3785,
+      "step": 28620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.734165003093458e-06,
+      "loss": 1.4688,
+      "step": 28630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.732796399263809e-06,
+      "loss": 1.3723,
+      "step": 28640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7314277250530157e-06,
+      "loss": 1.4241,
+      "step": 28650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7300589808748666e-06,
+      "loss": 1.4655,
+      "step": 28660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7286901671431742e-06,
+      "loss": 1.3776,
+      "step": 28670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.727321284271771e-06,
+      "loss": 1.4413,
+      "step": 28680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7259523326745102e-06,
+      "loss": 1.4603,
+      "step": 28690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7245833127652656e-06,
+      "loss": 1.449,
+      "step": 28700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.723214224957931e-06,
+      "loss": 1.5163,
+      "step": 28710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.721845069666422e-06,
+      "loss": 1.3313,
+      "step": 28720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7204758473046754e-06,
+      "loss": 1.5347,
+      "step": 28730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.719106558286647e-06,
+      "loss": 1.4337,
+      "step": 28740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7177372030263116e-06,
+      "loss": 1.3697,
+      "step": 28750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7163677819376653e-06,
+      "loss": 1.4623,
+      "step": 28760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.714998295434724e-06,
+      "loss": 1.4348,
+      "step": 28770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7136287439315253e-06,
+      "loss": 1.4179,
+      "step": 28780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.712259127842123e-06,
+      "loss": 1.4325,
+      "step": 28790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7108894475805912e-06,
+      "loss": 1.4737,
+      "step": 28800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.709519703561026e-06,
+      "loss": 1.4348,
+      "step": 28810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.70814989619754e-06,
+      "loss": 1.4241,
+      "step": 28820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7067800259042665e-06,
+      "loss": 1.4912,
+      "step": 28830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.705410093095357e-06,
+      "loss": 1.4137,
+      "step": 28840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.7040400981849817e-06,
+      "loss": 1.358,
+      "step": 28850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.702670041587331e-06,
+      "loss": 1.4449,
+      "step": 28860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.701299923716612e-06,
+      "loss": 1.3361,
+      "step": 28870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6999297449870522e-06,
+      "loss": 1.4547,
+      "step": 28880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6985595058128957e-06,
+      "loss": 1.3683,
+      "step": 28890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.697189206608406e-06,
+      "loss": 1.5136,
+      "step": 28900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6958188477878655e-06,
+      "loss": 1.4977,
+      "step": 28910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6944484297655727e-06,
+      "loss": 1.4472,
+      "step": 28920
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6930779529558447e-06,
+      "loss": 1.554,
+      "step": 28930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.691707417773018e-06,
+      "loss": 1.3803,
+      "step": 28940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6903368246314445e-06,
+      "loss": 1.4,
+      "step": 28950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6889661739454946e-06,
+      "loss": 1.5043,
+      "step": 28960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.687595466129555e-06,
+      "loss": 1.411,
+      "step": 28970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.686224701598032e-06,
+      "loss": 1.4547,
+      "step": 28980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.684853880765348e-06,
+      "loss": 1.4406,
+      "step": 28990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.683483004045941e-06,
+      "loss": 1.4136,
+      "step": 29000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6821120718542674e-06,
+      "loss": 1.3697,
+      "step": 29010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6807410846047997e-06,
+      "loss": 1.4635,
+      "step": 29020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6793700427120283e-06,
+      "loss": 1.5556,
+      "step": 29030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.677998946590459e-06,
+      "loss": 1.3757,
+      "step": 29040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.676627796654614e-06,
+      "loss": 1.4034,
+      "step": 29050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.675256593319031e-06,
+      "loss": 1.4441,
+      "step": 29060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6738853369982653e-06,
+      "loss": 1.3738,
+      "step": 29070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6725140281068882e-06,
+      "loss": 1.4347,
+      "step": 29080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6711426670594863e-06,
+      "loss": 1.3882,
+      "step": 29090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6697712542706626e-06,
+      "loss": 1.3562,
+      "step": 29100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.668399790155034e-06,
+      "loss": 1.4941,
+      "step": 29110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6670282751272335e-06,
+      "loss": 1.3823,
+      "step": 29120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6656567096019128e-06,
+      "loss": 1.4929,
+      "step": 29130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.664285093993734e-06,
+      "loss": 1.4127,
+      "step": 29140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2.6629134287173767e-06,
+      "loss": 1.4302,
+      "step": 29150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6615417141875367e-06,
+      "loss": 1.4768,
+      "step": 29160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6601699508189216e-06,
+      "loss": 1.4373,
+      "step": 29170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6587981390262563e-06,
+      "loss": 1.4484,
+      "step": 29180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6574262792242807e-06,
+      "loss": 1.3462,
+      "step": 29190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.656054371827746e-06,
+      "loss": 1.4601,
+      "step": 29200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.654682417251421e-06,
+      "loss": 1.5215,
+      "step": 29210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6533104159100868e-06,
+      "loss": 1.4009,
+      "step": 29220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6519383682185407e-06,
+      "loss": 1.5284,
+      "step": 29230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6505662745915915e-06,
+      "loss": 1.3279,
+      "step": 29240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.649194135444064e-06,
+      "loss": 1.4559,
+      "step": 29250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6478219511907954e-06,
+      "loss": 1.477,
+      "step": 29260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6464497222466377e-06,
+      "loss": 1.4256,
+      "step": 29270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6450774490264547e-06,
+      "loss": 1.4784,
+      "step": 29280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6437051319451256e-06,
+      "loss": 1.3762,
+      "step": 29290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6423327714175416e-06,
+      "loss": 1.3423,
+      "step": 29300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.640960367858607e-06,
+      "loss": 1.4003,
+      "step": 29310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.63958792168324e-06,
+      "loss": 1.4141,
+      "step": 29320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.638215433306371e-06,
+      "loss": 1.5162,
+      "step": 29330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6368429031429425e-06,
+      "loss": 1.4423,
+      "step": 29340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.635470331607911e-06,
+      "loss": 1.2504,
+      "step": 29350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6340977191162457e-06,
+      "loss": 1.4316,
+      "step": 29360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.632725066082926e-06,
+      "loss": 1.3998,
+      "step": 29370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.631352372922946e-06,
+      "loss": 1.5012,
+      "step": 29380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.629979640051311e-06,
+      "loss": 1.354,
+      "step": 29390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6286068678830375e-06,
+      "loss": 1.4331,
+      "step": 29400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.627234056833155e-06,
+      "loss": 1.4978,
+      "step": 29410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.625861207316703e-06,
+      "loss": 1.4094,
+      "step": 29420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6244883197487366e-06,
+      "loss": 1.519,
+      "step": 29430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.623115394544318e-06,
+      "loss": 1.3928,
+      "step": 29440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6217424321185224e-06,
+      "loss": 1.3971,
+      "step": 29450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6203694328864378e-06,
+      "loss": 1.4058,
+      "step": 29460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.61899639726316e-06,
+      "loss": 1.3286,
+      "step": 29470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.617623325663799e-06,
+      "loss": 1.5471,
+      "step": 29480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6162502185034746e-06,
+      "loss": 1.4519,
+      "step": 29490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6148770761973168e-06,
+      "loss": 1.4063,
+      "step": 29500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6135038991604665e-06,
+      "loss": 1.4491,
+      "step": 29510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6121306878080753e-06,
+      "loss": 1.3353,
+      "step": 29520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.610757442555304e-06,
+      "loss": 1.4875,
+      "step": 29530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.609384163817327e-06,
+      "loss": 1.3984,
+      "step": 29540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.608010852009325e-06,
+      "loss": 1.3309,
+      "step": 29550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.60663750754649e-06,
+      "loss": 1.4722,
+      "step": 29560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.605264130844025e-06,
+      "loss": 1.4161,
+      "step": 29570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6038907223171408e-06,
+      "loss": 1.4794,
+      "step": 29580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6025172823810597e-06,
+      "loss": 1.3594,
+      "step": 29590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.6011438114510135e-06,
+      "loss": 1.3901,
+      "step": 29600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5997703099422406e-06,
+      "loss": 1.5603,
+      "step": 29610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.598396778269992e-06,
+      "loss": 1.4283,
+      "step": 29620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5970232168495247e-06,
+      "loss": 1.5243,
+      "step": 29630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5956496260961094e-06,
+      "loss": 1.3335,
+      "step": 29640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.59427600642502e-06,
+      "loss": 1.4669,
+      "step": 29650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5929023582515426e-06,
+      "loss": 1.5486,
+      "step": 29660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.591528681990972e-06,
+      "loss": 1.4123,
+      "step": 29670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5901549780586085e-06,
+      "loss": 1.476,
+      "step": 29680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5887812468697652e-06,
+      "loss": 1.4444,
+      "step": 29690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5874074888397604e-06,
+      "loss": 1.4084,
+      "step": 29700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.586033704383921e-06,
+      "loss": 1.5234,
+      "step": 29710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.5846598939175816e-06,
+      "loss": 1.4293,
+      "step": 29720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.583286057856086e-06,
+      "loss": 1.4568,
+      "step": 29730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2.581912196614784e-06,
+      "loss": 1.4382,
+      "step": 29740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5805383106090353e-06,
+      "loss": 1.3706,
+      "step": 29750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5791644002542054e-06,
+      "loss": 1.4833,
+      "step": 29760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5777904659656664e-06,
+      "loss": 1.4558,
+      "step": 29770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5764165081588e-06,
+      "loss": 1.5365,
+      "step": 29780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5750425272489927e-06,
+      "loss": 1.3881,
+      "step": 29790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.57366852365164e-06,
+      "loss": 1.439,
+      "step": 29800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5722944977821428e-06,
+      "loss": 1.5417,
+      "step": 29810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5709204500559093e-06,
+      "loss": 1.3718,
+      "step": 29820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5695463808883536e-06,
+      "loss": 1.4858,
+      "step": 29830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5681722906948977e-06,
+      "loss": 1.3609,
+      "step": 29840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.566798179890969e-06,
+      "loss": 1.3799,
+      "step": 29850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.565424048892001e-06,
+      "loss": 1.5202,
+      "step": 29860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.564049898113435e-06,
+      "loss": 1.3381,
+      "step": 29870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5626757279707143e-06,
+      "loss": 1.5345,
+      "step": 29880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5613015388792924e-06,
+      "loss": 1.3566,
+      "step": 29890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.559927331254626e-06,
+      "loss": 1.4793,
+      "step": 29900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5585531055121805e-06,
+      "loss": 1.4427,
+      "step": 29910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.557178862067421e-06,
+      "loss": 1.3817,
+      "step": 29920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.555804601335822e-06,
+      "loss": 1.4107,
+      "step": 29930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.554430323732865e-06,
+      "loss": 1.3951,
+      "step": 29940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5530560296740326e-06,
+      "loss": 1.4315,
+      "step": 29950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5516817195748143e-06,
+      "loss": 1.4942,
+      "step": 29960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5503073938507036e-06,
+      "loss": 1.3688,
+      "step": 29970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.548933052917199e-06,
+      "loss": 1.4974,
+      "step": 29980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5475586971898052e-06,
+      "loss": 1.3428,
+      "step": 29990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.54618432708403e-06,
+      "loss": 1.328,
+      "step": 30000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5448099430153846e-06,
+      "loss": 1.5044,
+      "step": 30010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5434355453993855e-06,
+      "loss": 1.4922,
+      "step": 30020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5420611346515526e-06,
+      "loss": 1.4829,
+      "step": 30030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5406867111874127e-06,
+      "loss": 1.4165,
+      "step": 30040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.539312275422493e-06,
+      "loss": 1.4187,
+      "step": 30050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.537937827772324e-06,
+      "loss": 1.4581,
+      "step": 30060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5365633686524433e-06,
+      "loss": 1.4041,
+      "step": 30070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.535188898478389e-06,
+      "loss": 1.4747,
+      "step": 30080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.533814417665704e-06,
+      "loss": 1.4404,
+      "step": 30090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5324399266299336e-06,
+      "loss": 1.4327,
+      "step": 30100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.531065425786627e-06,
+      "loss": 1.5797,
+      "step": 30110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.529690915551335e-06,
+      "loss": 1.4685,
+      "step": 30120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5283163963396134e-06,
+      "loss": 1.3993,
+      "step": 30130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5269418685670184e-06,
+      "loss": 1.3204,
+      "step": 30140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.525567332649111e-06,
+      "loss": 1.4551,
+      "step": 30150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.524192789001452e-06,
+      "loss": 1.5083,
+      "step": 30160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5228182380396078e-06,
+      "loss": 1.4886,
+      "step": 30170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.521443680179144e-06,
+      "loss": 1.3562,
+      "step": 30180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5200691158356295e-06,
+      "loss": 1.4033,
+      "step": 30190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.518694545424636e-06,
+      "loss": 1.4674,
+      "step": 30200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5173199693617355e-06,
+      "loss": 1.4894,
+      "step": 30210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.515945388062503e-06,
+      "loss": 1.4034,
+      "step": 30220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5145708019425146e-06,
+      "loss": 1.4866,
+      "step": 30230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5131962114173474e-06,
+      "loss": 1.4379,
+      "step": 30240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5118216169025804e-06,
+      "loss": 1.4438,
+      "step": 30250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5104470188137935e-06,
+      "loss": 1.5526,
+      "step": 30260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5090724175665694e-06,
+      "loss": 1.4439,
+      "step": 30270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.507697813576487e-06,
+      "loss": 1.4901,
+      "step": 30280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5063232072591307e-06,
+      "loss": 1.4649,
+      "step": 30290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5049485990300855e-06,
+      "loss": 1.3634,
+      "step": 30300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5035739893049336e-06,
+      "loss": 1.5186,
+      "step": 30310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5021993784992616e-06,
+      "loss": 1.4774,
+      "step": 30320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2.5008247670286517e-06,
+      "loss": 1.3973,
+      "step": 30330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4994501553086915e-06,
+      "loss": 1.3454,
+      "step": 30340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.498075543754965e-06,
+      "loss": 1.3165,
+      "step": 30350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4967009327830584e-06,
+      "loss": 1.4715,
+      "step": 30360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4953263228085557e-06,
+      "loss": 1.4428,
+      "step": 30370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.493951714247042e-06,
+      "loss": 1.5088,
+      "step": 30380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.492577107514102e-06,
+      "loss": 1.3688,
+      "step": 30390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.491202503025319e-06,
+      "loss": 1.3687,
+      "step": 30400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4898279011962755e-06,
+      "loss": 1.4982,
+      "step": 30410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4884533024425546e-06,
+      "loss": 1.3565,
+      "step": 30420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4870787071797372e-06,
+      "loss": 1.5096,
+      "step": 30430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4857041158234032e-06,
+      "loss": 1.4072,
+      "step": 30440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4843295287891328e-06,
+      "loss": 1.4506,
+      "step": 30450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.482954946492503e-06,
+      "loss": 1.4296,
+      "step": 30460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4815803693490906e-06,
+      "loss": 1.3517,
+      "step": 30470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4802057977744703e-06,
+      "loss": 1.4976,
+      "step": 30480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4788312321842145e-06,
+      "loss": 1.4252,
+      "step": 30490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.477456672993895e-06,
+      "loss": 1.3511,
+      "step": 30500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.476082120619081e-06,
+      "loss": 1.434,
+      "step": 30510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4747075754753395e-06,
+      "loss": 1.4357,
+      "step": 30520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4733330379782366e-06,
+      "loss": 1.4881,
+      "step": 30530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.471958508543334e-06,
+      "loss": 1.3919,
+      "step": 30540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4705839875861927e-06,
+      "loss": 1.2248,
+      "step": 30550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.469209475522371e-06,
+      "loss": 1.5497,
+      "step": 30560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.467834972767424e-06,
+      "loss": 1.3933,
+      "step": 30570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4664604797369016e-06,
+      "loss": 1.4275,
+      "step": 30580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4650859968463557e-06,
+      "loss": 1.3678,
+      "step": 30590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.463711524511332e-06,
+      "loss": 1.4689,
+      "step": 30600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4623370631473727e-06,
+      "loss": 1.5018,
+      "step": 30610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.460962613170019e-06,
+      "loss": 1.4869,
+      "step": 30620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4595881749948065e-06,
+      "loss": 1.4109,
+      "step": 30630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4582137490372676e-06,
+      "loss": 1.4135,
+      "step": 30640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4568393357129315e-06,
+      "loss": 1.3796,
+      "step": 30650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.455464935437325e-06,
+      "loss": 1.5389,
+      "step": 30660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.454090548625967e-06,
+      "loss": 1.3987,
+      "step": 30670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4527161756943766e-06,
+      "loss": 1.4796,
+      "step": 30680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.451341817058065e-06,
+      "loss": 1.4026,
+      "step": 30690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4499674731325428e-06,
+      "loss": 1.3885,
+      "step": 30700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4485931443333132e-06,
+      "loss": 1.485,
+      "step": 30710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4472188310758766e-06,
+      "loss": 1.3492,
+      "step": 30720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4458445337757276e-06,
+      "loss": 1.4826,
+      "step": 30730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.444470252848356e-06,
+      "loss": 1.4302,
+      "step": 30740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.443095988709248e-06,
+      "loss": 1.4222,
+      "step": 30750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.441721741773883e-06,
+      "loss": 1.5156,
+      "step": 30760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.440347512457736e-06,
+      "loss": 1.4366,
+      "step": 30770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4389733011762768e-06,
+      "loss": 1.4755,
+      "step": 30780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.437599108344969e-06,
+      "loss": 1.4623,
+      "step": 30790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4362249343792722e-06,
+      "loss": 1.3878,
+      "step": 30800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4348507796946386e-06,
+      "loss": 1.4339,
+      "step": 30810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4334766447065157e-06,
+      "loss": 1.4075,
+      "step": 30820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.432102529830344e-06,
+      "loss": 1.5356,
+      "step": 30830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.43072843548156e-06,
+      "loss": 1.3639,
+      "step": 30840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4293543620755907e-06,
+      "loss": 1.4807,
+      "step": 30850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.427980310027859e-06,
+      "loss": 1.4382,
+      "step": 30860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4266062797537816e-06,
+      "loss": 1.4753,
+      "step": 30870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4252322716687676e-06,
+      "loss": 1.4268,
+      "step": 30880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.4238582861882195e-06,
+      "loss": 1.3833,
+      "step": 30890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.422484323727533e-06,
+      "loss": 1.5231,
+      "step": 30900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.421110384702098e-06,
+      "loss": 1.5493,
+      "step": 30910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2.419736469527296e-06,
+      "loss": 1.3977,
+      "step": 30920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4183625786185018e-06,
+      "loss": 1.4952,
+      "step": 30930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4169887123910823e-06,
+      "loss": 1.3561,
+      "step": 30940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.415614871260397e-06,
+      "loss": 1.3983,
+      "step": 30950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.414241055641799e-06,
+      "loss": 1.5402,
+      "step": 30960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4128672659506322e-06,
+      "loss": 1.3787,
+      "step": 30970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4114935026022337e-06,
+      "loss": 1.4655,
+      "step": 30980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.410119766011932e-06,
+      "loss": 1.368,
+      "step": 30990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4087460565950475e-06,
+      "loss": 1.3905,
+      "step": 31000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.407372374766894e-06,
+      "loss": 1.4492,
+      "step": 31010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.405998720942775e-06,
+      "loss": 1.4368,
+      "step": 31020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4046250955379852e-06,
+      "loss": 1.4775,
+      "step": 31030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4032514989678113e-06,
+      "loss": 1.4729,
+      "step": 31040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4018779316475333e-06,
+      "loss": 1.4161,
+      "step": 31050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.40050439399242e-06,
+      "loss": 1.4728,
+      "step": 31060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3991308864177317e-06,
+      "loss": 1.397,
+      "step": 31070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.39775740933872e-06,
+      "loss": 1.4479,
+      "step": 31080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.396383963170627e-06,
+      "loss": 1.4778,
+      "step": 31090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3950105483286857e-06,
+      "loss": 1.3388,
+      "step": 31100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3936371652281202e-06,
+      "loss": 1.3672,
+      "step": 31110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3922638142841433e-06,
+      "loss": 1.3829,
+      "step": 31120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3908904959119586e-06,
+      "loss": 1.4382,
+      "step": 31130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.389517210526761e-06,
+      "loss": 1.4472,
+      "step": 31140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3881439585437346e-06,
+      "loss": 1.4719,
+      "step": 31150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.386770740378054e-06,
+      "loss": 1.5017,
+      "step": 31160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.385397556444883e-06,
+      "loss": 1.452,
+      "step": 31170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.384024407159375e-06,
+      "loss": 1.478,
+      "step": 31180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.382651292936673e-06,
+      "loss": 1.4895,
+      "step": 31190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.38127821419191e-06,
+      "loss": 1.3628,
+      "step": 31200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.379905171340207e-06,
+      "loss": 1.3779,
+      "step": 31210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.378532164796675e-06,
+      "loss": 1.4261,
+      "step": 31220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3771591949764146e-06,
+      "loss": 1.5958,
+      "step": 31230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.375786262294514e-06,
+      "loss": 1.3661,
+      "step": 31240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3744133671660506e-06,
+      "loss": 1.3885,
+      "step": 31250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.373040510006092e-06,
+      "loss": 1.4651,
+      "step": 31260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.371667691229692e-06,
+      "loss": 1.383,
+      "step": 31270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3702949112518935e-06,
+      "loss": 1.4572,
+      "step": 31280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3689221704877292e-06,
+      "loss": 1.4917,
+      "step": 31290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.367549469352218e-06,
+      "loss": 1.4861,
+      "step": 31300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3661768082603665e-06,
+      "loss": 1.5055,
+      "step": 31310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.364804187627171e-06,
+      "loss": 1.373,
+      "step": 31320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.363431607867615e-06,
+      "loss": 1.4435,
+      "step": 31330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.362059069396669e-06,
+      "loss": 1.4561,
+      "step": 31340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3606865726292922e-06,
+      "loss": 1.3936,
+      "step": 31350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3593141179804295e-06,
+      "loss": 1.4346,
+      "step": 31360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.357941705865015e-06,
+      "loss": 1.3916,
+      "step": 31370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.356569336697969e-06,
+      "loss": 1.5179,
+      "step": 31380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3551970108941965e-06,
+      "loss": 1.3807,
+      "step": 31390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3538247288685935e-06,
+      "loss": 1.3369,
+      "step": 31400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.352452491036041e-06,
+      "loss": 1.4584,
+      "step": 31410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.351080297811406e-06,
+      "loss": 1.4028,
+      "step": 31420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.349708149609543e-06,
+      "loss": 1.4921,
+      "step": 31430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.348336046845292e-06,
+      "loss": 1.4368,
+      "step": 31440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3469639899334797e-06,
+      "loss": 1.4041,
+      "step": 31450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3455919792889186e-06,
+      "loss": 1.437,
+      "step": 31460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.34422001532641e-06,
+      "loss": 1.3518,
+      "step": 31470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.3428480984607356e-06,
+      "loss": 1.4276,
+      "step": 31480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.341476229106666e-06,
+      "loss": 1.3752,
+      "step": 31490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.340104407678959e-06,
+      "loss": 1.4534,
+      "step": 31500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.338732634592356e-06,
+      "loss": 1.499,
+      "step": 31510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3373609102615834e-06,
+      "loss": 1.4113,
+      "step": 31520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.335989235101354e-06,
+      "loss": 1.4057,
+      "step": 31530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3346176095263644e-06,
+      "loss": 1.4748,
+      "step": 31540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3332460339512984e-06,
+      "loss": 1.423,
+      "step": 31550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3318745087908225e-06,
+      "loss": 1.471,
+      "step": 31560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3305030344595884e-06,
+      "loss": 1.4691,
+      "step": 31570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3291316113722327e-06,
+      "loss": 1.478,
+      "step": 31580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.327760239943378e-06,
+      "loss": 1.4202,
+      "step": 31590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.326388920587628e-06,
+      "loss": 1.3963,
+      "step": 31600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.325017653719574e-06,
+      "loss": 1.5438,
+      "step": 31610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.32364643975379e-06,
+      "loss": 1.4479,
+      "step": 31620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3222752791048334e-06,
+      "loss": 1.5283,
+      "step": 31630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3209041721872464e-06,
+      "loss": 1.4142,
+      "step": 31640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3195331194155547e-06,
+      "loss": 1.4796,
+      "step": 31650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.318162121204267e-06,
+      "loss": 1.5657,
+      "step": 31660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3167911779678762e-06,
+      "loss": 1.3516,
+      "step": 31670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3154202901208593e-06,
+      "loss": 1.4257,
+      "step": 31680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3140494580776744e-06,
+      "loss": 1.3754,
+      "step": 31690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3126786822527642e-06,
+      "loss": 1.4144,
+      "step": 31700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3113079630605556e-06,
+      "loss": 1.4718,
+      "step": 31710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.309937300915456e-06,
+      "loss": 1.4536,
+      "step": 31720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3085666962318568e-06,
+      "loss": 1.4298,
+      "step": 31730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3071961494241323e-06,
+      "loss": 1.3669,
+      "step": 31740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3058256609066373e-06,
+      "loss": 1.4508,
+      "step": 31750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.304455231093712e-06,
+      "loss": 1.4746,
+      "step": 31760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.303084860399676e-06,
+      "loss": 1.379,
+      "step": 31770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.3017145492388333e-06,
+      "loss": 1.4383,
+      "step": 31780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.300344298025469e-06,
+      "loss": 1.4046,
+      "step": 31790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.298974107173849e-06,
+      "loss": 1.3864,
+      "step": 31800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.297603977098223e-06,
+      "loss": 1.4701,
+      "step": 31810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.296233908212821e-06,
+      "loss": 1.4239,
+      "step": 31820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.294863900931856e-06,
+      "loss": 1.4307,
+      "step": 31830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.293493955669519e-06,
+      "loss": 1.3445,
+      "step": 31840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.292124072839985e-06,
+      "loss": 1.4279,
+      "step": 31850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.29075425285741e-06,
+      "loss": 1.5007,
+      "step": 31860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.289384496135931e-06,
+      "loss": 1.3582,
+      "step": 31870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2880148030896648e-06,
+      "loss": 1.4617,
+      "step": 31880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2866451741327098e-06,
+      "loss": 1.4322,
+      "step": 31890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.285275609679145e-06,
+      "loss": 1.3768,
+      "step": 31900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2839061101430285e-06,
+      "loss": 1.488,
+      "step": 31910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2825366759384027e-06,
+      "loss": 1.4726,
+      "step": 31920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2811673074792847e-06,
+      "loss": 1.4494,
+      "step": 31930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.279798005179675e-06,
+      "loss": 1.4071,
+      "step": 31940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2784287694535545e-06,
+      "loss": 1.4105,
+      "step": 31950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2770596007148833e-06,
+      "loss": 1.4427,
+      "step": 31960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2756904993776e-06,
+      "loss": 1.4038,
+      "step": 31970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.274321465855625e-06,
+      "loss": 1.4991,
+      "step": 31980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2729525005628565e-06,
+      "loss": 1.4218,
+      "step": 31990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.271583603913173e-06,
+      "loss": 1.3988,
+      "step": 32000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.270214776320432e-06,
+      "loss": 1.5195,
+      "step": 32010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2688460181984688e-06,
+      "loss": 1.2928,
+      "step": 32020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2674773299611e-06,
+      "loss": 1.4596,
+      "step": 32030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.26610871202212e-06,
+      "loss": 1.4242,
+      "step": 32040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.264740164795301e-06,
+      "loss": 1.4287,
+      "step": 32050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.263371688694396e-06,
+      "loss": 1.5382,
+      "step": 32060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2620032841331347e-06,
+      "loss": 1.4099,
+      "step": 32070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.260634951525226e-06,
+      "loss": 1.4456,
+      "step": 32080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.2592666912843563e-06,
+      "loss": 1.4292,
+      "step": 32090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2.257898503824191e-06,
+      "loss": 1.4351,
+      "step": 32100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2565303895583734e-06,
+      "loss": 1.4437,
+      "step": 32110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2551623489005233e-06,
+      "loss": 1.383,
+      "step": 32120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.25379438226424e-06,
+      "loss": 1.4877,
+      "step": 32130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2524264900630997e-06,
+      "loss": 1.3392,
+      "step": 32140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2510586727106563e-06,
+      "loss": 1.4841,
+      "step": 32150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.24969093062044e-06,
+      "loss": 1.443,
+      "step": 32160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2483232642059605e-06,
+      "loss": 1.4084,
+      "step": 32170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.246955673880703e-06,
+      "loss": 1.4722,
+      "step": 32180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2455881600581297e-06,
+      "loss": 1.4133,
+      "step": 32190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.244220723151679e-06,
+      "loss": 1.3576,
+      "step": 32200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.242853363574768e-06,
+      "loss": 1.5421,
+      "step": 32210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.241486081740789e-06,
+      "loss": 1.3495,
+      "step": 32220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2401188780631116e-06,
+      "loss": 1.4698,
+      "step": 32230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2387517529550813e-06,
+      "loss": 1.3873,
+      "step": 32240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2373847068300196e-06,
+      "loss": 1.4581,
+      "step": 32250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.236017740101224e-06,
+      "loss": 1.513,
+      "step": 32260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.23465085318197e-06,
+      "loss": 1.451,
+      "step": 32270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2332840464855072e-06,
+      "loss": 1.484,
+      "step": 32280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2319173204250592e-06,
+      "loss": 1.3797,
+      "step": 32290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2305506754138283e-06,
+      "loss": 1.4098,
+      "step": 32300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2291841118649916e-06,
+      "loss": 1.4645,
+      "step": 32310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2278176301917004e-06,
+      "loss": 1.395,
+      "step": 32320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2264512308070827e-06,
+      "loss": 1.489,
+      "step": 32330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2250849141242405e-06,
+      "loss": 1.3748,
+      "step": 32340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.223718680556251e-06,
+      "loss": 1.3794,
+      "step": 32350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2223525305161675e-06,
+      "loss": 1.5384,
+      "step": 32360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2209864644170155e-06,
+      "loss": 1.4806,
+      "step": 32370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2196204826717986e-06,
+      "loss": 1.4455,
+      "step": 32380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2182545856934908e-06,
+      "loss": 1.4049,
+      "step": 32390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2168887738950438e-06,
+      "loss": 1.3521,
+      "step": 32400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2155230476893816e-06,
+      "loss": 1.4399,
+      "step": 32410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2141574074894043e-06,
+      "loss": 1.5103,
+      "step": 32420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.212791853707984e-06,
+      "loss": 1.4875,
+      "step": 32430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2114263867579673e-06,
+      "loss": 1.3913,
+      "step": 32440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.210061007052175e-06,
+      "loss": 1.4361,
+      "step": 32450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.208695715003402e-06,
+      "loss": 1.5022,
+      "step": 32460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.207330511024415e-06,
+      "loss": 1.4534,
+      "step": 32470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.205965395527955e-06,
+      "loss": 1.4684,
+      "step": 32480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.204600368926736e-06,
+      "loss": 1.4201,
+      "step": 32490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2032354316334458e-06,
+      "loss": 1.458,
+      "step": 32500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.201870584060744e-06,
+      "loss": 1.502,
+      "step": 32510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.2005058266212657e-06,
+      "loss": 1.3994,
+      "step": 32520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1991411597276154e-06,
+      "loss": 1.4527,
+      "step": 32530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1977765837923725e-06,
+      "loss": 1.4192,
+      "step": 32540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1964120992280876e-06,
+      "loss": 1.4117,
+      "step": 32550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.195047706447285e-06,
+      "loss": 1.4758,
+      "step": 32560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.193683405862459e-06,
+      "loss": 1.409,
+      "step": 32570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.192319197886078e-06,
+      "loss": 1.4963,
+      "step": 32580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1909550829305828e-06,
+      "loss": 1.3724,
+      "step": 32590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.189591061408384e-06,
+      "loss": 1.4201,
+      "step": 32600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.188227133731866e-06,
+      "loss": 1.5291,
+      "step": 32610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1868633003133828e-06,
+      "loss": 1.4026,
+      "step": 32620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1854995615652624e-06,
+      "loss": 1.4021,
+      "step": 32630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1841359178998024e-06,
+      "loss": 1.4283,
+      "step": 32640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.182772369729272e-06,
+      "loss": 1.4427,
+      "step": 32650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1814089174659114e-06,
+      "loss": 1.4525,
+      "step": 32660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1800455615219323e-06,
+      "loss": 1.3602,
+      "step": 32670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1786823023095164e-06,
+      "loss": 1.4459,
+      "step": 32680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2.1773191402408173e-06,
+      "loss": 1.3752,
+      "step": 32690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1759560757279583e-06,
+      "loss": 1.4652,
+      "step": 32700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1745931091830343e-06,
+      "loss": 1.4698,
+      "step": 32710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1732302410181084e-06,
+      "loss": 1.3657,
+      "step": 32720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.171867471645217e-06,
+      "loss": 1.4973,
+      "step": 32730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1705048014763656e-06,
+      "loss": 1.3938,
+      "step": 32740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1691422309235266e-06,
+      "loss": 1.3507,
+      "step": 32750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.167779760398646e-06,
+      "loss": 1.4944,
+      "step": 32760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.166417390313639e-06,
+      "loss": 1.4228,
+      "step": 32770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1650551210803897e-06,
+      "loss": 1.4227,
+      "step": 32780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1636929531107516e-06,
+      "loss": 1.4122,
+      "step": 32790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1623308868165474e-06,
+      "loss": 1.3944,
+      "step": 32800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.16096892260957e-06,
+      "loss": 1.4465,
+      "step": 32810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.159607060901581e-06,
+      "loss": 1.4093,
+      "step": 32820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1582453021043114e-06,
+      "loss": 1.4354,
+      "step": 32830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1568836466294596e-06,
+      "loss": 1.4938,
+      "step": 32840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1555220948886937e-06,
+      "loss": 1.4359,
+      "step": 32850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.154160647293651e-06,
+      "loss": 1.5218,
+      "step": 32860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.152799304255937e-06,
+      "loss": 1.4336,
+      "step": 32870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1514380661871253e-06,
+      "loss": 1.5571,
+      "step": 32880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.150076933498758e-06,
+      "loss": 1.2809,
+      "step": 32890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1487159066023448e-06,
+      "loss": 1.405,
+      "step": 32900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1473549859093645e-06,
+      "loss": 1.4394,
+      "step": 32910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1459941718312633e-06,
+      "loss": 1.4718,
+      "step": 32920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.144633464779454e-06,
+      "loss": 1.4237,
+      "step": 32930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1432728651653184e-06,
+      "loss": 1.3536,
+      "step": 32940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1419123734002058e-06,
+      "loss": 1.3509,
+      "step": 32950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.140551989895432e-06,
+      "loss": 1.4762,
+      "step": 32960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1391917150622805e-06,
+      "loss": 1.3482,
+      "step": 32970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1378315493120034e-06,
+      "loss": 1.469,
+      "step": 32980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1364714930558174e-06,
+      "loss": 1.4128,
+      "step": 32990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.135111546704908e-06,
+      "loss": 1.3859,
+      "step": 33000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.133751710670426e-06,
+      "loss": 1.4522,
+      "step": 33010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1323919853634895e-06,
+      "loss": 1.4112,
+      "step": 33020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.131032371195183e-06,
+      "loss": 1.5013,
+      "step": 33030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.129672868576558e-06,
+      "loss": 1.4225,
+      "step": 33040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1283134779186315e-06,
+      "loss": 1.4319,
+      "step": 33050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1269541996323877e-06,
+      "loss": 1.4674,
+      "step": 33060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1255950341287744e-06,
+      "loss": 1.4904,
+      "step": 33070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1242359818187085e-06,
+      "loss": 1.4257,
+      "step": 33080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.122877043113071e-06,
+      "loss": 1.4631,
+      "step": 33090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1215182184227103e-06,
+      "loss": 1.3397,
+      "step": 33100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1201595081584342e-06,
+      "loss": 1.4584,
+      "step": 33110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1188009127310244e-06,
+      "loss": 1.4722,
+      "step": 33120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.117442432551223e-06,
+      "loss": 1.4641,
+      "step": 33130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1160840680297374e-06,
+      "loss": 1.3916,
+      "step": 33140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1147258195772415e-06,
+      "loss": 1.4048,
+      "step": 33150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1133676876043734e-06,
+      "loss": 1.4769,
+      "step": 33160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.112009672521736e-06,
+      "loss": 1.3862,
+      "step": 33170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.110651774739896e-06,
+      "loss": 1.5772,
+      "step": 33180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1092939946693884e-06,
+      "loss": 1.3857,
+      "step": 33190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1079363327207067e-06,
+      "loss": 1.3836,
+      "step": 33200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.106578789304312e-06,
+      "loss": 1.4174,
+      "step": 33210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.1052213648306308e-06,
+      "loss": 1.4498,
+      "step": 33220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.103864059710051e-06,
+      "loss": 1.5622,
+      "step": 33230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.102506874352926e-06,
+      "loss": 1.4557,
+      "step": 33240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.101149809169573e-06,
+      "loss": 1.4277,
+      "step": 33250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.099792864570271e-06,
+      "loss": 1.4324,
+      "step": 33260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.0984360409652656e-06,
+      "loss": 1.462,
+      "step": 33270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2.0970793387647636e-06,
+      "loss": 1.444,
+      "step": 33280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.095722758378935e-06,
+      "loss": 1.364,
+      "step": 33290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0943663002179136e-06,
+      "loss": 1.4787,
+      "step": 33300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0930099646917975e-06,
+      "loss": 1.5185,
+      "step": 33310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0916537522106446e-06,
+      "loss": 1.4177,
+      "step": 33320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0902976631844798e-06,
+      "loss": 1.4541,
+      "step": 33330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.088941698023287e-06,
+      "loss": 1.4178,
+      "step": 33340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0875858571370143e-06,
+      "loss": 1.3648,
+      "step": 33350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0862301409355717e-06,
+      "loss": 1.4426,
+      "step": 33360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0848745498288332e-06,
+      "loss": 1.3638,
+      "step": 33370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0835190842266315e-06,
+      "loss": 1.4574,
+      "step": 33380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0821637445387646e-06,
+      "loss": 1.4103,
+      "step": 33390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0808085311749904e-06,
+      "loss": 1.4022,
+      "step": 33400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0794534445450304e-06,
+      "loss": 1.3727,
+      "step": 33410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0780984850585655e-06,
+      "loss": 1.3934,
+      "step": 33420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0767436531252417e-06,
+      "loss": 1.4137,
+      "step": 33430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0753889491546624e-06,
+      "loss": 1.4373,
+      "step": 33440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.074034373556395e-06,
+      "loss": 1.47,
+      "step": 33450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.072679926739967e-06,
+      "loss": 1.4206,
+      "step": 33460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.071325609114867e-06,
+      "loss": 1.3586,
+      "step": 33470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.069971421090545e-06,
+      "loss": 1.5196,
+      "step": 33480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0686173630764104e-06,
+      "loss": 1.3802,
+      "step": 33490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0672634354818365e-06,
+      "loss": 1.4311,
+      "step": 33500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0659096387161537e-06,
+      "loss": 1.4679,
+      "step": 33510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.064555973188655e-06,
+      "loss": 1.3996,
+      "step": 33520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0632024393085915e-06,
+      "loss": 1.5182,
+      "step": 33530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.061849037485178e-06,
+      "loss": 1.4123,
+      "step": 33540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.060495768127588e-06,
+      "loss": 1.435,
+      "step": 33550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0591426316449513e-06,
+      "loss": 1.4648,
+      "step": 33560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0577896284463615e-06,
+      "loss": 1.4099,
+      "step": 33570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0564367589408724e-06,
+      "loss": 1.4413,
+      "step": 33580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0550840235374947e-06,
+      "loss": 1.4474,
+      "step": 33590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0537314226452004e-06,
+      "loss": 1.4153,
+      "step": 33600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0523789566729197e-06,
+      "loss": 1.4959,
+      "step": 33610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.051026626029543e-06,
+      "loss": 1.4016,
+      "step": 33620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0496744311239185e-06,
+      "loss": 1.4765,
+      "step": 33630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0483223723648567e-06,
+      "loss": 1.392,
+      "step": 33640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.046970450161121e-06,
+      "loss": 1.4217,
+      "step": 33650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0456186649214388e-06,
+      "loss": 1.4808,
+      "step": 33660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0442670170544928e-06,
+      "loss": 1.4454,
+      "step": 33670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.042915506968928e-06,
+      "loss": 1.482,
+      "step": 33680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0415641350733434e-06,
+      "loss": 1.305,
+      "step": 33690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0402129017762986e-06,
+      "loss": 1.3741,
+      "step": 33700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0388618074863114e-06,
+      "loss": 1.4845,
+      "step": 33710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0375108526118563e-06,
+      "loss": 1.3882,
+      "step": 33720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0361600375613676e-06,
+      "loss": 1.4183,
+      "step": 33730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.034809362743235e-06,
+      "loss": 1.3706,
+      "step": 33740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.033458828565807e-06,
+      "loss": 1.4208,
+      "step": 33750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.032108435437389e-06,
+      "loss": 1.4814,
+      "step": 33760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0307581837662456e-06,
+      "loss": 1.4756,
+      "step": 33770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0294080739605955e-06,
+      "loss": 1.5113,
+      "step": 33780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.028058106428618e-06,
+      "loss": 1.383,
+      "step": 33790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.026708281578447e-06,
+      "loss": 1.3657,
+      "step": 33800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0253585998181745e-06,
+      "loss": 1.5111,
+      "step": 33810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.024009061555848e-06,
+      "loss": 1.3536,
+      "step": 33820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.022659667199473e-06,
+      "loss": 1.5089,
+      "step": 33830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0213104171570094e-06,
+      "loss": 1.4375,
+      "step": 33840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.019961311836376e-06,
+      "loss": 1.505,
+      "step": 33850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2.0186123516454464e-06,
+      "loss": 1.4267,
+      "step": 33860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0172635369920505e-06,
+      "loss": 1.3728,
+      "step": 33870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.015914868283974e-06,
+      "loss": 1.434,
+      "step": 33880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.01456634592896e-06,
+      "loss": 1.3303,
+      "step": 33890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0132179703347053e-06,
+      "loss": 1.4779,
+      "step": 33900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0118697419088646e-06,
+      "loss": 1.5215,
+      "step": 33910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.010521661059044e-06,
+      "loss": 1.2791,
+      "step": 33920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.009173728192809e-06,
+      "loss": 1.5056,
+      "step": 33930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.00782594371768e-06,
+      "loss": 1.4237,
+      "step": 33940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.00647830804113e-06,
+      "loss": 1.4357,
+      "step": 33950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.00513082157059e-06,
+      "loss": 1.5177,
+      "step": 33960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0037834847134436e-06,
+      "loss": 1.3414,
+      "step": 33970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0024362978770297e-06,
+      "loss": 1.4636,
+      "step": 33980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2.0010892614686427e-06,
+      "loss": 1.3925,
+      "step": 33990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.999742375895533e-06,
+      "loss": 1.3343,
+      "step": 34000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9983956415649002e-06,
+      "loss": 1.5047,
+      "step": 34010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.997049058883902e-06,
+      "loss": 1.3981,
+      "step": 34020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9957026282596508e-06,
+      "loss": 1.4867,
+      "step": 34030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9943563500992115e-06,
+      "loss": 1.367,
+      "step": 34040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9930102248096035e-06,
+      "loss": 1.3993,
+      "step": 34050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.991664252797799e-06,
+      "loss": 1.4143,
+      "step": 34060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.990318434470725e-06,
+      "loss": 1.4167,
+      "step": 34070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9889727702352618e-06,
+      "loss": 1.5206,
+      "step": 34080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9876272604982434e-06,
+      "loss": 1.4211,
+      "step": 34090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9862819056664555e-06,
+      "loss": 1.4434,
+      "step": 34100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.984936706146639e-06,
+      "loss": 1.5035,
+      "step": 34110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9835916623454865e-06,
+      "loss": 1.5218,
+      "step": 34120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.982246774669644e-06,
+      "loss": 1.4756,
+      "step": 34130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.980902043525711e-06,
+      "loss": 1.3978,
+      "step": 34140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.979557469320238e-06,
+      "loss": 1.4742,
+      "step": 34150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.97821305245973e-06,
+      "loss": 1.4694,
+      "step": 34160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9768687933506423e-06,
+      "loss": 1.3705,
+      "step": 34170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9755246923993855e-06,
+      "loss": 1.5277,
+      "step": 34180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9741807500123183e-06,
+      "loss": 1.3213,
+      "step": 34190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9728369665957547e-06,
+      "loss": 1.3633,
+      "step": 34200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.97149334255596e-06,
+      "loss": 1.4744,
+      "step": 34210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9701498782991507e-06,
+      "loss": 1.3966,
+      "step": 34220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9688065742314946e-06,
+      "loss": 1.4895,
+      "step": 34230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.967463430759112e-06,
+      "loss": 1.327,
+      "step": 34240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9661204482880757e-06,
+      "loss": 1.3928,
+      "step": 34250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9647776272244075e-06,
+      "loss": 1.4803,
+      "step": 34260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.963434967974081e-06,
+      "loss": 1.3421,
+      "step": 34270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.962092470943022e-06,
+      "loss": 1.4651,
+      "step": 34280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9607501365371056e-06,
+      "loss": 1.392,
+      "step": 34290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9594079651621596e-06,
+      "loss": 1.3792,
+      "step": 34300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9580659572239606e-06,
+      "loss": 1.5174,
+      "step": 34310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9567241131282378e-06,
+      "loss": 1.3929,
+      "step": 34320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.955382433280669e-06,
+      "loss": 1.4928,
+      "step": 34330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9540409180868834e-06,
+      "loss": 1.4201,
+      "step": 34340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.952699567952461e-06,
+      "loss": 1.3739,
+      "step": 34350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9513583832829314e-06,
+      "loss": 1.548,
+      "step": 34360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9500173644837713e-06,
+      "loss": 1.3618,
+      "step": 34370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.948676511960412e-06,
+      "loss": 1.5273,
+      "step": 34380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.947335826118231e-06,
+      "loss": 1.4322,
+      "step": 34390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9459953073625586e-06,
+      "loss": 1.4315,
+      "step": 34400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.944654956098671e-06,
+      "loss": 1.4644,
+      "step": 34410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.9433147727317966e-06,
+      "loss": 1.49,
+      "step": 34420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.941974757667111e-06,
+      "loss": 1.4588,
+      "step": 34430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.94063491130974e-06,
+      "loss": 1.4904,
+      "step": 34440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.93929523406476e-06,
+      "loss": 1.4495,
+      "step": 34450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9379557263371924e-06,
+      "loss": 1.5531,
+      "step": 34460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.93661638853201e-06,
+      "loss": 1.4149,
+      "step": 34470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9352772210541332e-06,
+      "loss": 1.4986,
+      "step": 34480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.933938224308433e-06,
+      "loss": 1.368,
+      "step": 34490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.932599398699726e-06,
+      "loss": 1.4758,
+      "step": 34500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.931260744632779e-06,
+      "loss": 1.4587,
+      "step": 34510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9299222625123054e-06,
+      "loss": 1.2581,
+      "step": 34520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.928583952742968e-06,
+      "loss": 1.519,
+      "step": 34530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9272458157293765e-06,
+      "loss": 1.3797,
+      "step": 34540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9259078518760894e-06,
+      "loss": 1.3866,
+      "step": 34550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.924570061587611e-06,
+      "loss": 1.4452,
+      "step": 34560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9232324452683954e-06,
+      "loss": 1.4537,
+      "step": 34570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.921895003322842e-06,
+      "loss": 1.4491,
+      "step": 34580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9205577361553e-06,
+      "loss": 1.4194,
+      "step": 34590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9192206441700636e-06,
+      "loss": 1.3636,
+      "step": 34600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.917883727771374e-06,
+      "loss": 1.3768,
+      "step": 34610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9165469873634213e-06,
+      "loss": 1.3326,
+      "step": 34620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9152104233503406e-06,
+      "loss": 1.4377,
+      "step": 34630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.913874036136214e-06,
+      "loss": 1.367,
+      "step": 34640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.912537826125071e-06,
+      "loss": 1.4725,
+      "step": 34650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9112017937208852e-06,
+      "loss": 1.4534,
+      "step": 34660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.90986593932758e-06,
+      "loss": 1.3589,
+      "step": 34670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9085302633490227e-06,
+      "loss": 1.4911,
+      "step": 34680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9071947661890264e-06,
+      "loss": 1.3716,
+      "step": 34690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9058594482513518e-06,
+      "loss": 1.3958,
+      "step": 34700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.904524309939704e-06,
+      "loss": 1.4459,
+      "step": 34710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9031893516577343e-06,
+      "loss": 1.5133,
+      "step": 34720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9018545738090406e-06,
+      "loss": 1.4632,
+      "step": 34730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.9005199767971633e-06,
+      "loss": 1.3937,
+      "step": 34740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.899185561025591e-06,
+      "loss": 1.4128,
+      "step": 34750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8978513268977564e-06,
+      "loss": 1.5129,
+      "step": 34760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8965172748170373e-06,
+      "loss": 1.4674,
+      "step": 34770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8951834051867573e-06,
+      "loss": 1.4214,
+      "step": 34780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8938497184101826e-06,
+      "loss": 1.385,
+      "step": 34790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8925162148905275e-06,
+      "loss": 1.3783,
+      "step": 34800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8911828950309475e-06,
+      "loss": 1.4161,
+      "step": 34810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8898497592345463e-06,
+      "loss": 1.4216,
+      "step": 34820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8885168079043664e-06,
+      "loss": 1.4387,
+      "step": 34830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8871840414434e-06,
+      "loss": 1.3756,
+      "step": 34840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8858514602545814e-06,
+      "loss": 1.4281,
+      "step": 34850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.884519064740788e-06,
+      "loss": 1.4779,
+      "step": 34860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8831868553048424e-06,
+      "loss": 1.3288,
+      "step": 34870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8818548323495101e-06,
+      "loss": 1.5253,
+      "step": 34880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8805229962775003e-06,
+      "loss": 1.3434,
+      "step": 34890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.879191347491466e-06,
+      "loss": 1.4042,
+      "step": 34900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.877859886394005e-06,
+      "loss": 1.5123,
+      "step": 34910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8765286133876549e-06,
+      "loss": 1.3658,
+      "step": 34920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.875197528874898e-06,
+      "loss": 1.4426,
+      "step": 34930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8738666332581618e-06,
+      "loss": 1.465,
+      "step": 34940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8725359269398141e-06,
+      "loss": 1.3992,
+      "step": 34950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8712054103221661e-06,
+      "loss": 1.4953,
+      "step": 34960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8698750838074719e-06,
+      "loss": 1.4662,
+      "step": 34970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8685449477979282e-06,
+      "loss": 1.4649,
+      "step": 34980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8672150026956738e-06,
+      "loss": 1.3979,
+      "step": 34990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.86588524890279e-06,
+      "loss": 1.3452,
+      "step": 35000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8645556868213e-06,
+      "loss": 1.4209,
+      "step": 35010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8632263168531692e-06,
+      "loss": 1.3704,
+      "step": 35020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8618971394003048e-06,
+      "loss": 1.5136,
+      "step": 35030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.860568154864556e-06,
+      "loss": 1.362,
+      "step": 35040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8592393636477141e-06,
+      "loss": 1.4624,
+      "step": 35050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.857910766151511e-06,
+      "loss": 1.5073,
+      "step": 35060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.856582362777621e-06,
+      "loss": 1.3991,
+      "step": 35070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8552541539276587e-06,
+      "loss": 1.5102,
+      "step": 35080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8539261400031811e-06,
+      "loss": 1.3988,
+      "step": 35090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.852598321405684e-06,
+      "loss": 1.4172,
+      "step": 35100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8512706985366072e-06,
+      "loss": 1.4562,
+      "step": 35110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8499432717973293e-06,
+      "loss": 1.3483,
+      "step": 35120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8486160415891699e-06,
+      "loss": 1.4457,
+      "step": 35130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8472890083133893e-06,
+      "loss": 1.4097,
+      "step": 35140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8459621723711893e-06,
+      "loss": 1.4527,
+      "step": 35150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8446355341637107e-06,
+      "loss": 1.5158,
+      "step": 35160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8433090940920345e-06,
+      "loss": 1.4149,
+      "step": 35170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8419828525571837e-06,
+      "loss": 1.4943,
+      "step": 35180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8406568099601177e-06,
+      "loss": 1.3737,
+      "step": 35190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8393309667017388e-06,
+      "loss": 1.3377,
+      "step": 35200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8380053231828888e-06,
+      "loss": 1.5476,
+      "step": 35210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8366798798043472e-06,
+      "loss": 1.4709,
+      "step": 35220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8353546369668352e-06,
+      "loss": 1.481,
+      "step": 35230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8340295950710119e-06,
+      "loss": 1.4264,
+      "step": 35240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8327047545174759e-06,
+      "loss": 1.4129,
+      "step": 35250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8313801157067657e-06,
+      "loss": 1.4616,
+      "step": 35260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8300556790393592e-06,
+      "loss": 1.4089,
+      "step": 35270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.82873144491567e-06,
+      "loss": 1.5195,
+      "step": 35280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8274074137360534e-06,
+      "loss": 1.4009,
+      "step": 35290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8260835859008037e-06,
+      "loss": 1.4699,
+      "step": 35300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8247599618101519e-06,
+      "loss": 1.4658,
+      "step": 35310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8234365418642685e-06,
+      "loss": 1.3364,
+      "step": 35320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8221133264632617e-06,
+      "loss": 1.469,
+      "step": 35330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8207903160071788e-06,
+      "loss": 1.4196,
+      "step": 35340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.819467510896003e-06,
+      "loss": 1.3711,
+      "step": 35350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8181449115296596e-06,
+      "loss": 1.4276,
+      "step": 35360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8168225183080058e-06,
+      "loss": 1.3564,
+      "step": 35370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8155003316308412e-06,
+      "loss": 1.4426,
+      "step": 35380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.814178351897901e-06,
+      "loss": 1.2582,
+      "step": 35390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8128565795088588e-06,
+      "loss": 1.3924,
+      "step": 35400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8115350148633249e-06,
+      "loss": 1.443,
+      "step": 35410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8102136583608465e-06,
+      "loss": 1.4572,
+      "step": 35420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8088925104009086e-06,
+      "loss": 1.5203,
+      "step": 35430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8075715713829323e-06,
+      "loss": 1.3975,
+      "step": 35440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.806250841706277e-06,
+      "loss": 1.3677,
+      "step": 35450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8049303217702368e-06,
+      "loss": 1.4581,
+      "step": 35460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8036100119740436e-06,
+      "loss": 1.5135,
+      "step": 35470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8022899127168655e-06,
+      "loss": 1.4805,
+      "step": 35480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8009700243978074e-06,
+      "loss": 1.3893,
+      "step": 35490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7996503474159094e-06,
+      "loss": 1.3856,
+      "step": 35500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.798330882170149e-06,
+      "loss": 1.4726,
+      "step": 35510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7970116290594391e-06,
+      "loss": 1.4402,
+      "step": 35520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7956925884826288e-06,
+      "loss": 1.4622,
+      "step": 35530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7943737608385017e-06,
+      "loss": 1.4528,
+      "step": 35540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7930551465257779e-06,
+      "loss": 1.3295,
+      "step": 35550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7917367459431135e-06,
+      "loss": 1.3803,
+      "step": 35560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7904185594890988e-06,
+      "loss": 1.3912,
+      "step": 35570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7891005875622607e-06,
+      "loss": 1.495,
+      "step": 35580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7877828305610603e-06,
+      "loss": 1.3634,
+      "step": 35590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7864652888838935e-06,
+      "loss": 1.2809,
+      "step": 35600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7851479629290924e-06,
+      "loss": 1.4338,
+      "step": 35610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7838308530949231e-06,
+      "loss": 1.4256,
+      "step": 35620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.7825139597795871e-06,
+      "loss": 1.4067,
+      "step": 35630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7811972833812167e-06,
+      "loss": 1.4022,
+      "step": 35640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7798808242978843e-06,
+      "loss": 1.4268,
+      "step": 35650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7785645829275926e-06,
+      "loss": 1.5425,
+      "step": 35660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.777248559668281e-06,
+      "loss": 1.4081,
+      "step": 35670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7759327549178208e-06,
+      "loss": 1.5075,
+      "step": 35680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7746171690740182e-06,
+      "loss": 1.3635,
+      "step": 35690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7733018025346127e-06,
+      "loss": 1.427,
+      "step": 35700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7719866556972798e-06,
+      "loss": 1.4514,
+      "step": 35710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7706717289596263e-06,
+      "loss": 1.3638,
+      "step": 35720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7693570227191908e-06,
+      "loss": 1.4394,
+      "step": 35730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7680425373734489e-06,
+      "loss": 1.367,
+      "step": 35740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7667282733198082e-06,
+      "loss": 1.377,
+      "step": 35750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7654142309556088e-06,
+      "loss": 1.4158,
+      "step": 35760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7641004106781233e-06,
+      "loss": 1.389,
+      "step": 35770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7627868128845594e-06,
+      "loss": 1.5165,
+      "step": 35780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7614734379720547e-06,
+      "loss": 1.4395,
+      "step": 35790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7601602863376816e-06,
+      "loss": 1.4306,
+      "step": 35800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7588473583784438e-06,
+      "loss": 1.5211,
+      "step": 35810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7575346544912775e-06,
+      "loss": 1.3546,
+      "step": 35820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7562221750730518e-06,
+      "loss": 1.4141,
+      "step": 35830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7549099205205673e-06,
+      "loss": 1.3519,
+      "step": 35840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.753597891230556e-06,
+      "loss": 1.417,
+      "step": 35850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7522860875996845e-06,
+      "loss": 1.4928,
+      "step": 35860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.750974510024548e-06,
+      "loss": 1.4054,
+      "step": 35870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7496631589016754e-06,
+      "loss": 1.4379,
+      "step": 35880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7483520346275258e-06,
+      "loss": 1.4182,
+      "step": 35890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7470411375984915e-06,
+      "loss": 1.4368,
+      "step": 35900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.745730468210893e-06,
+      "loss": 1.4549,
+      "step": 35910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.744420026860985e-06,
+      "loss": 1.465,
+      "step": 35920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7431098139449523e-06,
+      "loss": 1.5616,
+      "step": 35930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7417998298589104e-06,
+      "loss": 1.3657,
+      "step": 35940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7404900749989046e-06,
+      "loss": 1.3312,
+      "step": 35950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7391805497609137e-06,
+      "loss": 1.5074,
+      "step": 35960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.737871254540845e-06,
+      "loss": 1.4047,
+      "step": 35970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7365621897345364e-06,
+      "loss": 1.4197,
+      "step": 35980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7352533557377566e-06,
+      "loss": 1.4461,
+      "step": 35990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7339447529462033e-06,
+      "loss": 1.3874,
+      "step": 36000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7326363817555065e-06,
+      "loss": 1.4242,
+      "step": 36010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7313282425612243e-06,
+      "loss": 1.3822,
+      "step": 36020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7300203357588457e-06,
+      "loss": 1.4897,
+      "step": 36030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7287126617437889e-06,
+      "loss": 1.4042,
+      "step": 36040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7274052209114022e-06,
+      "loss": 1.4134,
+      "step": 36050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7260980136569621e-06,
+      "loss": 1.4998,
+      "step": 36060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7247910403756774e-06,
+      "loss": 1.3925,
+      "step": 36070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.723484301462684e-06,
+      "loss": 1.4893,
+      "step": 36080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.722177797313045e-06,
+      "loss": 1.5151,
+      "step": 36090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7208715283217562e-06,
+      "loss": 1.3998,
+      "step": 36100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7195654948837408e-06,
+      "loss": 1.4518,
+      "step": 36110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7182596973938512e-06,
+      "loss": 1.4327,
+      "step": 36120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7169541362468674e-06,
+      "loss": 1.5188,
+      "step": 36130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7156488118374992e-06,
+      "loss": 1.3945,
+      "step": 36140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.714343724560384e-06,
+      "loss": 1.3643,
+      "step": 36150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7130388748100868e-06,
+      "loss": 1.4531,
+      "step": 36160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7117342629811045e-06,
+      "loss": 1.4009,
+      "step": 36170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7104298894678564e-06,
+      "loss": 1.5196,
+      "step": 36180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.709125754664694e-06,
+      "loss": 1.4813,
+      "step": 36190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7078218589658946e-06,
+      "loss": 1.4357,
+      "step": 36200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7065182027656647e-06,
+      "loss": 1.5163,
+      "step": 36210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.7052147864581375e-06,
+      "loss": 1.4725,
+      "step": 36220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7039116104373734e-06,
+      "loss": 1.468,
+      "step": 36230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7026086750973612e-06,
+      "loss": 1.3194,
+      "step": 36240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7013059808320153e-06,
+      "loss": 1.3241,
+      "step": 36250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.7000035280351795e-06,
+      "loss": 1.4535,
+      "step": 36260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6987013171006222e-06,
+      "loss": 1.3251,
+      "step": 36270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6973993484220403e-06,
+      "loss": 1.515,
+      "step": 36280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6960976223930561e-06,
+      "loss": 1.3751,
+      "step": 36290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.69479613940722e-06,
+      "loss": 1.3752,
+      "step": 36300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6934948998580084e-06,
+      "loss": 1.4666,
+      "step": 36310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.692193904138824e-06,
+      "loss": 1.3666,
+      "step": 36320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6908931526429955e-06,
+      "loss": 1.5352,
+      "step": 36330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6895926457637785e-06,
+      "loss": 1.4041,
+      "step": 36340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6882923838943541e-06,
+      "loss": 1.4012,
+      "step": 36350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6869923674278288e-06,
+      "loss": 1.5645,
+      "step": 36360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6856925967572356e-06,
+      "loss": 1.3043,
+      "step": 36370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6843930722755337e-06,
+      "loss": 1.4509,
+      "step": 36380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6830937943756066e-06,
+      "loss": 1.3657,
+      "step": 36390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6817947634502646e-06,
+      "loss": 1.3648,
+      "step": 36400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6804959798922416e-06,
+      "loss": 1.5102,
+      "step": 36410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6791974440941992e-06,
+      "loss": 1.4382,
+      "step": 36420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6778991564487218e-06,
+      "loss": 1.5515,
+      "step": 36430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.676601117348321e-06,
+      "loss": 1.4303,
+      "step": 36440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6753033271854288e-06,
+      "loss": 1.4299,
+      "step": 36450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6740057863524079e-06,
+      "loss": 1.4826,
+      "step": 36460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6727084952415413e-06,
+      "loss": 1.4385,
+      "step": 36470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6714114542450388e-06,
+      "loss": 1.5177,
+      "step": 36480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6701146637550334e-06,
+      "loss": 1.3988,
+      "step": 36490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6688181241635825e-06,
+      "loss": 1.3855,
+      "step": 36500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6675218358626677e-06,
+      "loss": 1.4813,
+      "step": 36510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6662257992441957e-06,
+      "loss": 1.3504,
+      "step": 36520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6649300146999967e-06,
+      "loss": 1.544,
+      "step": 36530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6636344826218221e-06,
+      "loss": 1.4706,
+      "step": 36540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6623392034013494e-06,
+      "loss": 1.4062,
+      "step": 36550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6610441774301806e-06,
+      "loss": 1.5097,
+      "step": 36560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.659749405099839e-06,
+      "loss": 1.2801,
+      "step": 36570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6584548868017725e-06,
+      "loss": 1.5319,
+      "step": 36580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6571606229273514e-06,
+      "loss": 1.4019,
+      "step": 36590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6558666138678698e-06,
+      "loss": 1.3919,
+      "step": 36600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6545728600145438e-06,
+      "loss": 1.4384,
+      "step": 36610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6532793617585143e-06,
+      "loss": 1.4773,
+      "step": 36620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.651986119490842e-06,
+      "loss": 1.461,
+      "step": 36630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6506931336025122e-06,
+      "loss": 1.3683,
+      "step": 36640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6494004044844326e-06,
+      "loss": 1.4407,
+      "step": 36650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6481079325274319e-06,
+      "loss": 1.5385,
+      "step": 36660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6468157181222638e-06,
+      "loss": 1.4686,
+      "step": 36670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6455237616596011e-06,
+      "loss": 1.4961,
+      "step": 36680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.644232063530041e-06,
+      "loss": 1.3646,
+      "step": 36690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6429406241241009e-06,
+      "loss": 1.3822,
+      "step": 36700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6416494438322209e-06,
+      "loss": 1.5129,
+      "step": 36710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6403585230447617e-06,
+      "loss": 1.4194,
+      "step": 36720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6390678621520073e-06,
+      "loss": 1.4786,
+      "step": 36730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6377774615441614e-06,
+      "loss": 1.3733,
+      "step": 36740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6364873216113502e-06,
+      "loss": 1.4173,
+      "step": 36750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6351974427436198e-06,
+      "loss": 1.486,
+      "step": 36760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6339078253309398e-06,
+      "loss": 1.4148,
+      "step": 36770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6326184697631983e-06,
+      "loss": 1.4955,
+      "step": 36780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.631329376430205e-06,
+      "loss": 1.4297,
+      "step": 36790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.6300405457216912e-06,
+      "loss": 1.3736,
+      "step": 36800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.628751978027307e-06,
+      "loss": 1.5062,
+      "step": 36810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6274636737366245e-06,
+      "loss": 1.3052,
+      "step": 36820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6261756332391354e-06,
+      "loss": 1.4641,
+      "step": 36830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6248878569242522e-06,
+      "loss": 1.3304,
+      "step": 36840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6236003451813073e-06,
+      "loss": 1.3874,
+      "step": 36850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6223130983995524e-06,
+      "loss": 1.4704,
+      "step": 36860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6210261169681609e-06,
+      "loss": 1.4191,
+      "step": 36870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6197394012762245e-06,
+      "loss": 1.5331,
+      "step": 36880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6184529517127545e-06,
+      "loss": 1.3931,
+      "step": 36890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6171667686666828e-06,
+      "loss": 1.3889,
+      "step": 36900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6158808525268594e-06,
+      "loss": 1.3746,
+      "step": 36910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6145952036820538e-06,
+      "loss": 1.4109,
+      "step": 36920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6133098225209565e-06,
+      "loss": 1.5109,
+      "step": 36930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.612024709432175e-06,
+      "loss": 1.3502,
+      "step": 36940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6107398648042363e-06,
+      "loss": 1.3783,
+      "step": 36950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6094552890255869e-06,
+      "loss": 1.4271,
+      "step": 36960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6081709824845908e-06,
+      "loss": 1.5031,
+      "step": 36970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6068869455695325e-06,
+      "loss": 1.4551,
+      "step": 36980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6056031786686137e-06,
+      "loss": 1.4116,
+      "step": 36990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6043196821699536e-06,
+      "loss": 1.3573,
+      "step": 37000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6030364564615909e-06,
+      "loss": 1.5306,
+      "step": 37010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.6017535019314822e-06,
+      "loss": 1.3593,
+      "step": 37020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.600470818967503e-06,
+      "loss": 1.4997,
+      "step": 37030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5991884079574448e-06,
+      "loss": 1.4387,
+      "step": 37040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5979062692890184e-06,
+      "loss": 1.4464,
+      "step": 37050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5966244033498513e-06,
+      "loss": 1.4333,
+      "step": 37060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5953428105274888e-06,
+      "loss": 1.4117,
+      "step": 37070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5940614912093957e-06,
+      "loss": 1.435,
+      "step": 37080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.59278044578295e-06,
+      "loss": 1.4173,
+      "step": 37090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5914996746354494e-06,
+      "loss": 1.4807,
+      "step": 37100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5902191781541087e-06,
+      "loss": 1.436,
+      "step": 37110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.58893895672606e-06,
+      "loss": 1.4325,
+      "step": 37120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5876590107383508e-06,
+      "loss": 1.4123,
+      "step": 37130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5863793405779464e-06,
+      "loss": 1.4092,
+      "step": 37140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5850999466317287e-06,
+      "loss": 1.4408,
+      "step": 37150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5838208292864954e-06,
+      "loss": 1.4735,
+      "step": 37160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582541988928962e-06,
+      "loss": 1.4548,
+      "step": 37170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812634259457576e-06,
+      "loss": 1.4392,
+      "step": 37180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.57998514072343e-06,
+      "loss": 1.3967,
+      "step": 37190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5787071336484422e-06,
+      "loss": 1.4707,
+      "step": 37200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5774294051071731e-06,
+      "loss": 1.5362,
+      "step": 37210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576151955485916e-06,
+      "loss": 1.4363,
+      "step": 37220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5748747851708835e-06,
+      "loss": 1.5372,
+      "step": 37230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5735978945482e-06,
+      "loss": 1.4504,
+      "step": 37240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.572321284003907e-06,
+      "loss": 1.3664,
+      "step": 37250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.571044953923962e-06,
+      "loss": 1.5692,
+      "step": 37260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.569768904694235e-06,
+      "loss": 1.4104,
+      "step": 37270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5684931367005144e-06,
+      "loss": 1.4354,
+      "step": 37280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5672176503285013e-06,
+      "loss": 1.3471,
+      "step": 37290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5659424459638128e-06,
+      "loss": 1.4909,
+      "step": 37300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5646675239919807e-06,
+      "loss": 1.4943,
+      "step": 37310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5633928847984503e-06,
+      "loss": 1.4399,
+      "step": 37320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.562118528768583e-06,
+      "loss": 1.548,
+      "step": 37330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5608444562876535e-06,
+      "loss": 1.3201,
+      "step": 37340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5595706677408524e-06,
+      "loss": 1.4466,
+      "step": 37350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.55829716351328e-06,
+      "loss": 1.4585,
+      "step": 37360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5570239439899564e-06,
+      "loss": 1.3526,
+      "step": 37370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5557510095558123e-06,
+      "loss": 1.4159,
+      "step": 37380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.554478360595693e-06,
+      "loss": 1.4015,
+      "step": 37390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5532059974943569e-06,
+      "loss": 1.4128,
+      "step": 37400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5519339206364774e-06,
+      "loss": 1.4877,
+      "step": 37410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5506621304066399e-06,
+      "loss": 1.4821,
+      "step": 37420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5493906271893433e-06,
+      "loss": 1.4823,
+      "step": 37430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5481194113690017e-06,
+      "loss": 1.4132,
+      "step": 37440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.546848483329939e-06,
+      "loss": 1.4167,
+      "step": 37450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5455778434563941e-06,
+      "loss": 1.5262,
+      "step": 37460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5443074921325194e-06,
+      "loss": 1.4308,
+      "step": 37470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5430374297423784e-06,
+      "loss": 1.4741,
+      "step": 37480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5417676566699486e-06,
+      "loss": 1.3419,
+      "step": 37490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5404981732991186e-06,
+      "loss": 1.3718,
+      "step": 37500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5392289800136913e-06,
+      "loss": 1.4426,
+      "step": 37510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5379600771973805e-06,
+      "loss": 1.5031,
+      "step": 37520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5366914652338127e-06,
+      "loss": 1.4311,
+      "step": 37530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5354231445065253e-06,
+      "loss": 1.3541,
+      "step": 37540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5341551153989699e-06,
+      "loss": 1.3944,
+      "step": 37550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.532887378294508e-06,
+      "loss": 1.4591,
+      "step": 37560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5316199335764132e-06,
+      "loss": 1.3943,
+      "step": 37570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5303527816278724e-06,
+      "loss": 1.4028,
+      "step": 37580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5290859228319815e-06,
+      "loss": 1.4352,
+      "step": 37590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5278193575717492e-06,
+      "loss": 1.3571,
+      "step": 37600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5265530862300948e-06,
+      "loss": 1.5498,
+      "step": 37610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5252871091898503e-06,
+      "loss": 1.4162,
+      "step": 37620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.524021426833756e-06,
+      "loss": 1.4285,
+      "step": 37630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.522756039544465e-06,
+      "loss": 1.4168,
+      "step": 37640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5214909477045415e-06,
+      "loss": 1.3011,
+      "step": 37650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5202261516964586e-06,
+      "loss": 1.4522,
+      "step": 37660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5189616519026012e-06,
+      "loss": 1.435,
+      "step": 37670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5176974487052658e-06,
+      "loss": 1.4572,
+      "step": 37680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5164335424866566e-06,
+      "loss": 1.3045,
+      "step": 37690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5151699336288902e-06,
+      "loss": 1.3632,
+      "step": 37700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5139066225139922e-06,
+      "loss": 1.4949,
+      "step": 37710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5126436095238978e-06,
+      "loss": 1.4133,
+      "step": 37720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.511380895040453e-06,
+      "loss": 1.4731,
+      "step": 37730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5101184794454135e-06,
+      "loss": 1.3715,
+      "step": 37740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5088563631204441e-06,
+      "loss": 1.4339,
+      "step": 37750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5075945464471198e-06,
+      "loss": 1.4941,
+      "step": 37760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.506333029806924e-06,
+      "loss": 1.3379,
+      "step": 37770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5050718135812498e-06,
+      "loss": 1.5096,
+      "step": 37780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5038108981514004e-06,
+      "loss": 1.3507,
+      "step": 37790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5025502838985884e-06,
+      "loss": 1.4478,
+      "step": 37800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5012899712039313e-06,
+      "loss": 1.5175,
+      "step": 37810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5000299604484608e-06,
+      "loss": 1.3026,
+      "step": 37820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4987702520131142e-06,
+      "loss": 1.4499,
+      "step": 37830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4975108462787382e-06,
+      "loss": 1.3638,
+      "step": 37840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4962517436260875e-06,
+      "loss": 1.4093,
+      "step": 37850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4949929444358263e-06,
+      "loss": 1.4447,
+      "step": 37860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4937344490885259e-06,
+      "loss": 1.3634,
+      "step": 37870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4924762579646653e-06,
+      "loss": 1.471,
+      "step": 37880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4912183714446352e-06,
+      "loss": 1.4575,
+      "step": 37890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4899607899087282e-06,
+      "loss": 1.4139,
+      "step": 37900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4887035137371492e-06,
+      "loss": 1.4911,
+      "step": 37910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4874465433100093e-06,
+      "loss": 1.371,
+      "step": 37920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4861898790073276e-06,
+      "loss": 1.4496,
+      "step": 37930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4849335212090305e-06,
+      "loss": 1.403,
+      "step": 37940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4836774702949514e-06,
+      "loss": 1.4362,
+      "step": 37950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4824217266448312e-06,
+      "loss": 1.4432,
+      "step": 37960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.481166290638318e-06,
+      "loss": 1.4209,
+      "step": 37970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4799111626549668e-06,
+      "loss": 1.4867,
+      "step": 37980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4786563430742385e-06,
+      "loss": 1.443,
+      "step": 37990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4774018322755027e-06,
+      "loss": 1.4485,
+      "step": 38000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.476147630638034e-06,
+      "loss": 1.4509,
+      "step": 38010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4748937385410141e-06,
+      "loss": 1.4195,
+      "step": 38020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4736401563635323e-06,
+      "loss": 1.4522,
+      "step": 38030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4723868844845818e-06,
+      "loss": 1.4632,
+      "step": 38040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4711339232830641e-06,
+      "loss": 1.4854,
+      "step": 38050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4698812731377852e-06,
+      "loss": 1.5638,
+      "step": 38060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4686289344274587e-06,
+      "loss": 1.2963,
+      "step": 38070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4673769075307023e-06,
+      "loss": 1.5183,
+      "step": 38080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4661251928260407e-06,
+      "loss": 1.3801,
+      "step": 38090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4648737906919033e-06,
+      "loss": 1.3948,
+      "step": 38100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4636227015066262e-06,
+      "loss": 1.4519,
+      "step": 38110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4623719256484495e-06,
+      "loss": 1.4074,
+      "step": 38120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4611214634955193e-06,
+      "loss": 1.4864,
+      "step": 38130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4598713154258873e-06,
+      "loss": 1.3929,
+      "step": 38140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4586214818175098e-06,
+      "loss": 1.4247,
+      "step": 38150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4573719630482485e-06,
+      "loss": 1.4371,
+      "step": 38160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4561227594958675e-06,
+      "loss": 1.4941,
+      "step": 38170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4548738715380383e-06,
+      "loss": 1.4328,
+      "step": 38180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.453625299552337e-06,
+      "loss": 1.3315,
+      "step": 38190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4523770439162414e-06,
+      "loss": 1.3391,
+      "step": 38200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.451129105007138e-06,
+      "loss": 1.4546,
+      "step": 38210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4498814832023142e-06,
+      "loss": 1.3521,
+      "step": 38220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4486341788789622e-06,
+      "loss": 1.4599,
+      "step": 38230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4473871924141788e-06,
+      "loss": 1.3014,
+      "step": 38240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4461405241849656e-06,
+      "loss": 1.4558,
+      "step": 38250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.444894174568224e-06,
+      "loss": 1.5344,
+      "step": 38260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4436481439407634e-06,
+      "loss": 1.3877,
+      "step": 38270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4424024326792955e-06,
+      "loss": 1.5102,
+      "step": 38280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4411570411604348e-06,
+      "loss": 1.3095,
+      "step": 38290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4399119697606981e-06,
+      "loss": 1.4446,
+      "step": 38300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4386672188565093e-06,
+      "loss": 1.5185,
+      "step": 38310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4374227888241921e-06,
+      "loss": 1.4589,
+      "step": 38320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4361786800399739e-06,
+      "loss": 1.4478,
+      "step": 38330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4349348928799855e-06,
+      "loss": 1.3601,
+      "step": 38340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.433691427720259e-06,
+      "loss": 1.418,
+      "step": 38350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4324482849367304e-06,
+      "loss": 1.4269,
+      "step": 38360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4312054649052384e-06,
+      "loss": 1.4712,
+      "step": 38370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.429962968001524e-06,
+      "loss": 1.5113,
+      "step": 38380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4287207946012297e-06,
+      "loss": 1.3482,
+      "step": 38390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.427478945079901e-06,
+      "loss": 1.4029,
+      "step": 38400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4262374198129847e-06,
+      "loss": 1.5327,
+      "step": 38410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4249962191758316e-06,
+      "loss": 1.4222,
+      "step": 38420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.423755343543693e-06,
+      "loss": 1.4414,
+      "step": 38430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4225147932917193e-06,
+      "loss": 1.3783,
+      "step": 38440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4212745687949672e-06,
+      "loss": 1.3622,
+      "step": 38450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4200346704283917e-06,
+      "loss": 1.4018,
+      "step": 38460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4187950985668515e-06,
+      "loss": 1.4883,
+      "step": 38470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4175558535851037e-06,
+      "loss": 1.5233,
+      "step": 38480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4163169358578093e-06,
+      "loss": 1.282,
+      "step": 38490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.415078345759529e-06,
+      "loss": 1.3529,
+      "step": 38500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4138400836647237e-06,
+      "loss": 1.4589,
+      "step": 38510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4126021499477592e-06,
+      "loss": 1.3505,
+      "step": 38520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.411364544982896e-06,
+      "loss": 1.4915,
+      "step": 38530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4101272691442997e-06,
+      "loss": 1.4182,
+      "step": 38540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.408890322806034e-06,
+      "loss": 1.3789,
+      "step": 38550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4076537063420642e-06,
+      "loss": 1.4899,
+      "step": 38560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4064174201262558e-06,
+      "loss": 1.4756,
+      "step": 38570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.4051814645323742e-06,
+      "loss": 1.4332,
+      "step": 38580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4039458399340844e-06,
+      "loss": 1.3657,
+      "step": 38590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4027105467049521e-06,
+      "loss": 1.4622,
+      "step": 38600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4014755852184425e-06,
+      "loss": 1.466,
+      "step": 38610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.4002409558479202e-06,
+      "loss": 1.417,
+      "step": 38620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3990066589666502e-06,
+      "loss": 1.4936,
+      "step": 38630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.397772694947796e-06,
+      "loss": 1.3398,
+      "step": 38640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.396539064164421e-06,
+      "loss": 1.4478,
+      "step": 38650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3953057669894882e-06,
+      "loss": 1.5286,
+      "step": 38660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3940728037958589e-06,
+      "loss": 1.442,
+      "step": 38670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.392840174956294e-06,
+      "loss": 1.5515,
+      "step": 38680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.391607880843453e-06,
+      "loss": 1.4063,
+      "step": 38690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3903759218298946e-06,
+      "loss": 1.4558,
+      "step": 38700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.389144298288076e-06,
+      "loss": 1.5157,
+      "step": 38710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3879130105903526e-06,
+      "loss": 1.3234,
+      "step": 38720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.386682059108979e-06,
+      "loss": 1.4744,
+      "step": 38730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3854514442161077e-06,
+      "loss": 1.4058,
+      "step": 38740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.384221166283789e-06,
+      "loss": 1.4136,
+      "step": 38750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3829912256839725e-06,
+      "loss": 1.4091,
+      "step": 38760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3817616227885046e-06,
+      "loss": 1.4403,
+      "step": 38770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3805323579691305e-06,
+      "loss": 1.4708,
+      "step": 38780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3793034315974938e-06,
+      "loss": 1.4263,
+      "step": 38790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3780748440451314e-06,
+      "loss": 1.438,
+      "step": 38800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3768465956834847e-06,
+      "loss": 1.4019,
+      "step": 38810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3756186868838878e-06,
+      "loss": 1.439,
+      "step": 38820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.374391118017573e-06,
+      "loss": 1.48,
+      "step": 38830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.373163889455671e-06,
+      "loss": 1.3926,
+      "step": 38840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3719370015692078e-06,
+      "loss": 1.3827,
+      "step": 38850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3707104547291083e-06,
+      "loss": 1.4611,
+      "step": 38860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3694842493061932e-06,
+      "loss": 1.5121,
+      "step": 38870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3682583856711807e-06,
+      "loss": 1.4992,
+      "step": 38880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.367032864194684e-06,
+      "loss": 1.3452,
+      "step": 38890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.365807685247213e-06,
+      "loss": 1.4543,
+      "step": 38900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3645828491991774e-06,
+      "loss": 1.4861,
+      "step": 38910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3633583564208808e-06,
+      "loss": 1.4283,
+      "step": 38920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3621342072825216e-06,
+      "loss": 1.4493,
+      "step": 38930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3609104021541968e-06,
+      "loss": 1.5066,
+      "step": 38940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3596869414058982e-06,
+      "loss": 1.3639,
+      "step": 38950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3584638254075132e-06,
+      "loss": 1.4954,
+      "step": 38960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3572410545288272e-06,
+      "loss": 1.4466,
+      "step": 38970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3560186291395175e-06,
+      "loss": 1.4522,
+      "step": 38980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3547965496091592e-06,
+      "loss": 1.4244,
+      "step": 38990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3535748163072233e-06,
+      "loss": 1.3698,
+      "step": 39000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3523534296030744e-06,
+      "loss": 1.4106,
+      "step": 39010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3511323898659749e-06,
+      "loss": 1.408,
+      "step": 39020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.34991169746508e-06,
+      "loss": 1.43,
+      "step": 39030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3486913527694406e-06,
+      "loss": 1.3352,
+      "step": 39040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3474713561480024e-06,
+      "loss": 1.4091,
+      "step": 39050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3462517079696064e-06,
+      "loss": 1.5239,
+      "step": 39060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3450324086029882e-06,
+      "loss": 1.3582,
+      "step": 39070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.343813458416776e-06,
+      "loss": 1.5345,
+      "step": 39080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3425948577794954e-06,
+      "loss": 1.4434,
+      "step": 39090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3413766070595641e-06,
+      "loss": 1.3615,
+      "step": 39100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3401587066252944e-06,
+      "loss": 1.4212,
+      "step": 39110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3389411568448948e-06,
+      "loss": 1.4709,
+      "step": 39120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3377239580864653e-06,
+      "loss": 1.4327,
+      "step": 39130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3365071107180008e-06,
+      "loss": 1.3949,
+      "step": 39140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3352906151073895e-06,
+      "loss": 1.4092,
+      "step": 39150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3340744716224143e-06,
+      "loss": 1.4847,
+      "step": 39160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3328586806307495e-06,
+      "loss": 1.3871,
+      "step": 39170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3316432424999648e-06,
+      "loss": 1.4529,
+      "step": 39180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3304281575975226e-06,
+      "loss": 1.4399,
+      "step": 39190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3292134262907785e-06,
+      "loss": 1.4891,
+      "step": 39200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3279990489469813e-06,
+      "loss": 1.4862,
+      "step": 39210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.326785025933272e-06,
+      "loss": 1.3931,
+      "step": 39220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3255713576166868e-06,
+      "loss": 1.4609,
+      "step": 39230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3243580443641518e-06,
+      "loss": 1.3638,
+      "step": 39240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3231450865424888e-06,
+      "loss": 1.4309,
+      "step": 39250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3219324845184078e-06,
+      "loss": 1.4884,
+      "step": 39260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.320720238658515e-06,
+      "loss": 1.4332,
+      "step": 39270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.319508349329307e-06,
+      "loss": 1.5076,
+      "step": 39280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3182968168971744e-06,
+      "loss": 1.3433,
+      "step": 39290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3170856417283983e-06,
+      "loss": 1.4195,
+      "step": 39300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3158748241891523e-06,
+      "loss": 1.4349,
+      "step": 39310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3146643646455007e-06,
+      "loss": 1.3714,
+      "step": 39320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3134542634634028e-06,
+      "loss": 1.455,
+      "step": 39330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.312244521008708e-06,
+      "loss": 1.3798,
+      "step": 39340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3110351376471542e-06,
+      "loss": 1.4276,
+      "step": 39350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3098261137443744e-06,
+      "loss": 1.5288,
+      "step": 39360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.308617449665892e-06,
+      "loss": 1.3355,
+      "step": 39370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.307409145777121e-06,
+      "loss": 1.4925,
+      "step": 39380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3062012024433673e-06,
+      "loss": 1.448,
+      "step": 39390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3049936200298274e-06,
+      "loss": 1.3935,
+      "step": 39400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3037863989015886e-06,
+      "loss": 1.4504,
+      "step": 39410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.302579539423629e-06,
+      "loss": 1.3869,
+      "step": 39420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3013730419608177e-06,
+      "loss": 1.5067,
+      "step": 39430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3001669068779137e-06,
+      "loss": 1.4293,
+      "step": 39440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2989611345395674e-06,
+      "loss": 1.3458,
+      "step": 39450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2977557253103185e-06,
+      "loss": 1.4701,
+      "step": 39460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2965506795545973e-06,
+      "loss": 1.3904,
+      "step": 39470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.295345997636725e-06,
+      "loss": 1.4134,
+      "step": 39480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.294141679920911e-06,
+      "loss": 1.4125,
+      "step": 39490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2929377267712567e-06,
+      "loss": 1.3364,
+      "step": 39500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2917341385517518e-06,
+      "loss": 1.4288,
+      "step": 39510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.290530915626276e-06,
+      "loss": 1.4032,
+      "step": 39520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2893280583585992e-06,
+      "loss": 1.5039,
+      "step": 39530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.28812556711238e-06,
+      "loss": 1.3574,
+      "step": 39540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2869234422511663e-06,
+      "loss": 1.3449,
+      "step": 39550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2857216841383958e-06,
+      "loss": 1.5234,
+      "step": 39560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.284520293137395e-06,
+      "loss": 1.3765,
+      "step": 39570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2833192696113797e-06,
+      "loss": 1.4424,
+      "step": 39580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2821186139234542e-06,
+      "loss": 1.3317,
+      "step": 39590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2809183264366115e-06,
+      "loss": 1.3836,
+      "step": 39600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2797184075137337e-06,
+      "loss": 1.4163,
+      "step": 39610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2785188575175916e-06,
+      "loss": 1.4743,
+      "step": 39620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2773196768108437e-06,
+      "loss": 1.4453,
+      "step": 39630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2761208657560376e-06,
+      "loss": 1.4338,
+      "step": 39640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2749224247156089e-06,
+      "loss": 1.4016,
+      "step": 39650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2737243540518812e-06,
+      "loss": 1.5112,
+      "step": 39660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.272526654127066e-06,
+      "loss": 1.45,
+      "step": 39670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2713293253032632e-06,
+      "loss": 1.3971,
+      "step": 39680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2701323679424604e-06,
+      "loss": 1.4358,
+      "step": 39690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2689357824065335e-06,
+      "loss": 1.3354,
+      "step": 39700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2677395690572419e-06,
+      "loss": 1.4013,
+      "step": 39710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2665437282562398e-06,
+      "loss": 1.4323,
+      "step": 39720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2653482603650626e-06,
+      "loss": 1.4821,
+      "step": 39730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.264153165745136e-06,
+      "loss": 1.4324,
+      "step": 39740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2629584447577714e-06,
+      "loss": 1.3829,
+      "step": 39750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.2617640977641684e-06,
+      "loss": 1.4268,
+      "step": 39760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2605701251254128e-06,
+      "loss": 1.4233,
+      "step": 39770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2593765272024767e-06,
+      "loss": 1.4577,
+      "step": 39780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2581833043562208e-06,
+      "loss": 1.4124,
+      "step": 39790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2569904569473901e-06,
+      "loss": 1.4621,
+      "step": 39800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2557979853366174e-06,
+      "loss": 1.5042,
+      "step": 39810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2546058898844207e-06,
+      "loss": 1.4579,
+      "step": 39820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2534141709512073e-06,
+      "loss": 1.4372,
+      "step": 39830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2522228288972676e-06,
+      "loss": 1.3665,
+      "step": 39840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2510318640827796e-06,
+      "loss": 1.4026,
+      "step": 39850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.249841276867806e-06,
+      "loss": 1.4675,
+      "step": 39860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2486510676122958e-06,
+      "loss": 1.4416,
+      "step": 39870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2474612366760861e-06,
+      "loss": 1.3847,
+      "step": 39880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2462717844188948e-06,
+      "loss": 1.3536,
+      "step": 39890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2450827112003292e-06,
+      "loss": 1.3502,
+      "step": 39900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2438940173798806e-06,
+      "loss": 1.4988,
+      "step": 39910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2427057033169252e-06,
+      "loss": 1.4392,
+      "step": 39920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2415177693707265e-06,
+      "loss": 1.4383,
+      "step": 39930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.240330215900431e-06,
+      "loss": 1.4716,
+      "step": 39940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2391430432650708e-06,
+      "loss": 1.431,
+      "step": 39950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2379562518235628e-06,
+      "loss": 1.4267,
+      "step": 39960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2367698419347097e-06,
+      "loss": 1.3423,
+      "step": 39970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2355838139571954e-06,
+      "loss": 1.5091,
+      "step": 39980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2343981682495924e-06,
+      "loss": 1.4205,
+      "step": 39990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2332129051703553e-06,
+      "loss": 1.3694,
+      "step": 40000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2320280250778245e-06,
+      "loss": 1.4475,
+      "step": 40010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2308435283302222e-06,
+      "loss": 1.4081,
+      "step": 40020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.229659415285659e-06,
+      "loss": 1.4293,
+      "step": 40030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2284756863021249e-06,
+      "loss": 1.3588,
+      "step": 40040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2272923417374965e-06,
+      "loss": 1.334,
+      "step": 40050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.226109381949534e-06,
+      "loss": 1.4112,
+      "step": 40060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2249268072958786e-06,
+      "loss": 1.3265,
+      "step": 40070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2237446181340587e-06,
+      "loss": 1.474,
+      "step": 40080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.222562814821484e-06,
+      "loss": 1.4578,
+      "step": 40090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2213813977154485e-06,
+      "loss": 1.3434,
+      "step": 40100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2202003671731286e-06,
+      "loss": 1.5681,
+      "step": 40110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2190197235515846e-06,
+      "loss": 1.4388,
+      "step": 40120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2178394672077587e-06,
+      "loss": 1.4466,
+      "step": 40130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2166595984984786e-06,
+      "loss": 1.4214,
+      "step": 40140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2154801177804532e-06,
+      "loss": 1.3799,
+      "step": 40150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2143010254102716e-06,
+      "loss": 1.5557,
+      "step": 40160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.213122321744409e-06,
+      "loss": 1.3594,
+      "step": 40170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2119440071392222e-06,
+      "loss": 1.5001,
+      "step": 40180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2107660819509495e-06,
+      "loss": 1.3559,
+      "step": 40190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2095885465357128e-06,
+      "loss": 1.4713,
+      "step": 40200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2084114012495148e-06,
+      "loss": 1.4471,
+      "step": 40210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2072346464482413e-06,
+      "loss": 1.3626,
+      "step": 40220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2060582824876585e-06,
+      "loss": 1.5346,
+      "step": 40230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.204882309723418e-06,
+      "loss": 1.3659,
+      "step": 40240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2037067285110485e-06,
+      "loss": 1.3254,
+      "step": 40250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2025315392059631e-06,
+      "loss": 1.4698,
+      "step": 40260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2013567421634558e-06,
+      "loss": 1.4416,
+      "step": 40270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.2001823377387022e-06,
+      "loss": 1.5528,
+      "step": 40280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1990083262867592e-06,
+      "loss": 1.4568,
+      "step": 40290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1978347081625642e-06,
+      "loss": 1.3224,
+      "step": 40300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1966614837209365e-06,
+      "loss": 1.5011,
+      "step": 40310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.195488653316576e-06,
+      "loss": 1.3425,
+      "step": 40320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1943162173040637e-06,
+      "loss": 1.4597,
+      "step": 40330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.1931441760378611e-06,
+      "loss": 1.3948,
+      "step": 40340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1919725298723107e-06,
+      "loss": 1.3361,
+      "step": 40350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.190801279161635e-06,
+      "loss": 1.4029,
+      "step": 40360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1896304242599373e-06,
+      "loss": 1.4066,
+      "step": 40370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1884599655212015e-06,
+      "loss": 1.5142,
+      "step": 40380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.187289903299291e-06,
+      "loss": 1.4133,
+      "step": 40390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1861202379479504e-06,
+      "loss": 1.384,
+      "step": 40400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1849509698208029e-06,
+      "loss": 1.4916,
+      "step": 40410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1837820992713527e-06,
+      "loss": 1.3657,
+      "step": 40420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1826136266529836e-06,
+      "loss": 1.4451,
+      "step": 40430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.181445552318959e-06,
+      "loss": 1.4167,
+      "step": 40440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1802778766224218e-06,
+      "loss": 1.41,
+      "step": 40450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1791105999163942e-06,
+      "loss": 1.4497,
+      "step": 40460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1779437225537785e-06,
+      "loss": 1.3688,
+      "step": 40470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1767772448873553e-06,
+      "loss": 1.5056,
+      "step": 40480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1756111672697848e-06,
+      "loss": 1.3459,
+      "step": 40490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1744454900536068e-06,
+      "loss": 1.4545,
+      "step": 40500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1732802135912402e-06,
+      "loss": 1.4489,
+      "step": 40510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.172115338234979e-06,
+      "loss": 1.4088,
+      "step": 40520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1709508643370024e-06,
+      "loss": 1.5015,
+      "step": 40530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1697867922493633e-06,
+      "loss": 1.3987,
+      "step": 40540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.168623122323995e-06,
+      "loss": 1.4377,
+      "step": 40550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1674598549127088e-06,
+      "loss": 1.4546,
+      "step": 40560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1662969903671945e-06,
+      "loss": 1.3576,
+      "step": 40570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1651345290390197e-06,
+      "loss": 1.4685,
+      "step": 40580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.163972471279631e-06,
+      "loss": 1.4401,
+      "step": 40590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1628108174403526e-06,
+      "loss": 1.4066,
+      "step": 40600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.161649567872385e-06,
+      "loss": 1.5439,
+      "step": 40610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1604887229268073e-06,
+      "loss": 1.4436,
+      "step": 40620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1593282829545788e-06,
+      "loss": 1.4805,
+      "step": 40630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.158168248306534e-06,
+      "loss": 1.3586,
+      "step": 40640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1570086193333847e-06,
+      "loss": 1.4584,
+      "step": 40650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1558493963857206e-06,
+      "loss": 1.4188,
+      "step": 40660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1546905798140088e-06,
+      "loss": 1.4802,
+      "step": 40670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1535321699685934e-06,
+      "loss": 1.4612,
+      "step": 40680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1523741671996962e-06,
+      "loss": 1.4337,
+      "step": 40690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1512165718574137e-06,
+      "loss": 1.374,
+      "step": 40700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1500593842917212e-06,
+      "loss": 1.452,
+      "step": 40710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1489026048524707e-06,
+      "loss": 1.4943,
+      "step": 40720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1477462338893892e-06,
+      "loss": 1.4582,
+      "step": 40730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1465902717520834e-06,
+      "loss": 1.3527,
+      "step": 40740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1454347187900334e-06,
+      "loss": 1.4068,
+      "step": 40750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1442795753525969e-06,
+      "loss": 1.4729,
+      "step": 40760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1431248417890067e-06,
+      "loss": 1.382,
+      "step": 40770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.141970518448374e-06,
+      "loss": 1.4987,
+      "step": 40780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1408166056796825e-06,
+      "loss": 1.3694,
+      "step": 40790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1396631038317944e-06,
+      "loss": 1.4514,
+      "step": 40800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.138510013253447e-06,
+      "loss": 1.3854,
+      "step": 40810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1373573342932537e-06,
+      "loss": 1.3621,
+      "step": 40820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1362050672997015e-06,
+      "loss": 1.4794,
+      "step": 40830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1350532126211564e-06,
+      "loss": 1.3071,
+      "step": 40840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.133901770605857e-06,
+      "loss": 1.3775,
+      "step": 40850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1327507416019174e-06,
+      "loss": 1.4498,
+      "step": 40860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.131600125957328e-06,
+      "loss": 1.4028,
+      "step": 40870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1304499240199523e-06,
+      "loss": 1.4363,
+      "step": 40880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1293001361375308e-06,
+      "loss": 1.4748,
+      "step": 40890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1281507626576774e-06,
+      "loss": 1.3613,
+      "step": 40900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1270018039278817e-06,
+      "loss": 1.4739,
+      "step": 40910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1258532602955069e-06,
+      "loss": 1.3384,
+      "step": 40920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1247051321077917e-06,
+      "loss": 1.444,
+      "step": 40930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.123557419711848e-06,
+      "loss": 1.4001,
+      "step": 40940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1224101234546642e-06,
+      "loss": 1.4212,
+      "step": 40950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1212632436831015e-06,
+      "loss": 1.478,
+      "step": 40960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1201167807438933e-06,
+      "loss": 1.3528,
+      "step": 40970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1189707349836498e-06,
+      "loss": 1.5302,
+      "step": 40980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1178251067488543e-06,
+      "loss": 1.3729,
+      "step": 40990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1166798963858633e-06,
+      "loss": 1.4034,
+      "step": 41000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1155351042409074e-06,
+      "loss": 1.4148,
+      "step": 41010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1143907306600908e-06,
+      "loss": 1.4018,
+      "step": 41020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1132467759893905e-06,
+      "loss": 1.5349,
+      "step": 41030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1121032405746574e-06,
+      "loss": 1.4142,
+      "step": 41040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1109601247616175e-06,
+      "loss": 1.3532,
+      "step": 41050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1098174288958657e-06,
+      "loss": 1.4563,
+      "step": 41060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1086751533228735e-06,
+      "loss": 1.4022,
+      "step": 41070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1075332983879835e-06,
+      "loss": 1.5605,
+      "step": 41080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1063918644364122e-06,
+      "loss": 1.4646,
+      "step": 41090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1052508518132486e-06,
+      "loss": 1.356,
+      "step": 41100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1041102608634534e-06,
+      "loss": 1.5138,
+      "step": 41110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.102970091931861e-06,
+      "loss": 1.3849,
+      "step": 41120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.101830345363178e-06,
+      "loss": 1.427,
+      "step": 41130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.1006910215019825e-06,
+      "loss": 1.4439,
+      "step": 41140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0995521206927256e-06,
+      "loss": 1.4116,
+      "step": 41150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.09841364327973e-06,
+      "loss": 1.4296,
+      "step": 41160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0972755896071913e-06,
+      "loss": 1.5217,
+      "step": 41170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0961379600191758e-06,
+      "loss": 1.4911,
+      "step": 41180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.095000754859622e-06,
+      "loss": 1.3484,
+      "step": 41190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0938639744723405e-06,
+      "loss": 1.4796,
+      "step": 41200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0927276192010133e-06,
+      "loss": 1.5058,
+      "step": 41210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0915916893891937e-06,
+      "loss": 1.4229,
+      "step": 41220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0904561853803061e-06,
+      "loss": 1.4338,
+      "step": 41230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0893211075176465e-06,
+      "loss": 1.3644,
+      "step": 41240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0881864561443825e-06,
+      "loss": 1.4557,
+      "step": 41250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0870522316035522e-06,
+      "loss": 1.4939,
+      "step": 41260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0859184342380645e-06,
+      "loss": 1.4803,
+      "step": 41270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0847850643906997e-06,
+      "loss": 1.5065,
+      "step": 41280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0836521224041085e-06,
+      "loss": 1.4411,
+      "step": 41290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0825196086208122e-06,
+      "loss": 1.3239,
+      "step": 41300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.081387523383203e-06,
+      "loss": 1.5246,
+      "step": 41310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.080255867033543e-06,
+      "loss": 1.4779,
+      "step": 41320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0791246399139652e-06,
+      "loss": 1.4785,
+      "step": 41330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0779938423664724e-06,
+      "loss": 1.3923,
+      "step": 41340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.076863474732938e-06,
+      "loss": 1.3657,
+      "step": 41350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0757335373551047e-06,
+      "loss": 1.4601,
+      "step": 41360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.074604030574586e-06,
+      "loss": 1.3886,
+      "step": 41370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.073474954732864e-06,
+      "loss": 1.4278,
+      "step": 41380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0723463101712922e-06,
+      "loss": 1.3434,
+      "step": 41390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0712180972310923e-06,
+      "loss": 1.3769,
+      "step": 41400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.070090316253356e-06,
+      "loss": 1.5068,
+      "step": 41410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0689629675790454e-06,
+      "loss": 1.3305,
+      "step": 41420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.067836051548988e-06,
+      "loss": 1.4856,
+      "step": 41430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0667095685038864e-06,
+      "loss": 1.3491,
+      "step": 41440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0655835187843084e-06,
+      "loss": 1.38,
+      "step": 41450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0644579027306914e-06,
+      "loss": 1.4206,
+      "step": 41460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.063332720683342e-06,
+      "loss": 1.3237,
+      "step": 41470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.062207972982436e-06,
+      "loss": 1.4107,
+      "step": 41480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0610836599680168e-06,
+      "loss": 1.4215,
+      "step": 41490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0599597819799978e-06,
+      "loss": 1.3973,
+      "step": 41500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0588363393581602e-06,
+      "loss": 1.4478,
+      "step": 41510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.0577133324421524e-06,
+      "loss": 1.4714,
+      "step": 41520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0565907615714927e-06,
+      "loss": 1.5019,
+      "step": 41530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0554686270855661e-06,
+      "loss": 1.3743,
+      "step": 41540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0543469293236283e-06,
+      "loss": 1.3474,
+      "step": 41550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0532256686248008e-06,
+      "loss": 1.4836,
+      "step": 41560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0521048453280732e-06,
+      "loss": 1.4531,
+      "step": 41570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0509844597723029e-06,
+      "loss": 1.4126,
+      "step": 41580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0498645122962156e-06,
+      "loss": 1.3478,
+      "step": 41590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0487450032384045e-06,
+      "loss": 1.4182,
+      "step": 41600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0476259329373283e-06,
+      "loss": 1.4239,
+      "step": 41610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0465073017313155e-06,
+      "loss": 1.4189,
+      "step": 41620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0453891099585609e-06,
+      "loss": 1.4461,
+      "step": 41630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0442713579571257e-06,
+      "loss": 1.3771,
+      "step": 41640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.043154046064941e-06,
+      "loss": 1.4013,
+      "step": 41650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.042037174619801e-06,
+      "loss": 1.4156,
+      "step": 41660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0409207439593694e-06,
+      "loss": 1.3521,
+      "step": 41670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0398047544211756e-06,
+      "loss": 1.5173,
+      "step": 41680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0386892063426165e-06,
+      "loss": 1.3791,
+      "step": 41690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.037574100060953e-06,
+      "loss": 1.3873,
+      "step": 41700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0364594359133155e-06,
+      "loss": 1.4969,
+      "step": 41710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0353452142366993e-06,
+      "loss": 1.3944,
+      "step": 41720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.034231435367966e-06,
+      "loss": 1.4641,
+      "step": 41730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0331180996438437e-06,
+      "loss": 1.4344,
+      "step": 41740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0320052074009252e-06,
+      "loss": 1.4123,
+      "step": 41750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.030892758975672e-06,
+      "loss": 1.4815,
+      "step": 41760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0297807547044093e-06,
+      "loss": 1.4197,
+      "step": 41770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0286691949233284e-06,
+      "loss": 1.4678,
+      "step": 41780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0275580799684858e-06,
+      "loss": 1.3852,
+      "step": 41790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0264474101758037e-06,
+      "loss": 1.5283,
+      "step": 41800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0253371858810706e-06,
+      "loss": 1.436,
+      "step": 41810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.02422740741994e-06,
+      "loss": 1.3963,
+      "step": 41820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0231180751279293e-06,
+      "loss": 1.5025,
+      "step": 41830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0220091893404226e-06,
+      "loss": 1.317,
+      "step": 41840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0209007503926678e-06,
+      "loss": 1.4152,
+      "step": 41850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0197927586197797e-06,
+      "loss": 1.4653,
+      "step": 41860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0186852143567366e-06,
+      "loss": 1.3699,
+      "step": 41870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.01757811793838e-06,
+      "loss": 1.5329,
+      "step": 41880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.016471469699418e-06,
+      "loss": 1.3547,
+      "step": 41890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0153652699744226e-06,
+      "loss": 1.4504,
+      "step": 41900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0142595190978308e-06,
+      "loss": 1.4904,
+      "step": 41910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0131542174039427e-06,
+      "loss": 1.3863,
+      "step": 41920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0120493652269236e-06,
+      "loss": 1.3558,
+      "step": 41930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0109449629008026e-06,
+      "loss": 1.3245,
+      "step": 41940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0098410107594717e-06,
+      "loss": 1.3773,
+      "step": 41950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0087375091366906e-06,
+      "loss": 1.4574,
+      "step": 41960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0076344583660772e-06,
+      "loss": 1.5013,
+      "step": 41970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.006531858781117e-06,
+      "loss": 1.4886,
+      "step": 41980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0054297107151583e-06,
+      "loss": 1.351,
+      "step": 41990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.004328014501412e-06,
+      "loss": 1.411,
+      "step": 42000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0032267704729534e-06,
+      "loss": 1.4585,
+      "step": 42010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0021259789627208e-06,
+      "loss": 1.3844,
+      "step": 42020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.0010256403035155e-06,
+      "loss": 1.4913,
+      "step": 42030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.999257548280018e-07,
+      "loss": 1.4466,
+      "step": 42040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.988263228687073e-07,
+      "loss": 1.3284,
+      "step": 42050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.977273447580224e-07,
+      "loss": 1.5409,
+      "step": 42060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.966288208282e-07,
+      "loss": 1.4119,
+      "step": 42070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.955307514113566e-07,
+      "loss": 1.5206,
+      "step": 42080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.944331368394699e-07,
+      "loss": 1.3731,
+      "step": 42090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.933359774443813e-07,
+      "loss": 1.3829,
+      "step": 42100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.922392735577937e-07,
+      "loss": 1.5032,
+      "step": 42110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.911430255112731e-07,
+      "loss": 1.5052,
+      "step": 42120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.900472336362469e-07,
+      "loss": 1.3609,
+      "step": 42130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.889518982640051e-07,
+      "loss": 1.3735,
+      "step": 42140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.878570197256999e-07,
+      "loss": 1.3866,
+      "step": 42150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.867625983523444e-07,
+      "loss": 1.3801,
+      "step": 42160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.856686344748146e-07,
+      "loss": 1.3817,
+      "step": 42170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.845751284238475e-07,
+      "loss": 1.4076,
+      "step": 42180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.834820805300419e-07,
+      "loss": 1.4116,
+      "step": 42190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.823894911238582e-07,
+      "loss": 1.4142,
+      "step": 42200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.812973605356179e-07,
+      "loss": 1.5541,
+      "step": 42210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.80205689095504e-07,
+      "loss": 1.4406,
+      "step": 42220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.791144771335614e-07,
+      "loss": 1.506,
+      "step": 42230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.780237249796932e-07,
+      "loss": 1.3504,
+      "step": 42240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.769334329636675e-07,
+      "loss": 1.4801,
+      "step": 42250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.75843601415111e-07,
+      "loss": 1.4391,
+      "step": 42260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.747542306635114e-07,
+      "loss": 1.3597,
+      "step": 42270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.736653210382175e-07,
+      "loss": 1.473,
+      "step": 42280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.725768728684379e-07,
+      "loss": 1.475,
+      "step": 42290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.714888864832432e-07,
+      "loss": 1.4064,
+      "step": 42300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.70401362211562e-07,
+      "loss": 1.4017,
+      "step": 42310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.69314300382187e-07,
+      "loss": 1.4199,
+      "step": 42320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.682277013237662e-07,
+      "loss": 1.4148,
+      "step": 42330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.671415653648102e-07,
+      "loss": 1.3585,
+      "step": 42340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.660558928336916e-07,
+      "loss": 1.3514,
+      "step": 42350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.649706840586397e-07,
+      "loss": 1.4919,
+      "step": 42360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.638859393677452e-07,
+      "loss": 1.3841,
+      "step": 42370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.628016590889576e-07,
+      "loss": 1.4232,
+      "step": 42380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.617178435500868e-07,
+      "loss": 1.3803,
+      "step": 42390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.606344930788016e-07,
+      "loss": 1.3723,
+      "step": 42400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.595516080026318e-07,
+      "loss": 1.4239,
+      "step": 42410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.58469188648963e-07,
+      "loss": 1.4411,
+      "step": 42420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.57387235345043e-07,
+      "loss": 1.5041,
+      "step": 42430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.56305748417978e-07,
+      "loss": 1.3588,
+      "step": 42440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.552247281947323e-07,
+      "loss": 1.4206,
+      "step": 42450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.541441750021315e-07,
+      "loss": 1.5764,
+      "step": 42460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.530640891668574e-07,
+      "loss": 1.3939,
+      "step": 42470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.519844710154518e-07,
+      "loss": 1.4828,
+      "step": 42480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.509053208743144e-07,
+      "loss": 1.4118,
+      "step": 42490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.498266390697047e-07,
+      "loss": 1.4485,
+      "step": 42500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.48748425927738e-07,
+      "loss": 1.5287,
+      "step": 42510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.476706817743906e-07,
+      "loss": 1.4017,
+      "step": 42520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.465934069354957e-07,
+      "loss": 1.4496,
+      "step": 42530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.455166017367451e-07,
+      "loss": 1.3306,
+      "step": 42540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.444402665036878e-07,
+      "loss": 1.3518,
+      "step": 42550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.433644015617327e-07,
+      "loss": 1.547,
+      "step": 42560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.422890072361446e-07,
+      "loss": 1.3149,
+      "step": 42570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.412140838520461e-07,
+      "loss": 1.4285,
+      "step": 42580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.401396317344194e-07,
+      "loss": 1.4326,
+      "step": 42590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.390656512081009e-07,
+      "loss": 1.3547,
+      "step": 42600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.379921425977867e-07,
+      "loss": 1.4309,
+      "step": 42610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.3691910622803e-07,
+      "loss": 1.4644,
+      "step": 42620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.358465424232412e-07,
+      "loss": 1.4552,
+      "step": 42630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.347744515076873e-07,
+      "loss": 1.4328,
+      "step": 42640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.337028338054929e-07,
+      "loss": 1.4195,
+      "step": 42650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.326316896406385e-07,
+      "loss": 1.3694,
+      "step": 42660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.315610193369637e-07,
+      "loss": 1.3952,
+      "step": 42670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.304908232181636e-07,
+      "loss": 1.4514,
+      "step": 42680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.294211016077878e-07,
+      "loss": 1.3756,
+      "step": 42690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.283518548292453e-07,
+      "loss": 1.3637,
+      "step": 42700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.272830832058008e-07,
+      "loss": 1.4247,
+      "step": 42710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.262147870605748e-07,
+      "loss": 1.471,
+      "step": 42720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.251469667165447e-07,
+      "loss": 1.4424,
+      "step": 42730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.240796224965437e-07,
+      "loss": 1.4189,
+      "step": 42740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.230127547232609e-07,
+      "loss": 1.405,
+      "step": 42750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.21946363719241e-07,
+      "loss": 1.4881,
+      "step": 42760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.208804498068874e-07,
+      "loss": 1.4066,
+      "step": 42770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.198150133084547e-07,
+      "loss": 1.4954,
+      "step": 42780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.18750054546056e-07,
+      "loss": 1.4357,
+      "step": 42790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.176855738416596e-07,
+      "loss": 1.4369,
+      "step": 42800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.166215715170892e-07,
+      "loss": 1.4629,
+      "step": 42810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.155580478940235e-07,
+      "loss": 1.4404,
+      "step": 42820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.144950032939967e-07,
+      "loss": 1.4191,
+      "step": 42830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.134324380383983e-07,
+      "loss": 1.4148,
+      "step": 42840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.123703524484728e-07,
+      "loss": 1.4547,
+      "step": 42850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.1130874684532e-07,
+      "loss": 1.4913,
+      "step": 42860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.102476215498937e-07,
+      "loss": 1.3255,
+      "step": 42870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.09186976883003e-07,
+      "loss": 1.4006,
+      "step": 42880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.081268131653123e-07,
+      "loss": 1.4888,
+      "step": 42890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.070671307173399e-07,
+      "loss": 1.3471,
+      "step": 42900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.060079298594581e-07,
+      "loss": 1.4334,
+      "step": 42910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.049492109118951e-07,
+      "loss": 1.3419,
+      "step": 42920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.038909741947321e-07,
+      "loss": 1.4209,
+      "step": 42930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.028332200279047e-07,
+      "loss": 1.3751,
+      "step": 42940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.017759487312033e-07,
+      "loss": 1.3926,
+      "step": 42950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.007191606242718e-07,
+      "loss": 1.5281,
+      "step": 42960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.996628560266077e-07,
+      "loss": 1.4642,
+      "step": 42970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.986070352575632e-07,
+      "loss": 1.4857,
+      "step": 42980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.975516986363434e-07,
+      "loss": 1.4103,
+      "step": 42990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.964968464820073e-07,
+      "loss": 1.4608,
+      "step": 43000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.954424791134678e-07,
+      "loss": 1.5591,
+      "step": 43010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.943885968494908e-07,
+      "loss": 1.3566,
+      "step": 43020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.933352000086956e-07,
+      "loss": 1.4653,
+      "step": 43030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.922822889095556e-07,
+      "loss": 1.4197,
+      "step": 43040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.912298638703943e-07,
+      "loss": 1.4563,
+      "step": 43050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.901779252093925e-07,
+      "loss": 1.4488,
+      "step": 43060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.891264732445817e-07,
+      "loss": 1.4298,
+      "step": 43070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.880755082938463e-07,
+      "loss": 1.4703,
+      "step": 43080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.870250306749234e-07,
+      "loss": 1.4334,
+      "step": 43090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.859750407054035e-07,
+      "loss": 1.4453,
+      "step": 43100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.849255387027289e-07,
+      "loss": 1.4541,
+      "step": 43110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.838765249841949e-07,
+      "loss": 1.4996,
+      "step": 43120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.828279998669495e-07,
+      "loss": 1.4789,
+      "step": 43130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.817799636679907e-07,
+      "loss": 1.279,
+      "step": 43140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.807324167041708e-07,
+      "loss": 1.423,
+      "step": 43150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.796853592921955e-07,
+      "loss": 1.4207,
+      "step": 43160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.786387917486197e-07,
+      "loss": 1.4426,
+      "step": 43170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.775927143898516e-07,
+      "loss": 1.4899,
+      "step": 43180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.765471275321505e-07,
+      "loss": 1.3624,
+      "step": 43190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.755020314916282e-07,
+      "loss": 1.35,
+      "step": 43200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.744574265842476e-07,
+      "loss": 1.5042,
+      "step": 43210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.734133131258241e-07,
+      "loss": 1.4274,
+      "step": 43220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.723696914320218e-07,
+      "loss": 1.4343,
+      "step": 43230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.713265618183589e-07,
+      "loss": 1.3897,
+      "step": 43240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.702839246002042e-07,
+      "loss": 1.3654,
+      "step": 43250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.692417800927763e-07,
+      "loss": 1.5299,
+      "step": 43260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.682001286111478e-07,
+      "loss": 1.3998,
+      "step": 43270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.671589704702388e-07,
+      "loss": 1.466,
+      "step": 43280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 8.661183059848227e-07,
+      "loss": 1.3189,
+      "step": 43290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.650781354695223e-07,
+      "loss": 1.4485,
+      "step": 43300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.640384592388123e-07,
+      "loss": 1.4264,
+      "step": 43310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.629992776070156e-07,
+      "loss": 1.4333,
+      "step": 43320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.619605908883081e-07,
+      "loss": 1.4873,
+      "step": 43330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.609223993967145e-07,
+      "loss": 1.4479,
+      "step": 43340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.598847034461111e-07,
+      "loss": 1.4718,
+      "step": 43350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.588475033502228e-07,
+      "loss": 1.3866,
+      "step": 43360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.578107994226264e-07,
+      "loss": 1.4416,
+      "step": 43370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.567745919767478e-07,
+      "loss": 1.46,
+      "step": 43380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.557388813258622e-07,
+      "loss": 1.4053,
+      "step": 43390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.54703667783096e-07,
+      "loss": 1.4492,
+      "step": 43400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.536689516614233e-07,
+      "loss": 1.4355,
+      "step": 43410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.526347332736689e-07,
+      "loss": 1.3955,
+      "step": 43420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.516010129325083e-07,
+      "loss": 1.528,
+      "step": 43430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.505677909504645e-07,
+      "loss": 1.361,
+      "step": 43440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.495350676399109e-07,
+      "loss": 1.4775,
+      "step": 43450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.485028433130701e-07,
+      "loss": 1.4487,
+      "step": 43460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.474711182820125e-07,
+      "loss": 1.454,
+      "step": 43470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.464398928586603e-07,
+      "loss": 1.4488,
+      "step": 43480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.454091673547834e-07,
+      "loss": 1.4534,
+      "step": 43490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.443789420819984e-07,
+      "loss": 1.4757,
+      "step": 43500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.433492173517733e-07,
+      "loss": 1.4747,
+      "step": 43510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.423199934754237e-07,
+      "loss": 1.3616,
+      "step": 43520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.412912707641141e-07,
+      "loss": 1.4352,
+      "step": 43530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.402630495288578e-07,
+      "loss": 1.4281,
+      "step": 43540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.392353300805157e-07,
+      "loss": 1.3682,
+      "step": 43550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.382081127297972e-07,
+      "loss": 1.4762,
+      "step": 43560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.371813977872598e-07,
+      "loss": 1.3838,
+      "step": 43570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.361551855633105e-07,
+      "loss": 1.5052,
+      "step": 43580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.351294763682036e-07,
+      "loss": 1.4109,
+      "step": 43590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.34104270512039e-07,
+      "loss": 1.3524,
+      "step": 43600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.330795683047673e-07,
+      "loss": 1.4867,
+      "step": 43610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.32055370056186e-07,
+      "loss": 1.3677,
+      "step": 43620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.310316760759399e-07,
+      "loss": 1.4175,
+      "step": 43630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.300084866735217e-07,
+      "loss": 1.3602,
+      "step": 43640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.289858021582712e-07,
+      "loss": 1.4159,
+      "step": 43650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.279636228393761e-07,
+      "loss": 1.4522,
+      "step": 43660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.269419490258701e-07,
+      "loss": 1.3603,
+      "step": 43670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.259207810266373e-07,
+      "loss": 1.4617,
+      "step": 43680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.249001191504044e-07,
+      "loss": 1.3656,
+      "step": 43690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.238799637057477e-07,
+      "loss": 1.3925,
+      "step": 43700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.228603150010906e-07,
+      "loss": 1.4612,
+      "step": 43710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.218411733447018e-07,
+      "loss": 1.4122,
+      "step": 43720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.208225390446983e-07,
+      "loss": 1.4493,
+      "step": 43730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.198044124090427e-07,
+      "loss": 1.4066,
+      "step": 43740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.187867937455443e-07,
+      "loss": 1.3723,
+      "step": 43750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.17769683361859e-07,
+      "loss": 1.4678,
+      "step": 43760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.16753081565489e-07,
+      "loss": 1.3438,
+      "step": 43770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.157369886637826e-07,
+      "loss": 1.4309,
+      "step": 43780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.147214049639343e-07,
+      "loss": 1.3801,
+      "step": 43790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.13706330772985e-07,
+      "loss": 1.4299,
+      "step": 43800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.126917663978209e-07,
+      "loss": 1.4329,
+      "step": 43810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.116777121451747e-07,
+      "loss": 1.4207,
+      "step": 43820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.106641683216243e-07,
+      "loss": 1.4618,
+      "step": 43830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.096511352335937e-07,
+      "loss": 1.3634,
+      "step": 43840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.086386131873522e-07,
+      "loss": 1.3059,
+      "step": 43850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.076266024890154e-07,
+      "loss": 1.4621,
+      "step": 43860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.066151034445427e-07,
+      "loss": 1.4017,
+      "step": 43870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 8.056041163597406e-07,
+      "loss": 1.4613,
+      "step": 43880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.045936415402594e-07,
+      "loss": 1.4516,
+      "step": 43890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.035836792915958e-07,
+      "loss": 1.3934,
+      "step": 43900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.0257422991909e-07,
+      "loss": 1.5108,
+      "step": 43910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.015652937279289e-07,
+      "loss": 1.3861,
+      "step": 43920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 8.005568710231426e-07,
+      "loss": 1.4279,
+      "step": 43930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.995489621096072e-07,
+      "loss": 1.3594,
+      "step": 43940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.985415672920435e-07,
+      "loss": 1.3686,
+      "step": 43950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.975346868750142e-07,
+      "loss": 1.4706,
+      "step": 43960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.965283211629308e-07,
+      "loss": 1.3168,
+      "step": 43970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.955224704600462e-07,
+      "loss": 1.4944,
+      "step": 43980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.945171350704586e-07,
+      "loss": 1.3607,
+      "step": 43990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.935123152981103e-07,
+      "loss": 1.3581,
+      "step": 44000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.925080114467876e-07,
+      "loss": 1.5185,
+      "step": 44010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.915042238201204e-07,
+      "loss": 1.4608,
+      "step": 44020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.905009527215834e-07,
+      "loss": 1.5457,
+      "step": 44030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.894981984544958e-07,
+      "loss": 1.3025,
+      "step": 44040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.884959613220173e-07,
+      "loss": 1.406,
+      "step": 44050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.874942416271538e-07,
+      "loss": 1.4296,
+      "step": 44060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.864930396727558e-07,
+      "loss": 1.4352,
+      "step": 44070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.854923557615152e-07,
+      "loss": 1.4388,
+      "step": 44080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.844921901959679e-07,
+      "loss": 1.45,
+      "step": 44090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.834925432784932e-07,
+      "loss": 1.4168,
+      "step": 44100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.824934153113134e-07,
+      "loss": 1.6042,
+      "step": 44110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.814948065964942e-07,
+      "loss": 1.4511,
+      "step": 44120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.804967174359448e-07,
+      "loss": 1.4973,
+      "step": 44130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.794991481314151e-07,
+      "loss": 1.46,
+      "step": 44140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.785020989845005e-07,
+      "loss": 1.3558,
+      "step": 44150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.775055702966375e-07,
+      "loss": 1.4149,
+      "step": 44160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.765095623691055e-07,
+      "loss": 1.3658,
+      "step": 44170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.755140755030282e-07,
+      "loss": 1.4664,
+      "step": 44180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.745191099993693e-07,
+      "loss": 1.4183,
+      "step": 44190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.735246661589363e-07,
+      "loss": 1.3947,
+      "step": 44200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.725307442823785e-07,
+      "loss": 1.4779,
+      "step": 44210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.715373446701882e-07,
+      "loss": 1.3286,
+      "step": 44220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.705444676226972e-07,
+      "loss": 1.4523,
+      "step": 44230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.695521134400824e-07,
+      "loss": 1.4166,
+      "step": 44240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.685602824223615e-07,
+      "loss": 1.3976,
+      "step": 44250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.675689748693937e-07,
+      "loss": 1.492,
+      "step": 44260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.665781910808794e-07,
+      "loss": 1.397,
+      "step": 44270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.655879313563633e-07,
+      "loss": 1.4597,
+      "step": 44280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.645981959952286e-07,
+      "loss": 1.3953,
+      "step": 44290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.636089852967016e-07,
+      "loss": 1.4679,
+      "step": 44300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.626202995598503e-07,
+      "loss": 1.5314,
+      "step": 44310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.616321390835812e-07,
+      "loss": 1.3834,
+      "step": 44320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.606445041666455e-07,
+      "loss": 1.4295,
+      "step": 44330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.596573951076336e-07,
+      "loss": 1.4146,
+      "step": 44340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.586708122049777e-07,
+      "loss": 1.4344,
+      "step": 44350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.576847557569503e-07,
+      "loss": 1.3538,
+      "step": 44360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.566992260616651e-07,
+      "loss": 1.3795,
+      "step": 44370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.557142234170758e-07,
+      "loss": 1.3894,
+      "step": 44380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.547297481209789e-07,
+      "loss": 1.4595,
+      "step": 44390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.537458004710102e-07,
+      "loss": 1.3618,
+      "step": 44400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.527623807646442e-07,
+      "loss": 1.4913,
+      "step": 44410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.517794892991978e-07,
+      "loss": 1.4433,
+      "step": 44420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.507971263718283e-07,
+      "loss": 1.4595,
+      "step": 44430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.498152922795324e-07,
+      "loss": 1.3862,
+      "step": 44440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.48833987319147e-07,
+      "loss": 1.4069,
+      "step": 44450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.478532117873497e-07,
+      "loss": 1.4899,
+      "step": 44460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.46872965980657e-07,
+      "loss": 1.4848,
+      "step": 44470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.458932501954255e-07,
+      "loss": 1.425,
+      "step": 44480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.449140647278541e-07,
+      "loss": 1.3747,
+      "step": 44490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.439354098739767e-07,
+      "loss": 1.3919,
+      "step": 44500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.429572859296699e-07,
+      "loss": 1.4442,
+      "step": 44510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.419796931906495e-07,
+      "loss": 1.384,
+      "step": 44520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.410026319524702e-07,
+      "loss": 1.4439,
+      "step": 44530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.400261025105257e-07,
+      "loss": 1.4512,
+      "step": 44540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.390501051600498e-07,
+      "loss": 1.4866,
+      "step": 44550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.380746401961148e-07,
+      "loss": 1.4443,
+      "step": 44560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.370997079136322e-07,
+      "loss": 1.3787,
+      "step": 44570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.361253086073525e-07,
+      "loss": 1.4912,
+      "step": 44580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.351514425718653e-07,
+      "loss": 1.4162,
+      "step": 44590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.341781101015982e-07,
+      "loss": 1.3727,
+      "step": 44600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.332053114908186e-07,
+      "loss": 1.4233,
+      "step": 44610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.322330470336314e-07,
+      "loss": 1.4384,
+      "step": 44620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.312613170239807e-07,
+      "loss": 1.4226,
+      "step": 44630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.302901217556491e-07,
+      "loss": 1.3553,
+      "step": 44640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.29319461522257e-07,
+      "loss": 1.4091,
+      "step": 44650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.283493366172634e-07,
+      "loss": 1.5089,
+      "step": 44660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.27379747333965e-07,
+      "loss": 1.3004,
+      "step": 44670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.264106939654975e-07,
+      "loss": 1.3919,
+      "step": 44680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.254421768048339e-07,
+      "loss": 1.4854,
+      "step": 44690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.244741961447848e-07,
+      "loss": 1.4179,
+      "step": 44700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.235067522779996e-07,
+      "loss": 1.4814,
+      "step": 44710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.225398454969643e-07,
+      "loss": 1.3544,
+      "step": 44720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.215734760940032e-07,
+      "loss": 1.484,
+      "step": 44730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.206076443612781e-07,
+      "loss": 1.3278,
+      "step": 44740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.196423505907879e-07,
+      "loss": 1.5145,
+      "step": 44750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.1867759507437e-07,
+      "loss": 1.4726,
+      "step": 44760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.177133781036957e-07,
+      "loss": 1.4457,
+      "step": 44770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.167496999702786e-07,
+      "loss": 1.4518,
+      "step": 44780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.157865609654655e-07,
+      "loss": 1.4514,
+      "step": 44790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.148239613804414e-07,
+      "loss": 1.341,
+      "step": 44800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.138619015062287e-07,
+      "loss": 1.4158,
+      "step": 44810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.12900381633686e-07,
+      "loss": 1.3904,
+      "step": 44820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.119394020535087e-07,
+      "loss": 1.4922,
+      "step": 44830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.109789630562289e-07,
+      "loss": 1.4392,
+      "step": 44840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.100190649322164e-07,
+      "loss": 1.4046,
+      "step": 44850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.090597079716747e-07,
+      "loss": 1.4378,
+      "step": 44860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.081008924646451e-07,
+      "loss": 1.3184,
+      "step": 44870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.071426187010077e-07,
+      "loss": 1.5095,
+      "step": 44880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.061848869704754e-07,
+      "loss": 1.3906,
+      "step": 44890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.052276975625985e-07,
+      "loss": 1.4389,
+      "step": 44900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.042710507667632e-07,
+      "loss": 1.476,
+      "step": 44910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.033149468721923e-07,
+      "loss": 1.4107,
+      "step": 44920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.023593861679431e-07,
+      "loss": 1.5831,
+      "step": 44930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.014043689429109e-07,
+      "loss": 1.3764,
+      "step": 44940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 7.004498954858235e-07,
+      "loss": 1.4278,
+      "step": 44950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.994959660852468e-07,
+      "loss": 1.4638,
+      "step": 44960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.985425810295815e-07,
+      "loss": 1.382,
+      "step": 44970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.975897406070628e-07,
+      "loss": 1.4086,
+      "step": 44980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.966374451057642e-07,
+      "loss": 1.4421,
+      "step": 44990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.956856948135912e-07,
+      "loss": 1.4012,
+      "step": 45000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.947344900182859e-07,
+      "loss": 1.507,
+      "step": 45010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.937838310074249e-07,
+      "loss": 1.3381,
+      "step": 45020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.928337180684216e-07,
+      "loss": 1.4775,
+      "step": 45030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.918841514885207e-07,
+      "loss": 1.3837,
+      "step": 45040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.909351315548052e-07,
+      "loss": 1.4086,
+      "step": 45050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.899866585541908e-07,
+      "loss": 1.4743,
+      "step": 45060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.890387327734294e-07,
+      "loss": 1.373,
+      "step": 45070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.880913544991053e-07,
+      "loss": 1.4711,
+      "step": 45080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.871445240176403e-07,
+      "loss": 1.3384,
+      "step": 45090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.861982416152882e-07,
+      "loss": 1.369,
+      "step": 45100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.852525075781377e-07,
+      "loss": 1.4992,
+      "step": 45110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.843073221921123e-07,
+      "loss": 1.3798,
+      "step": 45120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.833626857429681e-07,
+      "loss": 1.5098,
+      "step": 45130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.824185985162968e-07,
+      "loss": 1.4003,
+      "step": 45140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.814750607975235e-07,
+      "loss": 1.4489,
+      "step": 45150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.80532072871907e-07,
+      "loss": 1.428,
+      "step": 45160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.795896350245401e-07,
+      "loss": 1.4407,
+      "step": 45170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.786477475403497e-07,
+      "loss": 1.4536,
+      "step": 45180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.777064107040948e-07,
+      "loss": 1.3926,
+      "step": 45190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.767656248003707e-07,
+      "loss": 1.3328,
+      "step": 45200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.758253901136041e-07,
+      "loss": 1.5295,
+      "step": 45210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.74885706928054e-07,
+      "loss": 1.3202,
+      "step": 45220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.739465755278149e-07,
+      "loss": 1.4721,
+      "step": 45230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.730079961968137e-07,
+      "loss": 1.3983,
+      "step": 45240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.720699692188099e-07,
+      "loss": 1.3567,
+      "step": 45250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.71132494877397e-07,
+      "loss": 1.4364,
+      "step": 45260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.701955734560004e-07,
+      "loss": 1.4397,
+      "step": 45270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.692592052378791e-07,
+      "loss": 1.3963,
+      "step": 45280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.683233905061234e-07,
+      "loss": 1.4233,
+      "step": 45290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.673881295436601e-07,
+      "loss": 1.3864,
+      "step": 45300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.664534226332436e-07,
+      "loss": 1.4831,
+      "step": 45310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.655192700574634e-07,
+      "loss": 1.3305,
+      "step": 45320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.645856720987415e-07,
+      "loss": 1.5085,
+      "step": 45330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.636526290393317e-07,
+      "loss": 1.4885,
+      "step": 45340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.627201411613199e-07,
+      "loss": 1.387,
+      "step": 45350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.617882087466246e-07,
+      "loss": 1.5112,
+      "step": 45360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.608568320769961e-07,
+      "loss": 1.3828,
+      "step": 45370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.599260114340173e-07,
+      "loss": 1.461,
+      "step": 45380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.589957470991016e-07,
+      "loss": 1.3601,
+      "step": 45390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.580660393534952e-07,
+      "loss": 1.4027,
+      "step": 45400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.571368884782764e-07,
+      "loss": 1.4367,
+      "step": 45410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.562082947543541e-07,
+      "loss": 1.3989,
+      "step": 45420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.552802584624696e-07,
+      "loss": 1.5346,
+      "step": 45430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.543527798831953e-07,
+      "loss": 1.3756,
+      "step": 45440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.534258592969348e-07,
+      "loss": 1.3873,
+      "step": 45450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.524994969839235e-07,
+      "loss": 1.4647,
+      "step": 45460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.515736932242278e-07,
+      "loss": 1.364,
+      "step": 45470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.506484482977452e-07,
+      "loss": 1.4448,
+      "step": 45480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.497237624842037e-07,
+      "loss": 1.4077,
+      "step": 45490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.487996360631634e-07,
+      "loss": 1.3907,
+      "step": 45500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.478760693140146e-07,
+      "loss": 1.529,
+      "step": 45510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.469530625159784e-07,
+      "loss": 1.3377,
+      "step": 45520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.460306159481067e-07,
+      "loss": 1.4126,
+      "step": 45530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.451087298892819e-07,
+      "loss": 1.3627,
+      "step": 45540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.441874046182172e-07,
+      "loss": 1.3712,
+      "step": 45550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.43266640413456e-07,
+      "loss": 1.4268,
+      "step": 45560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.423464375533731e-07,
+      "loss": 1.4182,
+      "step": 45570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.414267963161702e-07,
+      "loss": 1.4177,
+      "step": 45580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.405077169798843e-07,
+      "loss": 1.5034,
+      "step": 45590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.395891998223789e-07,
+      "loss": 1.3434,
+      "step": 45600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.386712451213484e-07,
+      "loss": 1.449,
+      "step": 45610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.377538531543173e-07,
+      "loss": 1.4818,
+      "step": 45620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.368370241986405e-07,
+      "loss": 1.4373,
+      "step": 45630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.359207585315017e-07,
+      "loss": 1.4156,
+      "step": 45640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.350050564299145e-07,
+      "loss": 1.4255,
+      "step": 45650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.340899181707239e-07,
+      "loss": 1.4507,
+      "step": 45660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.331753440306004e-07,
+      "loss": 1.4034,
+      "step": 45670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.322613342860473e-07,
+      "loss": 1.506,
+      "step": 45680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.313478892133975e-07,
+      "loss": 1.41,
+      "step": 45690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.304350090888117e-07,
+      "loss": 1.3902,
+      "step": 45700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.295226941882798e-07,
+      "loss": 1.4168,
+      "step": 45710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.286109447876215e-07,
+      "loss": 1.4023,
+      "step": 45720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.27699761162485e-07,
+      "loss": 1.4185,
+      "step": 45730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.267891435883478e-07,
+      "loss": 1.4546,
+      "step": 45740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.258790923405161e-07,
+      "loss": 1.4437,
+      "step": 45750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.249696076941258e-07,
+      "loss": 1.4774,
+      "step": 45760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.240606899241394e-07,
+      "loss": 1.3337,
+      "step": 45770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.231523393053487e-07,
+      "loss": 1.4515,
+      "step": 45780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.222445561123766e-07,
+      "loss": 1.384,
+      "step": 45790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.213373406196715e-07,
+      "loss": 1.3558,
+      "step": 45800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.204306931015112e-07,
+      "loss": 1.479,
+      "step": 45810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.195246138320016e-07,
+      "loss": 1.4051,
+      "step": 45820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.186191030850769e-07,
+      "loss": 1.4478,
+      "step": 45830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.177141611344991e-07,
+      "loss": 1.439,
+      "step": 45840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.168097882538601e-07,
+      "loss": 1.4133,
+      "step": 45850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.15905984716576e-07,
+      "loss": 1.5585,
+      "step": 45860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.150027507958941e-07,
+      "loss": 1.392,
+      "step": 45870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.141000867648886e-07,
+      "loss": 1.4404,
+      "step": 45880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.131979928964599e-07,
+      "loss": 1.3923,
+      "step": 45890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.122964694633387e-07,
+      "loss": 1.369,
+      "step": 45900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.113955167380817e-07,
+      "loss": 1.432,
+      "step": 45910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.104951349930729e-07,
+      "loss": 1.3847,
+      "step": 45920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.09595324500524e-07,
+      "loss": 1.4868,
+      "step": 45930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.086960855324747e-07,
+      "loss": 1.3626,
+      "step": 45940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.077974183607896e-07,
+      "loss": 1.3979,
+      "step": 45950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.068993232571626e-07,
+      "loss": 1.4884,
+      "step": 45960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.060018004931148e-07,
+      "loss": 1.4076,
+      "step": 45970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.051048503399928e-07,
+      "loss": 1.5034,
+      "step": 45980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.042084730689701e-07,
+      "loss": 1.403,
+      "step": 45990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.033126689510497e-07,
+      "loss": 1.3481,
+      "step": 46000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.02417438257058e-07,
+      "loss": 1.425,
+      "step": 46010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.015227812576497e-07,
+      "loss": 1.3833,
+      "step": 46020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.006286982233064e-07,
+      "loss": 1.4247,
+      "step": 46030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.997351894243342e-07,
+      "loss": 1.4463,
+      "step": 46040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.988422551308671e-07,
+      "loss": 1.4348,
+      "step": 46050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.979498956128657e-07,
+      "loss": 1.4147,
+      "step": 46060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.97058111140116e-07,
+      "loss": 1.4589,
+      "step": 46070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.961669019822308e-07,
+      "loss": 1.4999,
+      "step": 46080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.952762684086482e-07,
+      "loss": 1.3817,
+      "step": 46090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.943862106886325e-07,
+      "loss": 1.3809,
+      "step": 46100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.934967290912749e-07,
+      "loss": 1.4607,
+      "step": 46110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.926078238854923e-07,
+      "loss": 1.391,
+      "step": 46120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.917194953400252e-07,
+      "loss": 1.4499,
+      "step": 46130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.908317437234415e-07,
+      "loss": 1.4245,
+      "step": 46140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.899445693041348e-07,
+      "loss": 1.4183,
+      "step": 46150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.890579723503237e-07,
+      "loss": 1.5127,
+      "step": 46160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.881719531300522e-07,
+      "loss": 1.3997,
+      "step": 46170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.872865119111898e-07,
+      "loss": 1.4979,
+      "step": 46180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.864016489614316e-07,
+      "loss": 1.385,
+      "step": 46190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.855173645482964e-07,
+      "loss": 1.3914,
+      "step": 46200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.846336589391314e-07,
+      "loss": 1.4494,
+      "step": 46210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.837505324011041e-07,
+      "loss": 1.3949,
+      "step": 46220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 5.82867985201211e-07,
+      "loss": 1.44,
+      "step": 46230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.819860176062711e-07,
+      "loss": 1.4328,
+      "step": 46240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.811046298829292e-07,
+      "loss": 1.373,
+      "step": 46250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.802238222976548e-07,
+      "loss": 1.5391,
+      "step": 46260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.793435951167411e-07,
+      "loss": 1.3184,
+      "step": 46270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.784639486063074e-07,
+      "loss": 1.4579,
+      "step": 46280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.775848830322952e-07,
+      "loss": 1.3728,
+      "step": 46290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.767063986604729e-07,
+      "loss": 1.4028,
+      "step": 46300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.758284957564316e-07,
+      "loss": 1.4234,
+      "step": 46310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.749511745855865e-07,
+      "loss": 1.3992,
+      "step": 46320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.740744354131781e-07,
+      "loss": 1.4905,
+      "step": 46330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.731982785042697e-07,
+      "loss": 1.4591,
+      "step": 46340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.723227041237492e-07,
+      "loss": 1.2822,
+      "step": 46350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.714477125363286e-07,
+      "loss": 1.4455,
+      "step": 46360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.705733040065431e-07,
+      "loss": 1.3432,
+      "step": 46370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.696994787987517e-07,
+      "loss": 1.4887,
+      "step": 46380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.688262371771377e-07,
+      "loss": 1.3402,
+      "step": 46390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.679535794057076e-07,
+      "loss": 1.4357,
+      "step": 46400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.670815057482909e-07,
+      "loss": 1.4191,
+      "step": 46410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.66210016468541e-07,
+      "loss": 1.4331,
+      "step": 46420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.653391118299347e-07,
+      "loss": 1.502,
+      "step": 46430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.644687920957717e-07,
+      "loss": 1.3732,
+      "step": 46440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.635990575291752e-07,
+      "loss": 1.4543,
+      "step": 46450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.627299083930915e-07,
+      "loss": 1.4181,
+      "step": 46460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.618613449502897e-07,
+      "loss": 1.3713,
+      "step": 46470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.609933674633619e-07,
+      "loss": 1.4841,
+      "step": 46480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.60125976194722e-07,
+      "loss": 1.4045,
+      "step": 46490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.59259171406609e-07,
+      "loss": 1.4651,
+      "step": 46500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.583929533610827e-07,
+      "loss": 1.465,
+      "step": 46510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.575273223200265e-07,
+      "loss": 1.3952,
+      "step": 46520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.566622785451453e-07,
+      "loss": 1.5258,
+      "step": 46530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.557978222979676e-07,
+      "loss": 1.4449,
+      "step": 46540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.549339538398435e-07,
+      "loss": 1.4158,
+      "step": 46550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.54070673431946e-07,
+      "loss": 1.3901,
+      "step": 46560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.5320798133527e-07,
+      "loss": 1.4235,
+      "step": 46570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.523458778106317e-07,
+      "loss": 1.4802,
+      "step": 46580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.5148436311867e-07,
+      "loss": 1.3577,
+      "step": 46590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.506234375198472e-07,
+      "loss": 1.4166,
+      "step": 46600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.497631012744453e-07,
+      "loss": 1.5001,
+      "step": 46610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.489033546425698e-07,
+      "loss": 1.4434,
+      "step": 46620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.480441978841463e-07,
+      "loss": 1.4847,
+      "step": 46630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.471856312589238e-07,
+      "loss": 1.3856,
+      "step": 46640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.463276550264715e-07,
+      "loss": 1.3972,
+      "step": 46650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.454702694461813e-07,
+      "loss": 1.4323,
+      "step": 46660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.44613474777265e-07,
+      "loss": 1.3466,
+      "step": 46670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.437572712787567e-07,
+      "loss": 1.5951,
+      "step": 46680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.429016592095121e-07,
+      "loss": 1.3835,
+      "step": 46690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.420466388282069e-07,
+      "loss": 1.4423,
+      "step": 46700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.411922103933403e-07,
+      "loss": 1.3986,
+      "step": 46710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.403383741632301e-07,
+      "loss": 1.3853,
+      "step": 46720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.394851303960158e-07,
+      "loss": 1.4095,
+      "step": 46730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.386324793496578e-07,
+      "loss": 1.4252,
+      "step": 46740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.377804212819385e-07,
+      "loss": 1.4079,
+      "step": 46750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.36928956450458e-07,
+      "loss": 1.5026,
+      "step": 46760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.360780851126402e-07,
+      "loss": 1.402,
+      "step": 46770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.352278075257278e-07,
+      "loss": 1.3952,
+      "step": 46780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.343781239467851e-07,
+      "loss": 1.424,
+      "step": 46790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.335290346326952e-07,
+      "loss": 1.4071,
+      "step": 46800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.326805398401639e-07,
+      "loss": 1.5052,
+      "step": 46810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.318326398257154e-07,
+      "loss": 1.4137,
+      "step": 46820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.309853348456948e-07,
+      "loss": 1.5332,
+      "step": 46830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.301386251562674e-07,
+      "loss": 1.3582,
+      "step": 46840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.29292511013417e-07,
+      "loss": 1.3926,
+      "step": 46850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.284469926729494e-07,
+      "loss": 1.4917,
+      "step": 46860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.276020703904894e-07,
+      "loss": 1.3938,
+      "step": 46870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.267577444214819e-07,
+      "loss": 1.4957,
+      "step": 46880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.259140150211908e-07,
+      "loss": 1.3731,
+      "step": 46890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.250708824447003e-07,
+      "loss": 1.3806,
+      "step": 46900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.242283469469137e-07,
+      "loss": 1.4502,
+      "step": 46910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.23386408782555e-07,
+      "loss": 1.4418,
+      "step": 46920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.225450682061666e-07,
+      "loss": 1.4983,
+      "step": 46930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.217043254721094e-07,
+      "loss": 1.3657,
+      "step": 46940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.208641808345647e-07,
+      "loss": 1.2894,
+      "step": 46950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.200246345475329e-07,
+      "loss": 1.5009,
+      "step": 46960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.191856868648335e-07,
+      "loss": 1.4395,
+      "step": 46970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.183473380401046e-07,
+      "loss": 1.5449,
+      "step": 46980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.175095883268036e-07,
+      "loss": 1.4384,
+      "step": 46990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.166724379782067e-07,
+      "loss": 1.4349,
+      "step": 47000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.158358872474084e-07,
+      "loss": 1.5114,
+      "step": 47010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.149999363873243e-07,
+      "loss": 1.4394,
+      "step": 47020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.141645856506844e-07,
+      "loss": 1.4441,
+      "step": 47030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.13329835290041e-07,
+      "loss": 1.4333,
+      "step": 47040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.124956855577626e-07,
+      "loss": 1.4746,
+      "step": 47050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.116621367060376e-07,
+      "loss": 1.3608,
+      "step": 47060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.108291889868719e-07,
+      "loss": 1.3234,
+      "step": 47070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.099968426520902e-07,
+      "loss": 1.4026,
+      "step": 47080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.091650979533347e-07,
+      "loss": 1.4106,
+      "step": 47090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.083339551420663e-07,
+      "loss": 1.3977,
+      "step": 47100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.075034144695637e-07,
+      "loss": 1.4525,
+      "step": 47110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.066734761869235e-07,
+      "loss": 1.3644,
+      "step": 47120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.058441405450603e-07,
+      "loss": 1.5661,
+      "step": 47130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.050154077947067e-07,
+      "loss": 1.37,
+      "step": 47140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.041872781864127e-07,
+      "loss": 1.4565,
+      "step": 47150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.033597519705457e-07,
+      "loss": 1.4616,
+      "step": 47160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.025328293972914e-07,
+      "loss": 1.4157,
+      "step": 47170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.017065107166525e-07,
+      "loss": 1.429,
+      "step": 47180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.008807961784489e-07,
+      "loss": 1.364,
+      "step": 47190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.000556860323191e-07,
+      "loss": 1.4365,
+      "step": 47200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.992311805277169e-07,
+      "loss": 1.4689,
+      "step": 47210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.984072799139153e-07,
+      "loss": 1.3805,
+      "step": 47220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.97583984440003e-07,
+      "loss": 1.4502,
+      "step": 47230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.967612943548863e-07,
+      "loss": 1.3264,
+      "step": 47240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.959392099072885e-07,
+      "loss": 1.4061,
+      "step": 47250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.951177313457497e-07,
+      "loss": 1.5018,
+      "step": 47260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.942968589186267e-07,
+      "loss": 1.3754,
+      "step": 47270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.934765928740933e-07,
+      "loss": 1.4551,
+      "step": 47280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.926569334601408e-07,
+      "loss": 1.4996,
+      "step": 47290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.918378809245738e-07,
+      "loss": 1.3972,
+      "step": 47300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.910194355150178e-07,
+      "loss": 1.389,
+      "step": 47310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.902015974789121e-07,
+      "loss": 1.4789,
+      "step": 47320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.893843670635131e-07,
+      "loss": 1.445,
+      "step": 47330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.885677445158934e-07,
+      "loss": 1.3839,
+      "step": 47340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.877517300829418e-07,
+      "loss": 1.3691,
+      "step": 47350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.869363240113634e-07,
+      "loss": 1.4249,
+      "step": 47360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.861215265476787e-07,
+      "loss": 1.4543,
+      "step": 47370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.853073379382262e-07,
+      "loss": 1.4159,
+      "step": 47380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.844937584291571e-07,
+      "loss": 1.4114,
+      "step": 47390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.836807882664401e-07,
+      "loss": 1.4541,
+      "step": 47400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.828684276958615e-07,
+      "loss": 1.4846,
+      "step": 47410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.820566769630209e-07,
+      "loss": 1.3522,
+      "step": 47420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.81245536313334e-07,
+      "loss": 1.417,
+      "step": 47430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.804350059920324e-07,
+      "loss": 1.3837,
+      "step": 47440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.796250862441628e-07,
+      "loss": 1.3392,
+      "step": 47450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.788157773145883e-07,
+      "loss": 1.4441,
+      "step": 47460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.780070794479863e-07,
+      "loss": 1.4156,
+      "step": 47470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.771989928888493e-07,
+      "loss": 1.4842,
+      "step": 47480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.763915178814857e-07,
+      "loss": 1.4558,
+      "step": 47490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.755846546700182e-07,
+      "loss": 1.3453,
+      "step": 47500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.747784034983863e-07,
+      "loss": 1.4716,
+      "step": 47510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.739727646103426e-07,
+      "loss": 1.4145,
+      "step": 47520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.731677382494554e-07,
+      "loss": 1.4437,
+      "step": 47530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.723633246591078e-07,
+      "loss": 1.4031,
+      "step": 47540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.7155952408249687e-07,
+      "loss": 1.3838,
+      "step": 47550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.707563367626364e-07,
+      "loss": 1.4856,
+      "step": 47560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6995376294235146e-07,
+      "loss": 1.4267,
+      "step": 47570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6915180286428443e-07,
+      "loss": 1.4274,
+      "step": 47580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.683504567708913e-07,
+      "loss": 1.3999,
+      "step": 47590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.675497249044425e-07,
+      "loss": 1.3568,
+      "step": 47600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6674960750702156e-07,
+      "loss": 1.4897,
+      "step": 47610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6595010482052894e-07,
+      "loss": 1.4509,
+      "step": 47620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6515121708667675e-07,
+      "loss": 1.5275,
+      "step": 47630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.643529445469924e-07,
+      "loss": 1.3559,
+      "step": 47640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6355528744281755e-07,
+      "loss": 1.4296,
+      "step": 47650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.627582460153057e-07,
+      "loss": 1.4596,
+      "step": 47660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.6196182050542633e-07,
+      "loss": 1.369,
+      "step": 47670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.611660111539626e-07,
+      "loss": 1.5022,
+      "step": 47680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.603708182015104e-07,
+      "loss": 1.36,
+      "step": 47690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5957624188848e-07,
+      "loss": 1.3819,
+      "step": 47700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5878228245509413e-07,
+      "loss": 1.4813,
+      "step": 47710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5798894014139146e-07,
+      "loss": 1.4783,
+      "step": 47720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5719621518722194e-07,
+      "loss": 1.42,
+      "step": 47730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.564041078322498e-07,
+      "loss": 1.3459,
+      "step": 47740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5561261831595124e-07,
+      "loss": 1.3931,
+      "step": 47750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.548217468776167e-07,
+      "loss": 1.4713,
+      "step": 47760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5403149375635057e-07,
+      "loss": 1.3481,
+      "step": 47770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.532418591910687e-07,
+      "loss": 1.3738,
+      "step": 47780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.524528434205008e-07,
+      "loss": 1.4214,
+      "step": 47790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.5166444668318987e-07,
+      "loss": 1.3967,
+      "step": 47800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.508766692174904e-07,
+      "loss": 1.4636,
+      "step": 47810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.500895112615708e-07,
+      "loss": 1.3546,
+      "step": 47820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.493029730534129e-07,
+      "loss": 1.429,
+      "step": 47830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4851705483080904e-07,
+      "loss": 1.3649,
+      "step": 47840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4773175683136524e-07,
+      "loss": 1.4655,
+      "step": 47850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4694707929250044e-07,
+      "loss": 1.492,
+      "step": 47860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4616302245144545e-07,
+      "loss": 1.3774,
+      "step": 47870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4537958654524317e-07,
+      "loss": 1.4117,
+      "step": 47880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.445967718107497e-07,
+      "loss": 1.3769,
+      "step": 47890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.438145784846323e-07,
+      "loss": 1.4328,
+      "step": 47900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.4303300680337143e-07,
+      "loss": 1.5382,
+      "step": 47910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.422520570032582e-07,
+      "loss": 1.3829,
+      "step": 47920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.414717293203971e-07,
+      "loss": 1.4565,
+      "step": 47930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.406920239907039e-07,
+      "loss": 1.3447,
+      "step": 47940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3991294124990617e-07,
+      "loss": 1.3774,
+      "step": 47950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.391344813335435e-07,
+      "loss": 1.5379,
+      "step": 47960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3835664447696686e-07,
+      "loss": 1.4295,
+      "step": 47970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3757943091533914e-07,
+      "loss": 1.4384,
+      "step": 47980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3680284088363443e-07,
+      "loss": 1.3165,
+      "step": 47990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.3602687461663904e-07,
+      "loss": 1.3421,
+      "step": 48000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.352515323489498e-07,
+      "loss": 1.4579,
+      "step": 48010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3447681431497523e-07,
+      "loss": 1.4271,
+      "step": 48020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3370272074893583e-07,
+      "loss": 1.4708,
+      "step": 48030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.329292518848621e-07,
+      "loss": 1.4231,
+      "step": 48040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3215640795659625e-07,
+      "loss": 1.4906,
+      "step": 48050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.3138418919779187e-07,
+      "loss": 1.4827,
+      "step": 48060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.30612595841913e-07,
+      "loss": 1.3629,
+      "step": 48070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.298416281222348e-07,
+      "loss": 1.4559,
+      "step": 48080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2907128627184355e-07,
+      "loss": 1.3742,
+      "step": 48090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2830157052363595e-07,
+      "loss": 1.4134,
+      "step": 48100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.275324811103196e-07,
+      "loss": 1.4899,
+      "step": 48110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2676401826441225e-07,
+      "loss": 1.3598,
+      "step": 48120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2599618221824317e-07,
+      "loss": 1.4786,
+      "step": 48130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2522897320395133e-07,
+      "loss": 1.4267,
+      "step": 48140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.244623914534868e-07,
+      "loss": 1.4249,
+      "step": 48150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.236964371986088e-07,
+      "loss": 1.4635,
+      "step": 48160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2293111067088863e-07,
+      "loss": 1.3405,
+      "step": 48170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.2216641210170603e-07,
+      "loss": 1.4927,
+      "step": 48180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.214023417222521e-07,
+      "loss": 1.3973,
+      "step": 48190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.20638899763528e-07,
+      "loss": 1.398,
+      "step": 48200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.198760864563431e-07,
+      "loss": 1.5177,
+      "step": 48210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1911390203131954e-07,
+      "loss": 1.4317,
+      "step": 48220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.183523467188874e-07,
+      "loss": 1.4663,
+      "step": 48230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.17591420749287e-07,
+      "loss": 1.3792,
+      "step": 48240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1683112435256884e-07,
+      "loss": 1.3854,
+      "step": 48250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1607145775859226e-07,
+      "loss": 1.478,
+      "step": 48260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1531242119702693e-07,
+      "loss": 1.374,
+      "step": 48270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1455401489735175e-07,
+      "loss": 1.4921,
+      "step": 48280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.137962390888553e-07,
+      "loss": 1.3602,
+      "step": 48290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.130390940006346e-07,
+      "loss": 1.4066,
+      "step": 48300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.122825798615962e-07,
+      "loss": 1.5081,
+      "step": 48310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1152669690045834e-07,
+      "loss": 1.2979,
+      "step": 48320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.107714453457451e-07,
+      "loss": 1.4542,
+      "step": 48330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.1001682542579164e-07,
+      "loss": 1.3373,
+      "step": 48340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.092628373687413e-07,
+      "loss": 1.4668,
+      "step": 48350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.085094814025467e-07,
+      "loss": 1.4773,
+      "step": 48360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.077567577549693e-07,
+      "loss": 1.4363,
+      "step": 48370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.070046666535801e-07,
+      "loss": 1.4333,
+      "step": 48380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.06253208325757e-07,
+      "loss": 1.4381,
+      "step": 48390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0550238299868846e-07,
+      "loss": 1.388,
+      "step": 48400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0475219089937066e-07,
+      "loss": 1.3736,
+      "step": 48410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.040026322546084e-07,
+      "loss": 1.4067,
+      "step": 48420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.032537072910156e-07,
+      "loss": 1.446,
+      "step": 48430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0250541623501423e-07,
+      "loss": 1.4021,
+      "step": 48440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0175775931283414e-07,
+      "loss": 1.4056,
+      "step": 48450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.0101073675051395e-07,
+      "loss": 1.5053,
+      "step": 48460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.002643487739011e-07,
+      "loss": 1.4107,
+      "step": 48470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.995185956086489e-07,
+      "loss": 1.4602,
+      "step": 48480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.98773477480221e-07,
+      "loss": 1.387,
+      "step": 48490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.980289946138888e-07,
+      "loss": 1.4034,
+      "step": 48500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9728514723473083e-07,
+      "loss": 1.4364,
+      "step": 48510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.965419355676331e-07,
+      "loss": 1.3768,
+      "step": 48520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.957993598372917e-07,
+      "loss": 1.4204,
+      "step": 48530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.950574202682081e-07,
+      "loss": 1.3817,
+      "step": 48540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9431611708469265e-07,
+      "loss": 1.4582,
+      "step": 48550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.93575450510863e-07,
+      "loss": 1.5165,
+      "step": 48560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.928354207706436e-07,
+      "loss": 1.4593,
+      "step": 48570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9209602808776724e-07,
+      "loss": 1.4847,
+      "step": 48580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.913572726857742e-07,
+      "loss": 1.3678,
+      "step": 48590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.906191547880114e-07,
+      "loss": 1.4314,
+      "step": 48600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.89881674617634e-07,
+      "loss": 1.523,
+      "step": 48610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8914483239760326e-07,
+      "loss": 1.4145,
+      "step": 48620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.884086283506877e-07,
+      "loss": 1.4412,
+      "step": 48630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.876730626994646e-07,
+      "loss": 1.4413,
+      "step": 48640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8693813566631657e-07,
+      "loss": 1.4156,
+      "step": 48650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8620384747343253e-07,
+      "loss": 1.4558,
+      "step": 48660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8547019834280996e-07,
+      "loss": 1.4185,
+      "step": 48670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8473718849625226e-07,
+      "loss": 1.4738,
+      "step": 48680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.840048181553696e-07,
+      "loss": 1.4066,
+      "step": 48690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.832730875415791e-07,
+      "loss": 1.38,
+      "step": 48700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8254199687610404e-07,
+      "loss": 1.5064,
+      "step": 48710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.818115463799746e-07,
+      "loss": 1.4435,
+      "step": 48720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.810817362740268e-07,
+      "loss": 1.4811,
+      "step": 48730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.8035256677890517e-07,
+      "loss": 1.3647,
+      "step": 48740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7962403811505703e-07,
+      "loss": 1.358,
+      "step": 48750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.788961505027389e-07,
+      "loss": 1.4616,
+      "step": 48760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.781689041620118e-07,
+      "loss": 1.3302,
+      "step": 48770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.774422993127441e-07,
+      "loss": 1.511,
+      "step": 48780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.767163361746093e-07,
+      "loss": 1.4006,
+      "step": 48790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7599101496708726e-07,
+      "loss": 1.3808,
+      "step": 48800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.752663359094641e-07,
+      "loss": 1.4536,
+      "step": 48810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.745422992208311e-07,
+      "loss": 1.2969,
+      "step": 48820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.738189051200855e-07,
+      "loss": 1.4457,
+      "step": 48830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.730961538259312e-07,
+      "loss": 1.3842,
+      "step": 48840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.723740455568761e-07,
+      "loss": 1.467,
+      "step": 48850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7165258053123515e-07,
+      "loss": 1.496,
+      "step": 48860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.7093175896712827e-07,
+      "loss": 1.4224,
+      "step": 48870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.702115810824805e-07,
+      "loss": 1.4317,
+      "step": 48880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.694920470950231e-07,
+      "loss": 1.3871,
+      "step": 48890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.687731572222919e-07,
+      "loss": 1.3798,
+      "step": 48900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6805491168162816e-07,
+      "loss": 1.486,
+      "step": 48910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6733731069017895e-07,
+      "loss": 1.4072,
+      "step": 48920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.666203544648955e-07,
+      "loss": 1.4905,
+      "step": 48930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.65904043222535e-07,
+      "loss": 1.3719,
+      "step": 48940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.651883771796588e-07,
+      "loss": 1.3854,
+      "step": 48950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.6447335655263397e-07,
+      "loss": 1.4479,
+      "step": 48960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.63758981557632e-07,
+      "loss": 1.3247,
+      "step": 48970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.630452524106293e-07,
+      "loss": 1.4758,
+      "step": 48980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.623321693274068e-07,
+      "loss": 1.4335,
+      "step": 48990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.616197325235507e-07,
+      "loss": 1.3498,
+      "step": 49000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.609079422144521e-07,
+      "loss": 1.4723,
+      "step": 49010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.601967986153038e-07,
+      "loss": 1.3277,
+      "step": 49020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.594863019411071e-07,
+      "loss": 1.4129,
+      "step": 49030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.587764524066656e-07,
+      "loss": 1.3417,
+      "step": 49040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5806725022658725e-07,
+      "loss": 1.4463,
+      "step": 49050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.573586956152847e-07,
+      "loss": 1.486,
+      "step": 49060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5665078878697467e-07,
+      "loss": 1.3549,
+      "step": 49070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5594352995567793e-07,
+      "loss": 1.4583,
+      "step": 49080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.552369193352195e-07,
+      "loss": 1.3435,
+      "step": 49090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5453095713922915e-07,
+      "loss": 1.3667,
+      "step": 49100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.538256435811385e-07,
+      "loss": 1.3847,
+      "step": 49110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5312097887418435e-07,
+      "loss": 1.4047,
+      "step": 49120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.5241696323140895e-07,
+      "loss": 1.4031,
+      "step": 49130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.517135968656557e-07,
+      "loss": 1.4219,
+      "step": 49140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.510108799895731e-07,
+      "loss": 1.3217,
+      "step": 49150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.503088128156129e-07,
+      "loss": 1.5297,
+      "step": 49160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.496073955560306e-07,
+      "loss": 1.3137,
+      "step": 49170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.489066284228851e-07,
+      "loss": 1.4849,
+      "step": 49180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4820651162803895e-07,
+      "loss": 1.4364,
+      "step": 49190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4750704538315694e-07,
+      "loss": 1.4259,
+      "step": 49200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4680822989970913e-07,
+      "loss": 1.4434,
+      "step": 49210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4611006538896705e-07,
+      "loss": 1.3752,
+      "step": 49220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4541255206200637e-07,
+      "loss": 1.5033,
+      "step": 49230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4471569012970633e-07,
+      "loss": 1.428,
+      "step": 49240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.440194798027485e-07,
+      "loss": 1.3749,
+      "step": 49250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.433239212916173e-07,
+      "loss": 1.4102,
+      "step": 49260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4262901480660027e-07,
+      "loss": 1.473,
+      "step": 49270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.419347605577888e-07,
+      "loss": 1.3931,
+      "step": 49280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.4124115875507484e-07,
+      "loss": 1.3659,
+      "step": 49290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.405482096081553e-07,
+      "loss": 1.4061,
+      "step": 49300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3985591332652837e-07,
+      "loss": 1.4629,
+      "step": 49310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.391642701194961e-07,
+      "loss": 1.4228,
+      "step": 49320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3847328019616126e-07,
+      "loss": 1.5359,
+      "step": 49330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3778294376543173e-07,
+      "loss": 1.3813,
+      "step": 49340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.370932610360159e-07,
+      "loss": 1.4292,
+      "step": 49350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.364042322164246e-07,
+      "loss": 1.4853,
+      "step": 49360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3571585751497215e-07,
+      "loss": 1.4138,
+      "step": 49370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3502813713977316e-07,
+      "loss": 1.4995,
+      "step": 49380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3434107129874633e-07,
+      "loss": 1.4163,
+      "step": 49390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.336546601996113e-07,
+      "loss": 1.3349,
+      "step": 49400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.329689040498904e-07,
+      "loss": 1.4719,
+      "step": 49410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3228380305690784e-07,
+      "loss": 1.4396,
+      "step": 49420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.31599357427789e-07,
+      "loss": 1.5106,
+      "step": 49430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.309155673694625e-07,
+      "loss": 1.4532,
+      "step": 49440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.3023243308865804e-07,
+      "loss": 1.3263,
+      "step": 49450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2954995479190733e-07,
+      "loss": 1.4274,
+      "step": 49460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.288681326855425e-07,
+      "loss": 1.4345,
+      "step": 49470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.281869669756987e-07,
+      "loss": 1.5218,
+      "step": 49480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.275064578683124e-07,
+      "loss": 1.43,
+      "step": 49490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.268266055691208e-07,
+      "loss": 1.3889,
+      "step": 49500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.261474102836637e-07,
+      "loss": 1.4529,
+      "step": 49510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2546887221728163e-07,
+      "loss": 1.3747,
+      "step": 49520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2479099157511603e-07,
+      "loss": 1.4347,
+      "step": 49530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2411376856210956e-07,
+      "loss": 1.4163,
+      "step": 49540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2343720338300856e-07,
+      "loss": 1.3831,
+      "step": 49550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2276129624235625e-07,
+      "loss": 1.4793,
+      "step": 49560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2208604734450005e-07,
+      "loss": 1.3654,
+      "step": 49570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.214114568935872e-07,
+      "loss": 1.4996,
+      "step": 49580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.20737525093566e-07,
+      "loss": 1.4277,
+      "step": 49590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.2006425214818586e-07,
+      "loss": 1.3792,
+      "step": 49600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1939163826099686e-07,
+      "loss": 1.4179,
+      "step": 49610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1871968363534983e-07,
+      "loss": 1.3383,
+      "step": 49620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1804838847439594e-07,
+      "loss": 1.4853,
+      "step": 49630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.173777529810876e-07,
+      "loss": 1.3882,
+      "step": 49640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1670777735817727e-07,
+      "loss": 1.4197,
+      "step": 49650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.160384618082185e-07,
+      "loss": 1.4986,
+      "step": 49660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.153698065335647e-07,
+      "loss": 1.4147,
+      "step": 49670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.147018117363698e-07,
+      "loss": 1.5588,
+      "step": 49680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.14034477618588e-07,
+      "loss": 1.4699,
+      "step": 49690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1336780438197436e-07,
+      "loss": 1.4212,
+      "step": 49700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1270179222808313e-07,
+      "loss": 1.4059,
+      "step": 49710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1203644135826967e-07,
+      "loss": 1.4489,
+      "step": 49720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1137175197368863e-07,
+      "loss": 1.4735,
+      "step": 49730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.107077242752951e-07,
+      "loss": 1.4026,
+      "step": 49740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.1004435846384405e-07,
+      "loss": 1.3457,
+      "step": 49750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.0938165473989037e-07,
+      "loss": 1.4287,
+      "step": 49760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.087196133037887e-07,
+      "loss": 1.37,
+      "step": 49770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0805823435569345e-07,
+      "loss": 1.4296,
+      "step": 49780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0739751809555877e-07,
+      "loss": 1.4362,
+      "step": 49790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0673746472313857e-07,
+      "loss": 1.339,
+      "step": 49800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0607807443798597e-07,
+      "loss": 1.4388,
+      "step": 49810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.054193474394543e-07,
+      "loss": 1.4443,
+      "step": 49820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0476128392669585e-07,
+      "loss": 1.3954,
+      "step": 49830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.041038840986621e-07,
+      "loss": 1.3238,
+      "step": 49840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.034471481541046e-07,
+      "loss": 1.3946,
+      "step": 49850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0279107629157337e-07,
+      "loss": 1.4905,
+      "step": 49860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0213566870941846e-07,
+      "loss": 1.3907,
+      "step": 49870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.014809256057888e-07,
+      "loss": 1.3657,
+      "step": 49880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.0082684717863186e-07,
+      "loss": 1.3607,
+      "step": 49890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.001734336256948e-07,
+      "loss": 1.3897,
+      "step": 49900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9952068514452454e-07,
+      "loss": 1.4905,
+      "step": 49910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9886860193246436e-07,
+      "loss": 1.3367,
+      "step": 49920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9821718418665825e-07,
+      "loss": 1.4469,
+      "step": 49930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.975664321040503e-07,
+      "loss": 1.4357,
+      "step": 49940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9691634588138063e-07,
+      "loss": 1.4173,
+      "step": 49950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.962669257151896e-07,
+      "loss": 1.4367,
+      "step": 49960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.956181718018158e-07,
+      "loss": 1.39,
+      "step": 49970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9497008433739677e-07,
+      "loss": 1.3858,
+      "step": 49980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.943226635178678e-07,
+      "loss": 1.343,
+      "step": 49990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.936759095389641e-07,
+      "loss": 1.403,
+      "step": 50000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.930298225962169e-07,
+      "loss": 1.4926,
+      "step": 50010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9238440288495764e-07,
+      "loss": 1.3851,
+      "step": 50020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9173965060031506e-07,
+      "loss": 1.3805,
+      "step": 50030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.910955659372178e-07,
+      "loss": 1.3661,
+      "step": 50040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.9045214909039083e-07,
+      "loss": 1.3442,
+      "step": 50050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.898094002543575e-07,
+      "loss": 1.4496,
+      "step": 50060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8916731962344013e-07,
+      "loss": 1.5284,
+      "step": 50070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8852590739175817e-07,
+      "loss": 1.5047,
+      "step": 50080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.878851637532298e-07,
+      "loss": 1.3519,
+      "step": 50090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.872450889015693e-07,
+      "loss": 1.3669,
+      "step": 50100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8660568303029075e-07,
+      "loss": 1.4639,
+      "step": 50110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.85966946332705e-07,
+      "loss": 1.4427,
+      "step": 50120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.853288790019207e-07,
+      "loss": 1.545,
+      "step": 50130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8469148123084434e-07,
+      "loss": 1.3799,
+      "step": 50140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.840547532121801e-07,
+      "loss": 1.3877,
+      "step": 50150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8341869513842933e-07,
+      "loss": 1.4372,
+      "step": 50160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8278330720189097e-07,
+      "loss": 1.3612,
+      "step": 50170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8214858959466185e-07,
+      "loss": 1.4295,
+      "step": 50180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8151454250863433e-07,
+      "loss": 1.3531,
+      "step": 50190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8088116613550013e-07,
+      "loss": 1.4504,
+      "step": 50200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.8024846066674745e-07,
+      "loss": 1.469,
+      "step": 50210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7961642629366126e-07,
+      "loss": 1.3953,
+      "step": 50220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.789850632073243e-07,
+      "loss": 1.4325,
+      "step": 50230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7835437159861563e-07,
+      "loss": 1.2445,
+      "step": 50240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.777243516582126e-07,
+      "loss": 1.4571,
+      "step": 50250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.770950035765879e-07,
+      "loss": 1.5176,
+      "step": 50260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7646632754401216e-07,
+      "loss": 1.4227,
+      "step": 50270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.758383237505527e-07,
+      "loss": 1.5034,
+      "step": 50280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.752109923860727e-07,
+      "loss": 1.3712,
+      "step": 50290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.745843336402332e-07,
+      "loss": 1.3095,
+      "step": 50300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7395834770249126e-07,
+      "loss": 1.4313,
+      "step": 50310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.733330347621005e-07,
+      "loss": 1.4667,
+      "step": 50320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.727083950081119e-07,
+      "loss": 1.3535,
+      "step": 50330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.720844286293717e-07,
+      "loss": 1.3745,
+      "step": 50340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.714611358145228e-07,
+      "loss": 1.3496,
+      "step": 50350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.7083851675200583e-07,
+      "loss": 1.533,
+      "step": 50360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.70216571630057e-07,
+      "loss": 1.3164,
+      "step": 50370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.695953006367069e-07,
+      "loss": 1.3811,
+      "step": 50380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.689747039597848e-07,
+      "loss": 1.4504,
+      "step": 50390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.68354781786915e-07,
+      "loss": 1.3711,
+      "step": 50400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6773553430551826e-07,
+      "loss": 1.4229,
+      "step": 50410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6711696170281104e-07,
+      "loss": 1.4003,
+      "step": 50420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.664990641658058e-07,
+      "loss": 1.4979,
+      "step": 50430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.658818418813111e-07,
+      "loss": 1.3472,
+      "step": 50440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.652652950359308e-07,
+      "loss": 1.387,
+      "step": 50450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6464942381606627e-07,
+      "loss": 1.4425,
+      "step": 50460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.640342284079123e-07,
+      "loss": 1.3715,
+      "step": 50470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.634197089974605e-07,
+      "loss": 1.4721,
+      "step": 50480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.6280586577049795e-07,
+      "loss": 1.3822,
+      "step": 50490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.621926989126078e-07,
+      "loss": 1.4419,
+      "step": 50500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.615802086091679e-07,
+      "loss": 1.4192,
+      "step": 50510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.609683950453523e-07,
+      "loss": 1.3691,
+      "step": 50520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.603572584061298e-07,
+      "loss": 1.481,
+      "step": 50530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5974679887626475e-07,
+      "loss": 1.4137,
+      "step": 50540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.591370166403173e-07,
+      "loss": 1.3879,
+      "step": 50550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5852791188264197e-07,
+      "loss": 1.498,
+      "step": 50560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5791948478738923e-07,
+      "loss": 1.2795,
+      "step": 50570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.573117355385044e-07,
+      "loss": 1.4671,
+      "step": 50580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.567046643197274e-07,
+      "loss": 1.3694,
+      "step": 50590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5609827131459375e-07,
+      "loss": 1.3779,
+      "step": 50600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.55492556706434e-07,
+      "loss": 1.468,
+      "step": 50610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5488752067837294e-07,
+      "loss": 1.4168,
+      "step": 50620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5428316341333107e-07,
+      "loss": 1.4952,
+      "step": 50630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.53679485094023e-07,
+      "loss": 1.3855,
+      "step": 50640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.530764859029583e-07,
+      "loss": 1.2957,
+      "step": 50650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5247416602244114e-07,
+      "loss": 1.4868,
+      "step": 50660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5187252563457055e-07,
+      "loss": 1.3649,
+      "step": 50670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.512715649212399e-07,
+      "loss": 1.4809,
+      "step": 50680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.50671284064137e-07,
+      "loss": 1.4311,
+      "step": 50690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.5007168324474475e-07,
+      "loss": 1.319,
+      "step": 50700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.494727626443394e-07,
+      "loss": 1.4634,
+      "step": 50710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.488745224439923e-07,
+      "loss": 1.3219,
+      "step": 50720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4827696282456957e-07,
+      "loss": 1.463,
+      "step": 50730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4768008396672954e-07,
+      "loss": 1.3756,
+      "step": 50740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4708388605092725e-07,
+      "loss": 1.2764,
+      "step": 50750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.464883692574105e-07,
+      "loss": 1.4847,
+      "step": 50760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4589353376622134e-07,
+      "loss": 1.3769,
+      "step": 50770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4529937975719566e-07,
+      "loss": 1.4222,
+      "step": 50780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.447059074099639e-07,
+      "loss": 1.4225,
+      "step": 50790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4411311690394986e-07,
+      "loss": 1.476,
+      "step": 50800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.435210084183712e-07,
+      "loss": 1.4765,
+      "step": 50810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4292958213224057e-07,
+      "loss": 1.4309,
+      "step": 50820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4233883822436246e-07,
+      "loss": 1.448,
+      "step": 50830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4174877687333534e-07,
+      "loss": 1.3691,
+      "step": 50840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.4115939825755356e-07,
+      "loss": 1.4325,
+      "step": 50850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.405707025552026e-07,
+      "loss": 1.4239,
+      "step": 50860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3998268994426256e-07,
+      "loss": 1.3311,
+      "step": 50870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.393953606025068e-07,
+      "loss": 1.4666,
+      "step": 50880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.388087147075019e-07,
+      "loss": 1.321,
+      "step": 50890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3822275243660848e-07,
+      "loss": 1.3567,
+      "step": 50900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.376374739669801e-07,
+      "loss": 1.3799,
+      "step": 50910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3705287947556277e-07,
+      "loss": 1.3473,
+      "step": 50920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3646896913909702e-07,
+      "loss": 1.5558,
+      "step": 50930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.3588574313411578e-07,
+      "loss": 1.4116,
+      "step": 50940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.353032016369447e-07,
+      "loss": 1.4822,
+      "step": 50950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.347213448237043e-07,
+      "loss": 1.4546,
+      "step": 50960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3414017287030633e-07,
+      "loss": 1.3667,
+      "step": 50970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.335596859524561e-07,
+      "loss": 1.4014,
+      "step": 50980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3297988424565153e-07,
+      "loss": 1.4236,
+      "step": 50990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.324007679251844e-07,
+      "loss": 1.3804,
+      "step": 51000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3182233716613712e-07,
+      "loss": 1.4616,
+      "step": 51010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3124459214338675e-07,
+      "loss": 1.3859,
+      "step": 51020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3066753303160273e-07,
+      "loss": 1.489,
+      "step": 51030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.3009116000524688e-07,
+      "loss": 1.4241,
+      "step": 51040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2951547323857265e-07,
+      "loss": 1.3409,
+      "step": 51050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2894047290562854e-07,
+      "loss": 1.4544,
+      "step": 51060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2836615918025302e-07,
+      "loss": 1.3423,
+      "step": 51070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2779253223607772e-07,
+      "loss": 1.4778,
+      "step": 51080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2721959224652785e-07,
+      "loss": 1.3451,
+      "step": 51090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2664733938481841e-07,
+      "loss": 1.3925,
+      "step": 51100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2607577382395883e-07,
+      "loss": 1.4626,
+      "step": 51110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2550489573675034e-07,
+      "loss": 1.3469,
+      "step": 51120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2493470529578564e-07,
+      "loss": 1.4822,
+      "step": 51130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2436520267344996e-07,
+      "loss": 1.3929,
+      "step": 51140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2379638804192027e-07,
+      "loss": 1.3547,
+      "step": 51150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.232282615731668e-07,
+      "loss": 1.3785,
+      "step": 51160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2266082343895019e-07,
+      "loss": 1.4369,
+      "step": 51170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2209407381082377e-07,
+      "loss": 1.4258,
+      "step": 51180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2152801286013214e-07,
+      "loss": 1.4636,
+      "step": 51190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.2096264075801204e-07,
+      "loss": 1.3544,
+      "step": 51200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.20397957675392e-07,
+      "loss": 1.4798,
+      "step": 51210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1983396378299245e-07,
+      "loss": 1.3784,
+      "step": 51220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1927065925132496e-07,
+      "loss": 1.5425,
+      "step": 51230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.187080442506928e-07,
+      "loss": 1.4578,
+      "step": 51240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1814611895119132e-07,
+      "loss": 1.3617,
+      "step": 51250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1758488352270623e-07,
+      "loss": 1.4697,
+      "step": 51260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1702433813491684e-07,
+      "loss": 1.4668,
+      "step": 51270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1646448295729093e-07,
+      "loss": 1.5344,
+      "step": 51280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1590531815908977e-07,
+      "loss": 1.4254,
+      "step": 51290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.153468439093645e-07,
+      "loss": 1.3839,
+      "step": 51300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1478906037695895e-07,
+      "loss": 1.4251,
+      "step": 51310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1423196773050715e-07,
+      "loss": 1.4035,
+      "step": 51320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1367556613843404e-07,
+      "loss": 1.4797,
+      "step": 51330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1311985576895678e-07,
+      "loss": 1.4694,
+      "step": 51340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1256483679008206e-07,
+      "loss": 1.4229,
+      "step": 51350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.12010509369609e-07,
+      "loss": 1.5139,
+      "step": 51360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1145687367512658e-07,
+      "loss": 1.3324,
+      "step": 51370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.109039298740148e-07,
+      "loss": 1.3772,
+      "step": 51380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1035167813344525e-07,
+      "loss": 1.4007,
+      "step": 51390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0980011862037963e-07,
+      "loss": 1.4037,
+      "step": 51400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0924925150157012e-07,
+      "loss": 1.4886,
+      "step": 51410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.086990769435604e-07,
+      "loss": 1.4342,
+      "step": 51420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0814959511268384e-07,
+      "loss": 1.477,
+      "step": 51430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0760080617506505e-07,
+      "loss": 1.3055,
+      "step": 51440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.070527102966191e-07,
+      "loss": 1.3814,
+      "step": 51450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0650530764305093e-07,
+      "loss": 1.3991,
+      "step": 51460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0595859837985676e-07,
+      "loss": 1.5147,
+      "step": 51470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0541258267232272e-07,
+      "loss": 1.4247,
+      "step": 51480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.048672606855251e-07,
+      "loss": 1.3551,
+      "step": 51490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.043226325843306e-07,
+      "loss": 1.4179,
+      "step": 51500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0377869853339642e-07,
+      "loss": 1.4827,
+      "step": 51510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0323545869716937e-07,
+      "loss": 1.3623,
+      "step": 51520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.0269291323988722e-07,
+      "loss": 1.5045,
+      "step": 51530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.021510623255768e-07,
+      "loss": 1.3529,
+      "step": 51540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0160990611805569e-07,
+      "loss": 1.4203,
+      "step": 51550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.010694447809311e-07,
+      "loss": 1.4673,
+      "step": 51560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0052967847760062e-07,
+      "loss": 1.3546,
+      "step": 51570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.999906073712507e-07,
+      "loss": 1.4195,
+      "step": 51580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9945223162485872e-07,
+      "loss": 1.3924,
+      "step": 51590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9891455140119143e-07,
+      "loss": 1.3589,
+      "step": 51600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9837756686280517e-07,
+      "loss": 1.4272,
+      "step": 51610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9784127817204569e-07,
+      "loss": 1.3824,
+      "step": 51620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9730568549104962e-07,
+      "loss": 1.4179,
+      "step": 51630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9677078898174108e-07,
+      "loss": 1.3794,
+      "step": 51640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9623658880583512e-07,
+      "loss": 1.3738,
+      "step": 51650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9570308512483675e-07,
+      "loss": 1.4912,
+      "step": 51660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9517027810003946e-07,
+      "loss": 1.3882,
+      "step": 51670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9463816789252605e-07,
+      "loss": 1.4549,
+      "step": 51680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9410675466316919e-07,
+      "loss": 1.4513,
+      "step": 51690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9357603857263069e-07,
+      "loss": 1.3091,
+      "step": 51700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9304601978136106e-07,
+      "loss": 1.475,
+      "step": 51710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.925166984496013e-07,
+      "loss": 1.3974,
+      "step": 51720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9198807473737978e-07,
+      "loss": 1.4558,
+      "step": 51730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9146014880451536e-07,
+      "loss": 1.3956,
+      "step": 51740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.9093292081061477e-07,
+      "loss": 1.3346,
+      "step": 51750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.904063909150755e-07,
+      "loss": 1.436,
+      "step": 51760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.898805592770825e-07,
+      "loss": 1.36,
+      "step": 51770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8935542605560965e-07,
+      "loss": 1.4772,
+      "step": 51780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8883099140942052e-07,
+      "loss": 1.4087,
+      "step": 51790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.883072554970669e-07,
+      "loss": 1.3828,
+      "step": 51800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8778421847688938e-07,
+      "loss": 1.4386,
+      "step": 51810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8726188050701726e-07,
+      "loss": 1.3423,
+      "step": 51820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8674024174536842e-07,
+      "loss": 1.4599,
+      "step": 51830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8621930234964947e-07,
+      "loss": 1.4496,
+      "step": 51840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.856990624773558e-07,
+      "loss": 1.4165,
+      "step": 51850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.851795222857708e-07,
+      "loss": 1.4897,
+      "step": 51860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8466068193196713e-07,
+      "loss": 1.4517,
+      "step": 51870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8414254157280514e-07,
+      "loss": 1.4351,
+      "step": 51880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.83625101364934e-07,
+      "loss": 1.398,
+      "step": 51890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8310836146479105e-07,
+      "loss": 1.386,
+      "step": 51900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8259232202860105e-07,
+      "loss": 1.3451,
+      "step": 51910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.820769832123781e-07,
+      "loss": 1.3208,
+      "step": 51920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.815623451719245e-07,
+      "loss": 1.4625,
+      "step": 51930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8104840806283025e-07,
+      "loss": 1.4223,
+      "step": 51940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8053517204047355e-07,
+      "loss": 1.448,
+      "step": 51950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8002263726002e-07,
+      "loss": 1.5047,
+      "step": 51960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7951080387642457e-07,
+      "loss": 1.3197,
+      "step": 51970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7899967204442953e-07,
+      "loss": 1.4907,
+      "step": 51980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7848924191856494e-07,
+      "loss": 1.3541,
+      "step": 51990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7797951365314813e-07,
+      "loss": 1.4095,
+      "step": 52000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7747048740228478e-07,
+      "loss": 1.4639,
+      "step": 52010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7696216331986897e-07,
+      "loss": 1.4228,
+      "step": 52020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.764545415595817e-07,
+      "loss": 1.5053,
+      "step": 52030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7594762227489182e-07,
+      "loss": 1.395,
+      "step": 52040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7544140561905565e-07,
+      "loss": 1.3773,
+      "step": 52050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7493589174511744e-07,
+      "loss": 1.5421,
+      "step": 52060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7443108080590853e-07,
+      "loss": 1.3356,
+      "step": 52070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.739269729540488e-07,
+      "loss": 1.4562,
+      "step": 52080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7342356834194384e-07,
+      "loss": 1.379,
+      "step": 52090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7292086712178802e-07,
+      "loss": 1.39,
+      "step": 52100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.7241886944556223e-07,
+      "loss": 1.4216,
+      "step": 52110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.719175754650354e-07,
+      "loss": 1.3665,
+      "step": 52120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.714169853317632e-07,
+      "loss": 1.3731,
+      "step": 52130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7091709919708877e-07,
+      "loss": 1.3674,
+      "step": 52140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.7041791721214202e-07,
+      "loss": 1.4534,
+      "step": 52150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6991943952784029e-07,
+      "loss": 1.4916,
+      "step": 52160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6942166629488777e-07,
+      "loss": 1.3948,
+      "step": 52170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.689245976637763e-07,
+      "loss": 1.4135,
+      "step": 52180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6842823378478397e-07,
+      "loss": 1.4668,
+      "step": 52190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6793257480797604e-07,
+      "loss": 1.3389,
+      "step": 52200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6743762088320458e-07,
+      "loss": 1.3862,
+      "step": 52210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.669433721601091e-07,
+      "loss": 1.4068,
+      "step": 52220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.664498287881147e-07,
+      "loss": 1.4009,
+      "step": 52230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6595699091643435e-07,
+      "loss": 1.4537,
+      "step": 52240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6546485869406742e-07,
+      "loss": 1.4098,
+      "step": 52250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6497343226979962e-07,
+      "loss": 1.4263,
+      "step": 52260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.644827117922035e-07,
+      "loss": 1.3768,
+      "step": 52270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6399269740963815e-07,
+      "loss": 1.4599,
+      "step": 52280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.635033892702495e-07,
+      "loss": 1.2761,
+      "step": 52290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6301478752196952e-07,
+      "loss": 1.4092,
+      "step": 52300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6252689231251674e-07,
+      "loss": 1.4569,
+      "step": 52310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6203970378939587e-07,
+      "loss": 1.3444,
+      "step": 52320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.615532220998986e-07,
+      "loss": 1.5106,
+      "step": 52330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6106744739110254e-07,
+      "loss": 1.4383,
+      "step": 52340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6058237980987157e-07,
+      "loss": 1.3098,
+      "step": 52350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.6009801950285537e-07,
+      "loss": 1.4821,
+      "step": 52360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5961436661649037e-07,
+      "loss": 1.5052,
+      "step": 52370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5913142129699939e-07,
+      "loss": 1.3972,
+      "step": 52380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5864918369039028e-07,
+      "loss": 1.384,
+      "step": 52390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5816765394245815e-07,
+      "loss": 1.3582,
+      "step": 52400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5768683219878288e-07,
+      "loss": 1.4288,
+      "step": 52410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.572067186047313e-07,
+      "loss": 1.3708,
+      "step": 52420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.567273133054556e-07,
+      "loss": 1.4145,
+      "step": 52430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5624861644589402e-07,
+      "loss": 1.3524,
+      "step": 52440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5577062817077105e-07,
+      "loss": 1.3672,
+      "step": 52450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5529334862459527e-07,
+      "loss": 1.4146,
+      "step": 52460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.548167779516632e-07,
+      "loss": 1.3676,
+      "step": 52470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.543409162960563e-07,
+      "loss": 1.5683,
+      "step": 52480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5386576380164053e-07,
+      "loss": 1.4069,
+      "step": 52490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5339132061206907e-07,
+      "loss": 1.3562,
+      "step": 52500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5291758687077968e-07,
+      "loss": 1.4527,
+      "step": 52510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5244456272099612e-07,
+      "loss": 1.3738,
+      "step": 52520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5197224830572675e-07,
+      "loss": 1.5537,
+      "step": 52530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5150064376776713e-07,
+      "loss": 1.3499,
+      "step": 52540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5102974924969593e-07,
+      "loss": 1.4154,
+      "step": 52550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5055956489387847e-07,
+      "loss": 1.4182,
+      "step": 52560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.5009009084246572e-07,
+      "loss": 1.3336,
+      "step": 52570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.496213272373931e-07,
+      "loss": 1.4042,
+      "step": 52580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4915327422038161e-07,
+      "loss": 1.3794,
+      "step": 52590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4868593193293757e-07,
+      "loss": 1.3849,
+      "step": 52600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4821930051635154e-07,
+      "loss": 1.4598,
+      "step": 52610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4775338011170037e-07,
+      "loss": 1.3333,
+      "step": 52620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4728817085984588e-07,
+      "loss": 1.459,
+      "step": 52630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4682367290143302e-07,
+      "loss": 1.4248,
+      "step": 52640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.463598863768939e-07,
+      "loss": 1.4405,
+      "step": 52650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4589681142644446e-07,
+      "loss": 1.4004,
+      "step": 52660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4543444819008572e-07,
+      "loss": 1.4346,
+      "step": 52670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4497279680760424e-07,
+      "loss": 1.4889,
+      "step": 52680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4451185741857028e-07,
+      "loss": 1.4576,
+      "step": 52690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4405163016233903e-07,
+      "loss": 1.4237,
+      "step": 52700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4359211517805117e-07,
+      "loss": 1.4952,
+      "step": 52710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4313331260463137e-07,
+      "loss": 1.3291,
+      "step": 52720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4267522258078843e-07,
+      "loss": 1.3974,
+      "step": 52730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4221784524501654e-07,
+      "loss": 1.4058,
+      "step": 52740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4176118073559453e-07,
+      "loss": 1.4019,
+      "step": 52750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4130522919058532e-07,
+      "loss": 1.4619,
+      "step": 52760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.408499907478361e-07,
+      "loss": 1.4119,
+      "step": 52770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4039546554497907e-07,
+      "loss": 1.5243,
+      "step": 52780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994165371943065e-07,
+      "loss": 1.339,
+      "step": 52790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3948855540839078e-07,
+      "loss": 1.3373,
+      "step": 52800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3903617074884524e-07,
+      "loss": 1.424,
+      "step": 52810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3858449987756206e-07,
+      "loss": 1.4378,
+      "step": 52820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.381335429310951e-07,
+      "loss": 1.4546,
+      "step": 52830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.376833000457814e-07,
+      "loss": 1.3471,
+      "step": 52840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.372337713577429e-07,
+      "loss": 1.4562,
+      "step": 52850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3678495700288507e-07,
+      "loss": 1.5194,
+      "step": 52860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3633685711689798e-07,
+      "loss": 1.3273,
+      "step": 52870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3588947183525437e-07,
+      "loss": 1.4913,
+      "step": 52880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3544280129321268e-07,
+      "loss": 1.3919,
+      "step": 52890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3499684562581467e-07,
+      "loss": 1.4064,
+      "step": 52900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3455160496788522e-07,
+      "loss": 1.4415,
+      "step": 52910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3410707945403334e-07,
+      "loss": 1.4397,
+      "step": 52920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3366326921865263e-07,
+      "loss": 1.4159,
+      "step": 52930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.332201743959194e-07,
+      "loss": 1.3724,
+      "step": 52940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3277779511979478e-07,
+      "loss": 1.3059,
+      "step": 52950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.323361315240221e-07,
+      "loss": 1.4885,
+      "step": 52960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3189518374212978e-07,
+      "loss": 1.4303,
+      "step": 52970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3145495190742874e-07,
+      "loss": 1.5585,
+      "step": 52980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3101543615301497e-07,
+      "loss": 1.3442,
+      "step": 52990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3057663661176557e-07,
+      "loss": 1.3901,
+      "step": 53000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3013855341634323e-07,
+      "loss": 1.4148,
+      "step": 53010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.297011866991929e-07,
+      "loss": 1.3241,
+      "step": 53020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.292645365925435e-07,
+      "loss": 1.3827,
+      "step": 53030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2882860322840694e-07,
+      "loss": 1.3882,
+      "step": 53040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2839338673857893e-07,
+      "loss": 1.3594,
+      "step": 53050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.279588872546378e-07,
+      "loss": 1.4616,
+      "step": 53060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.275251049079454e-07,
+      "loss": 1.4564,
+      "step": 53070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2709203982964714e-07,
+      "loss": 1.4762,
+      "step": 53080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2665969215067074e-07,
+      "loss": 1.4079,
+      "step": 53090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.262280620017281e-07,
+      "loss": 1.4409,
+      "step": 53100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2579714951331312e-07,
+      "loss": 1.4811,
+      "step": 53110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2536695481570327e-07,
+      "loss": 1.4273,
+      "step": 53120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2493747803895922e-07,
+      "loss": 1.5069,
+      "step": 53130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2450871931292436e-07,
+      "loss": 1.3575,
+      "step": 53140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.240806787672247e-07,
+      "loss": 1.3954,
+      "step": 53150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2365335653126946e-07,
+      "loss": 1.451,
+      "step": 53160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2322675273425067e-07,
+      "loss": 1.3758,
+      "step": 53170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2280086750514314e-07,
+      "loss": 1.4251,
+      "step": 53180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.223757009727042e-07,
+      "loss": 1.4147,
+      "step": 53190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.219512532654743e-07,
+      "loss": 1.4168,
+      "step": 53200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.215275245117764e-07,
+      "loss": 1.5213,
+      "step": 53210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.2110451483971607e-07,
+      "loss": 1.3854,
+      "step": 53220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.206822243771813e-07,
+      "loss": 1.4103,
+      "step": 53230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.20260653251843e-07,
+      "loss": 1.4043,
+      "step": 53240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.198398015911542e-07,
+      "loss": 1.3109,
+      "step": 53250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1941966952235067e-07,
+      "loss": 1.4676,
+      "step": 53260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1900025717245106e-07,
+      "loss": 1.4025,
+      "step": 53270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1858156466825532e-07,
+      "loss": 1.4561,
+      "step": 53280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1816359213634692e-07,
+      "loss": 1.3913,
+      "step": 53290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1774633970309091e-07,
+      "loss": 1.3393,
+      "step": 53300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1732980749463469e-07,
+      "loss": 1.4936,
+      "step": 53310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1691399563690836e-07,
+      "loss": 1.4188,
+      "step": 53320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1649890425562415e-07,
+      "loss": 1.4139,
+      "step": 53330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1608453347627585e-07,
+      "loss": 1.4166,
+      "step": 53340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1567088342414046e-07,
+      "loss": 1.3771,
+      "step": 53350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.15257954224276e-07,
+      "loss": 1.4688,
+      "step": 53360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1484574600152232e-07,
+      "loss": 1.4478,
+      "step": 53370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1443425888050336e-07,
+      "loss": 1.4458,
+      "step": 53380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1402349298562321e-07,
+      "loss": 1.4141,
+      "step": 53390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1361344844106809e-07,
+      "loss": 1.4236,
+      "step": 53400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1320412537080661e-07,
+      "loss": 1.4997,
+      "step": 53410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1279552389858895e-07,
+      "loss": 1.4172,
+      "step": 53420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.123876441479474e-07,
+      "loss": 1.4604,
+      "step": 53430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1198048624219637e-07,
+      "loss": 1.3913,
+      "step": 53440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1157405030443075e-07,
+      "loss": 1.4141,
+      "step": 53450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1116833645752834e-07,
+      "loss": 1.4549,
+      "step": 53460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1076334482414797e-07,
+      "loss": 1.4209,
+      "step": 53470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1035907552673086e-07,
+      "loss": 1.4437,
+      "step": 53480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0995552868749954e-07,
+      "loss": 1.4142,
+      "step": 53490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0955270442845751e-07,
+      "loss": 1.3541,
+      "step": 53500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0915060287139067e-07,
+      "loss": 1.439,
+      "step": 53510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0874922413786599e-07,
+      "loss": 1.4569,
+      "step": 53520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0834856834923218e-07,
+      "loss": 1.4528,
+      "step": 53530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.079486356266185e-07,
+      "loss": 1.3716,
+      "step": 53540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0754942609093655e-07,
+      "loss": 1.379,
+      "step": 53550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.071509398628795e-07,
+      "loss": 1.4226,
+      "step": 53560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0675317706292071e-07,
+      "loss": 1.4432,
+      "step": 53570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0635613781131566e-07,
+      "loss": 1.4423,
+      "step": 53580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0595982222810108e-07,
+      "loss": 1.4391,
+      "step": 53590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0556423043309505e-07,
+      "loss": 1.3942,
+      "step": 53600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0516936254589576e-07,
+      "loss": 1.5535,
+      "step": 53610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0477521868588442e-07,
+      "loss": 1.4212,
+      "step": 53620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0438179897222095e-07,
+      "loss": 1.4806,
+      "step": 53630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.03989103523848e-07,
+      "loss": 1.4382,
+      "step": 53640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0359713245948921e-07,
+      "loss": 1.3836,
+      "step": 53650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0320588589764868e-07,
+      "loss": 1.4957,
+      "step": 53660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0281536395661151e-07,
+      "loss": 1.4251,
+      "step": 53670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0242556675444382e-07,
+      "loss": 1.4442,
+      "step": 53680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0203649440899327e-07,
+      "loss": 1.4085,
+      "step": 53690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0164814703788716e-07,
+      "loss": 1.362,
+      "step": 53700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0126052475853493e-07,
+      "loss": 1.4542,
+      "step": 53710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0087362768812531e-07,
+      "loss": 1.4018,
+      "step": 53720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.004874559436289e-07,
+      "loss": 1.4513,
+      "step": 53730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.0010200964179678e-07,
+      "loss": 1.3712,
+      "step": 53740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.971728889916043e-08,
+      "loss": 1.3423,
+      "step": 53750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.933329383203239e-08,
+      "loss": 1.4579,
+      "step": 53760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.895002455650532e-08,
+      "loss": 1.439,
+      "step": 53770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.856748118845294e-08,
+      "loss": 1.4587,
+      "step": 53780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.818566384352857e-08,
+      "loss": 1.372,
+      "step": 53790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.780457263716819e-08,
+      "loss": 1.4337,
+      "step": 53800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.742420768458577e-08,
+      "loss": 1.4777,
+      "step": 53810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.704456910077653e-08,
+      "loss": 1.2913,
+      "step": 53820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.6665657000517e-08,
+      "loss": 1.4315,
+      "step": 53830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.628747149836276e-08,
+      "loss": 1.3342,
+      "step": 53840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.59100127086507e-08,
+      "loss": 1.3771,
+      "step": 53850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.553328074549756e-08,
+      "loss": 1.4648,
+      "step": 53860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.515727572280031e-08,
+      "loss": 1.3666,
+      "step": 53870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.478199775423636e-08,
+      "loss": 1.5229,
+      "step": 53880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.440744695326298e-08,
+      "loss": 1.4055,
+      "step": 53890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.403362343311823e-08,
+      "loss": 1.4439,
+      "step": 53900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.366052730681918e-08,
+      "loss": 1.4546,
+      "step": 53910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.32881586871645e-08,
+      "loss": 1.319,
+      "step": 53920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.291651768673137e-08,
+      "loss": 1.4593,
+      "step": 53930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.254560441787824e-08,
+      "loss": 1.3392,
+      "step": 53940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.217541899274262e-08,
+      "loss": 1.4773,
+      "step": 53950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.18059615232425e-08,
+      "loss": 1.4622,
+      "step": 53960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.143723212107574e-08,
+      "loss": 1.2704,
+      "step": 53970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.106923089772013e-08,
+      "loss": 1.4603,
+      "step": 53980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.070195796443276e-08,
+      "loss": 1.3552,
+      "step": 53990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.03354134322515e-08,
+      "loss": 1.3797,
+      "step": 54000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.996959741199296e-08,
+      "loss": 1.4563,
+      "step": 54010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.960451001425396e-08,
+      "loss": 1.4195,
+      "step": 54020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.924015134941177e-08,
+      "loss": 1.354,
+      "step": 54030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.887652152762188e-08,
+      "loss": 1.3584,
+      "step": 54040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.851362065882052e-08,
+      "loss": 1.3394,
+      "step": 54050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.815144885272325e-08,
+      "loss": 1.4866,
+      "step": 54060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.7790006218825e-08,
+      "loss": 1.4845,
+      "step": 54070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.742929286640084e-08,
+      "loss": 1.4619,
+      "step": 54080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.706930890450437e-08,
+      "loss": 1.473,
+      "step": 54090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.671005444196967e-08,
+      "loss": 1.3674,
+      "step": 54100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.635152958740956e-08,
+      "loss": 1.5182,
+      "step": 54110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.599373444921677e-08,
+      "loss": 1.3667,
+      "step": 54120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.563666913556313e-08,
+      "loss": 1.3779,
+      "step": 54130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.528033375440003e-08,
+      "loss": 1.3495,
+      "step": 54140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.4924728413458e-08,
+      "loss": 1.3519,
+      "step": 54150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.456985322024686e-08,
+      "loss": 1.4335,
+      "step": 54160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.421570828205577e-08,
+      "loss": 1.4452,
+      "step": 54170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.386229370595244e-08,
+      "loss": 1.4509,
+      "step": 54180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.350960959878557e-08,
+      "loss": 1.3634,
+      "step": 54190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.315765606718123e-08,
+      "loss": 1.3806,
+      "step": 54200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.280643321754517e-08,
+      "loss": 1.4367,
+      "step": 54210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.24559411560627e-08,
+      "loss": 1.43,
+      "step": 54220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.210617998869742e-08,
+      "loss": 1.4556,
+      "step": 54230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.175714982119248e-08,
+      "loss": 1.4079,
+      "step": 54240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.140885075907017e-08,
+      "loss": 1.333,
+      "step": 54250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.106128290763071e-08,
+      "loss": 1.4531,
+      "step": 54260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.071444637195475e-08,
+      "loss": 1.4301,
+      "step": 54270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.036834125690036e-08,
+      "loss": 1.5075,
+      "step": 54280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 8.002296766710577e-08,
+      "loss": 1.3608,
+      "step": 54290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.967832570698774e-08,
+      "loss": 1.3843,
+      "step": 54300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.9334415480741e-08,
+      "loss": 1.5047,
+      "step": 54310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.899123709233986e-08,
+      "loss": 1.3825,
+      "step": 54320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.864879064553716e-08,
+      "loss": 1.4266,
+      "step": 54330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.830707624386507e-08,
+      "loss": 1.3716,
+      "step": 54340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.796609399063266e-08,
+      "loss": 1.3939,
+      "step": 54350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.762584398892936e-08,
+      "loss": 1.449,
+      "step": 54360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.728632634162292e-08,
+      "loss": 1.3846,
+      "step": 54370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.694754115135927e-08,
+      "loss": 1.4921,
+      "step": 54380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.66094885205626e-08,
+      "loss": 1.4132,
+      "step": 54390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.627216855143671e-08,
+      "loss": 1.3961,
+      "step": 54400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.593558134596308e-08,
+      "loss": 1.4267,
+      "step": 54410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.559972700590168e-08,
+      "loss": 1.3113,
+      "step": 54420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.526460563279158e-08,
+      "loss": 1.4717,
+      "step": 54430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.493021732794926e-08,
+      "loss": 1.3917,
+      "step": 54440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.459656219247019e-08,
+      "loss": 1.3452,
+      "step": 54450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.426364032722789e-08,
+      "loss": 1.4512,
+      "step": 54460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.393145183287459e-08,
+      "loss": 1.408,
+      "step": 54470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 7.359999680984053e-08,
+      "loss": 1.4585,
+      "step": 54480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.32692753583339e-08,
+      "loss": 1.3809,
+      "step": 54490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.29392875783419e-08,
+      "loss": 1.3017,
+      "step": 54500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.261003356962949e-08,
+      "loss": 1.5324,
+      "step": 54510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.228151343173978e-08,
+      "loss": 1.3828,
+      "step": 54520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.195372726399363e-08,
+      "loss": 1.4393,
+      "step": 54530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.162667516549038e-08,
+      "loss": 1.369,
+      "step": 54540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.130035723510758e-08,
+      "loss": 1.3052,
+      "step": 54550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.097477357150079e-08,
+      "loss": 1.4718,
+      "step": 54560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.064992427310318e-08,
+      "loss": 1.4051,
+      "step": 54570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.032580943812623e-08,
+      "loss": 1.4143,
+      "step": 54580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 7.000242916455958e-08,
+      "loss": 1.4309,
+      "step": 54590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.967978355017002e-08,
+      "loss": 1.3639,
+      "step": 54600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.935787269250371e-08,
+      "loss": 1.4125,
+      "step": 54610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.903669668888252e-08,
+      "loss": 1.4196,
+      "step": 54620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.871625563640822e-08,
+      "loss": 1.4056,
+      "step": 54630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.839654963195885e-08,
+      "loss": 1.3971,
+      "step": 54640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.807757877219129e-08,
+      "loss": 1.4524,
+      "step": 54650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.775934315353922e-08,
+      "loss": 1.444,
+      "step": 54660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.74418428722154e-08,
+      "loss": 1.4319,
+      "step": 54670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.712507802420859e-08,
+      "loss": 1.5029,
+      "step": 54680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.680904870528664e-08,
+      "loss": 1.3974,
+      "step": 54690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.64937550109937e-08,
+      "loss": 1.3942,
+      "step": 54700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.617919703665348e-08,
+      "loss": 1.4129,
+      "step": 54710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.586537487736522e-08,
+      "loss": 1.4766,
+      "step": 54720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.555228862800634e-08,
+      "loss": 1.4698,
+      "step": 54730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.523993838323223e-08,
+      "loss": 1.3845,
+      "step": 54740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.492832423747541e-08,
+      "loss": 1.4268,
+      "step": 54750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.46174462849461e-08,
+      "loss": 1.44,
+      "step": 54760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.430730461963186e-08,
+      "loss": 1.3578,
+      "step": 54770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.399789933529715e-08,
+      "loss": 1.4259,
+      "step": 54780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.368923052548437e-08,
+      "loss": 1.334,
+      "step": 54790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.338129828351331e-08,
+      "loss": 1.3289,
+      "step": 54800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.307410270248059e-08,
+      "loss": 1.4263,
+      "step": 54810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.276764387526085e-08,
+      "loss": 1.3551,
+      "step": 54820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.246192189450495e-08,
+      "loss": 1.4544,
+      "step": 54830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.215693685264201e-08,
+      "loss": 1.3726,
+      "step": 54840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.185268884187773e-08,
+      "loss": 1.4597,
+      "step": 54850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.154917795419519e-08,
+      "loss": 1.5419,
+      "step": 54860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.12464042813546e-08,
+      "loss": 1.345,
+      "step": 54870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.094436791489327e-08,
+      "loss": 1.4131,
+      "step": 54880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.064306894612543e-08,
+      "loss": 1.4209,
+      "step": 54890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.03425074661429e-08,
+      "loss": 1.4017,
+      "step": 54900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.004268356581383e-08,
+      "loss": 1.4758,
+      "step": 54910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.974359733578406e-08,
+      "loss": 1.3769,
+      "step": 54920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.944524886647596e-08,
+      "loss": 1.4658,
+      "step": 54930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.914763824808878e-08,
+      "loss": 1.4163,
+      "step": 54940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.8850765570599434e-08,
+      "loss": 1.4002,
+      "step": 54950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.855463092376057e-08,
+      "loss": 1.4529,
+      "step": 54960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.825923439710251e-08,
+      "loss": 1.3202,
+      "step": 54970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.796457607993272e-08,
+      "loss": 1.5159,
+      "step": 54980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.7670656061334096e-08,
+      "loss": 1.3837,
+      "step": 54990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.737747443016778e-08,
+      "loss": 1.4538,
+      "step": 55000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.708503127507148e-08,
+      "loss": 1.4234,
+      "step": 55010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.679332668445864e-08,
+      "loss": 1.3723,
+      "step": 55020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.650236074652038e-08,
+      "loss": 1.4745,
+      "step": 55030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.621213354922439e-08,
+      "loss": 1.3503,
+      "step": 55040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.5922645180314375e-08,
+      "loss": 1.373,
+      "step": 55050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.563389572731143e-08,
+      "loss": 1.4822,
+      "step": 55060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 5.5345885277513224e-08,
+      "loss": 1.3983,
+      "step": 55070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.505861391799289e-08,
+      "loss": 1.4278,
+      "step": 55080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.477208173560122e-08,
+      "loss": 1.3658,
+      "step": 55090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.448628881696588e-08,
+      "loss": 1.368,
+      "step": 55100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.420123524848969e-08,
+      "loss": 1.4528,
+      "step": 55110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.391692111635316e-08,
+      "loss": 1.4094,
+      "step": 55120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.3633346506512255e-08,
+      "loss": 1.4209,
+      "step": 55130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.335051150470033e-08,
+      "loss": 1.3975,
+      "step": 55140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.30684161964265e-08,
+      "loss": 1.3758,
+      "step": 55150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.278706066697642e-08,
+      "loss": 1.5241,
+      "step": 55160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.250644500141206e-08,
+      "loss": 1.3133,
+      "step": 55170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.222656928457137e-08,
+      "loss": 1.5117,
+      "step": 55180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.194743360106919e-08,
+      "loss": 1.3504,
+      "step": 55190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.166903803529688e-08,
+      "loss": 1.4286,
+      "step": 55200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.139138267142074e-08,
+      "loss": 1.456,
+      "step": 55210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.111446759338473e-08,
+      "loss": 1.3822,
+      "step": 55220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.0838292884907995e-08,
+      "loss": 1.4633,
+      "step": 55230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.056285862948651e-08,
+      "loss": 1.3758,
+      "step": 55240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.028816491039201e-08,
+      "loss": 1.3598,
+      "step": 55250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 5.001421181067195e-08,
+      "loss": 1.5431,
+      "step": 55260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.9740999413150635e-08,
+      "loss": 1.4361,
+      "step": 55270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.9468527800428376e-08,
+      "loss": 1.3551,
+      "step": 55280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.919679705488123e-08,
+      "loss": 1.3869,
+      "step": 55290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.892580725866069e-08,
+      "loss": 1.429,
+      "step": 55300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.8655558493695966e-08,
+      "loss": 1.4245,
+      "step": 55310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.838605084169057e-08,
+      "loss": 1.4348,
+      "step": 55320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.811728438412433e-08,
+      "loss": 1.454,
+      "step": 55330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.784925920225364e-08,
+      "loss": 1.4262,
+      "step": 55340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.758197537711007e-08,
+      "loss": 1.4138,
+      "step": 55350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.7315432989501476e-08,
+      "loss": 1.4681,
+      "step": 55360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.70496321200109e-08,
+      "loss": 1.4173,
+      "step": 55370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.678457284899823e-08,
+      "loss": 1.4367,
+      "step": 55380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.6520255256598534e-08,
+      "loss": 1.3639,
+      "step": 55390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.625667942272261e-08,
+      "loss": 1.4294,
+      "step": 55400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.599384542705726e-08,
+      "loss": 1.4288,
+      "step": 55410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.5731753349064765e-08,
+      "loss": 1.4241,
+      "step": 55420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.547040326798341e-08,
+      "loss": 1.5087,
+      "step": 55430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.520979526282665e-08,
+      "loss": 1.3587,
+      "step": 55440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.494992941238396e-08,
+      "loss": 1.3564,
+      "step": 55450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.469080579522056e-08,
+      "loss": 1.5175,
+      "step": 55460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.443242448967683e-08,
+      "loss": 1.4543,
+      "step": 55470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.4174785573869174e-08,
+      "loss": 1.4469,
+      "step": 55480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.391788912568945e-08,
+      "loss": 1.405,
+      "step": 55490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.366173522280442e-08,
+      "loss": 1.2828,
+      "step": 55500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.340632394265743e-08,
+      "loss": 1.4629,
+      "step": 55510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.315165536246668e-08,
+      "loss": 1.383,
+      "step": 55520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.289772955922589e-08,
+      "loss": 1.4183,
+      "step": 55530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.264454660970391e-08,
+      "loss": 1.4313,
+      "step": 55540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.2392106590445625e-08,
+      "loss": 1.3913,
+      "step": 55550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.21404095777711e-08,
+      "loss": 1.4932,
+      "step": 55560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.188945564777558e-08,
+      "loss": 1.4703,
+      "step": 55570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.163924487632948e-08,
+      "loss": 1.4238,
+      "step": 55580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.1389777339079505e-08,
+      "loss": 1.3989,
+      "step": 55590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.1141053111446174e-08,
+      "loss": 1.3576,
+      "step": 55600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.089307226862682e-08,
+      "loss": 1.4674,
+      "step": 55610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.064583488559259e-08,
+      "loss": 1.3726,
+      "step": 55620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.03993410370912e-08,
+      "loss": 1.4836,
+      "step": 55630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.015359079764441e-08,
+      "loss": 1.3285,
+      "step": 55640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.990858424155003e-08,
+      "loss": 1.3585,
+      "step": 55650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.966432144288046e-08,
+      "loss": 1.5297,
+      "step": 55660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.9420802475483575e-08,
+      "loss": 1.4059,
+      "step": 55670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.9178027412982435e-08,
+      "loss": 1.4362,
+      "step": 55680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.893599632877498e-08,
+      "loss": 1.3122,
+      "step": 55690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8694709296033806e-08,
+      "loss": 1.3875,
+      "step": 55700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8454166387707495e-08,
+      "loss": 1.5723,
+      "step": 55710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.821436767651926e-08,
+      "loss": 1.4391,
+      "step": 55720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.797531323496695e-08,
+      "loss": 1.3925,
+      "step": 55730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.7737003135323856e-08,
+      "loss": 1.4466,
+      "step": 55740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.749943744963791e-08,
+      "loss": 1.3567,
+      "step": 55750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.726261624973221e-08,
+      "loss": 1.5051,
+      "step": 55760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.702653960720476e-08,
+      "loss": 1.2955,
+      "step": 55770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6791207593428747e-08,
+      "loss": 1.4178,
+      "step": 55780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6556620279551424e-08,
+      "loss": 1.3787,
+      "step": 55790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6322777736495504e-08,
+      "loss": 1.4153,
+      "step": 55800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.608968003495833e-08,
+      "loss": 1.4687,
+      "step": 55810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.585732724541241e-08,
+      "loss": 1.4,
+      "step": 55820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.562571943810461e-08,
+      "loss": 1.3756,
+      "step": 55830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.53948566830567e-08,
+      "loss": 1.3524,
+      "step": 55840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.516473905006534e-08,
+      "loss": 1.4009,
+      "step": 55850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.493536660870184e-08,
+      "loss": 1.4374,
+      "step": 55860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.470673942831182e-08,
+      "loss": 1.4292,
+      "step": 55870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.4478857578016664e-08,
+      "loss": 1.3667,
+      "step": 55880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.425172112671071e-08,
+      "loss": 1.3448,
+      "step": 55890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.402533014306458e-08,
+      "loss": 1.3514,
+      "step": 55900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.379968469552297e-08,
+      "loss": 1.3724,
+      "step": 55910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.357478485230492e-08,
+      "loss": 1.4065,
+      "step": 55920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.335063068140409e-08,
+      "loss": 1.4528,
+      "step": 55930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.312722225058879e-08,
+      "loss": 1.4164,
+      "step": 55940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.2904559627402185e-08,
+      "loss": 1.4266,
+      "step": 55950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.268264287916156e-08,
+      "loss": 1.4848,
+      "step": 55960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.24614720729588e-08,
+      "loss": 1.4204,
+      "step": 55970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.224104727566041e-08,
+      "loss": 1.4338,
+      "step": 55980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.2021368553906976e-08,
+      "loss": 1.4024,
+      "step": 55990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.1802435974113696e-08,
+      "loss": 1.3607,
+      "step": 56000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.1584249602470696e-08,
+      "loss": 1.3903,
+      "step": 56010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.1366809504941864e-08,
+      "loss": 1.5017,
+      "step": 56020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.115011574726573e-08,
+      "loss": 1.4818,
+      "step": 56030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.093416839495517e-08,
+      "loss": 1.4038,
+      "step": 56040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.071896751329739e-08,
+      "loss": 1.3959,
+      "step": 56050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.050451316735398e-08,
+      "loss": 1.4798,
+      "step": 56060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.029080542196056e-08,
+      "loss": 1.3082,
+      "step": 56070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.0077844341727114e-08,
+      "loss": 1.5152,
+      "step": 56080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.986562999103826e-08,
+      "loss": 1.4491,
+      "step": 56090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9654162434052393e-08,
+      "loss": 1.4133,
+      "step": 56100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.944344173470226e-08,
+      "loss": 1.3606,
+      "step": 56110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9233467956694962e-08,
+      "loss": 1.4502,
+      "step": 56120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.9024241163511935e-08,
+      "loss": 1.5457,
+      "step": 56130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.8815761418408416e-08,
+      "loss": 1.3683,
+      "step": 56140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.8608028784414266e-08,
+      "loss": 1.2972,
+      "step": 56150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.84010433243323e-08,
+      "loss": 1.4327,
+      "step": 56160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.819480510074052e-08,
+      "loss": 1.4075,
+      "step": 56170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7989314175991266e-08,
+      "loss": 1.4777,
+      "step": 56180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.778457061220985e-08,
+      "loss": 1.4184,
+      "step": 56190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7580574471296474e-08,
+      "loss": 1.3792,
+      "step": 56200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.7377325814925136e-08,
+      "loss": 1.4738,
+      "step": 56210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.717482470454391e-08,
+      "loss": 1.4527,
+      "step": 56220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6973071201374657e-08,
+      "loss": 1.4414,
+      "step": 56230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6772065366413857e-08,
+      "loss": 1.3095,
+      "step": 56240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.6571807260430404e-08,
+      "loss": 1.2619,
+      "step": 56250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.637229694396892e-08,
+      "loss": 1.3683,
+      "step": 56260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.6173534477347274e-08,
+      "loss": 1.4021,
+      "step": 56270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5975519920656554e-08,
+      "loss": 1.4318,
+      "step": 56280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.577825333376277e-08,
+      "loss": 1.3844,
+      "step": 56290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5581734776305433e-08,
+      "loss": 1.3774,
+      "step": 56300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5385964307697574e-08,
+      "loss": 1.4257,
+      "step": 56310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.5190941987126295e-08,
+      "loss": 1.3965,
+      "step": 56320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4996667873553038e-08,
+      "loss": 1.434,
+      "step": 56330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4803142025712214e-08,
+      "loss": 1.3078,
+      "step": 56340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4610364502112016e-08,
+      "loss": 1.4488,
+      "step": 56350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4418335361034994e-08,
+      "loss": 1.4883,
+      "step": 56360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4227054660537207e-08,
+      "loss": 1.4723,
+      "step": 56370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.4036522458448508e-08,
+      "loss": 1.428,
+      "step": 56380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3846738812371983e-08,
+      "loss": 1.3881,
+      "step": 56390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.365770377968507e-08,
+      "loss": 1.4084,
+      "step": 56400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3469417417538443e-08,
+      "loss": 1.3893,
+      "step": 56410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.328187978285684e-08,
+      "loss": 1.4046,
+      "step": 56420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.3095090932337693e-08,
+      "loss": 1.3907,
+      "step": 56430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.29090509224536e-08,
+      "loss": 1.3269,
+      "step": 56440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2723759809449018e-08,
+      "loss": 1.437,
+      "step": 56450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.253921764934358e-08,
+      "loss": 1.5101,
+      "step": 56460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2355424497929324e-08,
+      "loss": 1.4015,
+      "step": 56470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.2172380410772354e-08,
+      "loss": 1.4769,
+      "step": 56480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1990085443212294e-08,
+      "loss": 1.4463,
+      "step": 56490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1808539650362282e-08,
+      "loss": 1.4467,
+      "step": 56500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1627743087108965e-08,
+      "loss": 1.4366,
+      "step": 56510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1447695808112233e-08,
+      "loss": 1.383,
+      "step": 56520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1268397867805768e-08,
+      "loss": 1.4819,
+      "step": 56530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.1089849320396483e-08,
+      "loss": 1.2811,
+      "step": 56540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.091205021986481e-08,
+      "loss": 1.3743,
+      "step": 56550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0735000619964696e-08,
+      "loss": 1.479,
+      "step": 56560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.055870057422332e-08,
+      "loss": 1.3423,
+      "step": 56570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0383150135941386e-08,
+      "loss": 1.5044,
+      "step": 56580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.0208349358192826e-08,
+      "loss": 1.4069,
+      "step": 56590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2.003429829382536e-08,
+      "loss": 1.3744,
+      "step": 56600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9860996995459127e-08,
+      "loss": 1.5031,
+      "step": 56610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.96884455154886e-08,
+      "loss": 1.3419,
+      "step": 56620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9516643906080935e-08,
+      "loss": 1.4501,
+      "step": 56630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.934559221917709e-08,
+      "loss": 1.3681,
+      "step": 56640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.917529050649042e-08,
+      "loss": 1.4003,
+      "step": 56650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.9005738819508344e-08,
+      "loss": 1.5445,
+      "step": 56660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8836937209491535e-08,
+      "loss": 1.3727,
+      "step": 56670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8668885727473608e-08,
+      "loss": 1.5362,
+      "step": 56680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8501584424261144e-08,
+      "loss": 1.3967,
+      "step": 56690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8335033350434238e-08,
+      "loss": 1.389,
+      "step": 56700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8169232556346228e-08,
+      "loss": 1.3902,
+      "step": 56710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.800418209212368e-08,
+      "loss": 1.4007,
+      "step": 56720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7839882007665842e-08,
+      "loss": 1.4724,
+      "step": 56730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7676332352645765e-08,
+      "loss": 1.42,
+      "step": 56740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7513533176509166e-08,
+      "loss": 1.4329,
+      "step": 56750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.7351484528475016e-08,
+      "loss": 1.4692,
+      "step": 56760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.719018645753495e-08,
+      "loss": 1.4941,
+      "step": 56770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.702963901245497e-08,
+      "loss": 1.4002,
+      "step": 56780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.686984224177235e-08,
+      "loss": 1.3744,
+      "step": 56790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6710796193798727e-08,
+      "loss": 1.3982,
+      "step": 56800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6552500916618142e-08,
+      "loss": 1.4012,
+      "step": 56810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.639495645808814e-08,
+      "loss": 1.3696,
+      "step": 56820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6238162865838956e-08,
+      "loss": 1.5051,
+      "step": 56830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6082120187274064e-08,
+      "loss": 1.3529,
+      "step": 56840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5926828469569055e-08,
+      "loss": 1.3662,
+      "step": 56850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5772287759673876e-08,
+      "loss": 1.4311,
+      "step": 56860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.561849810431032e-08,
+      "loss": 1.3233,
+      "step": 56870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5465459549973692e-08,
+      "loss": 1.4342,
+      "step": 56880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.531317214293171e-08,
+      "loss": 1.3151,
+      "step": 56890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.5161635929225315e-08,
+      "loss": 1.4097,
+      "step": 56900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.50108509546687e-08,
+      "loss": 1.4676,
+      "step": 56910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.486081726484817e-08,
+      "loss": 1.4705,
+      "step": 56920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4711534905123559e-08,
+      "loss": 1.4892,
+      "step": 56930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4563003920627372e-08,
+      "loss": 1.4004,
+      "step": 56940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4415224356264523e-08,
+      "loss": 1.3664,
+      "step": 56950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.4268196256713163e-08,
+      "loss": 1.4591,
+      "step": 56960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.41219196664244e-08,
+      "loss": 1.3903,
+      "step": 56970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3976394629621749e-08,
+      "loss": 1.3952,
+      "step": 56980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3831621190301403e-08,
+      "loss": 1.3514,
+      "step": 56990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3687599392233075e-08,
+      "loss": 1.3644,
+      "step": 57000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3544329278958601e-08,
+      "loss": 1.4596,
+      "step": 57010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.34018108937925e-08,
+      "loss": 1.4346,
+      "step": 57020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3260044279822804e-08,
+      "loss": 1.4213,
+      "step": 57030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3119029479909117e-08,
+      "loss": 1.3629,
+      "step": 57040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2978766536684284e-08,
+      "loss": 1.3947,
+      "step": 57050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2839255492554659e-08,
+      "loss": 1.4391,
+      "step": 57060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2700496389697613e-08,
+      "loss": 1.4129,
+      "step": 57070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2562489270064314e-08,
+      "loss": 1.5273,
+      "step": 57080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2425234175378609e-08,
+      "loss": 1.3548,
+      "step": 57090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2288731147136745e-08,
+      "loss": 1.5393,
+      "step": 57100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2152980226607103e-08,
+      "loss": 1.4161,
+      "step": 57110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2017981454831574e-08,
+      "loss": 1.4248,
+      "step": 57120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1883734872624175e-08,
+      "loss": 1.4417,
+      "step": 57130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1750240520571332e-08,
+      "loss": 1.4527,
+      "step": 57140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1617498439032427e-08,
+      "loss": 1.3645,
+      "step": 57150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1485508668139522e-08,
+      "loss": 1.4699,
+      "step": 57160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.1354271247796533e-08,
+      "loss": 1.3862,
+      "step": 57170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.122378621768061e-08,
+      "loss": 1.4375,
+      "step": 57180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.109405361724103e-08,
+      "loss": 1.3411,
+      "step": 57190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.096507348569975e-08,
+      "loss": 1.3988,
+      "step": 57200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0836845862051415e-08,
+      "loss": 1.5539,
+      "step": 57210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0709370785063066e-08,
+      "loss": 1.4504,
+      "step": 57220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.05826482932736e-08,
+      "loss": 1.51,
+      "step": 57230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0456678424995703e-08,
+      "loss": 1.3703,
+      "step": 57240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0331461218313077e-08,
+      "loss": 1.3746,
+      "step": 57250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0206996711082939e-08,
+      "loss": 1.4715,
+      "step": 57260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.008328494093408e-08,
+      "loss": 1.2609,
+      "step": 57270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.9603259452688e-09,
+      "loss": 1.4638,
+      "step": 57280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.8381197612607e-09,
+      "loss": 1.3758,
+      "step": 57290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.716666425856336e-09,
+      "loss": 1.379,
+      "step": 57300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.595965975774946e-09,
+      "loss": 1.4711,
+      "step": 57310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.476018447507341e-09,
+      "loss": 1.3523,
+      "step": 57320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.356823877317567e-09,
+      "loss": 1.428,
+      "step": 57330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.238382301241521e-09,
+      "loss": 1.3756,
+      "step": 57340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.120693755087229e-09,
+      "loss": 1.4235,
+      "step": 57350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.00375827443567e-09,
+      "loss": 1.4821,
+      "step": 57360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.887575894639678e-09,
+      "loss": 1.3512,
+      "step": 57370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.772146650824764e-09,
+      "loss": 1.5427,
+      "step": 57380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.657470577888571e-09,
+      "loss": 1.4077,
+      "step": 57390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.543547710500866e-09,
+      "loss": 1.4207,
+      "step": 57400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.430378083104096e-09,
+      "loss": 1.4657,
+      "step": 57410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.317961729912283e-09,
+      "loss": 1.4179,
+      "step": 57420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.206298684912684e-09,
+      "loss": 1.4807,
+      "step": 57430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 8.09538898186385e-09,
+      "loss": 1.3826,
+      "step": 57440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.985232654297293e-09,
+      "loss": 1.392,
+      "step": 57450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.875829735516372e-09,
+      "loss": 1.462,
+      "step": 57460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.767180258596851e-09,
+      "loss": 1.3076,
+      "step": 57470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.659284256386901e-09,
+      "loss": 1.5027,
+      "step": 57480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.552141761505981e-09,
+      "loss": 1.3624,
+      "step": 57490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.445752806347073e-09,
+      "loss": 1.3826,
+      "step": 57500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.340117423074167e-09,
+      "loss": 1.4715,
+      "step": 57510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.2352356436242165e-09,
+      "loss": 1.3668,
+      "step": 57520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.131107499705747e-09,
+      "loss": 1.3686,
+      "step": 57530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 7.027733022800243e-09,
+      "loss": 1.3814,
+      "step": 57540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.9251122441602035e-09,
+      "loss": 1.4115,
+      "step": 57550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.823245194811368e-09,
+      "loss": 1.4413,
+      "step": 57560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.722131905550766e-09,
+      "loss": 1.3536,
+      "step": 57570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.62177240694839e-09,
+      "loss": 1.4421,
+      "step": 57580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.522166729345247e-09,
+      "loss": 1.305,
+      "step": 57590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.423314902855582e-09,
+      "loss": 1.4427,
+      "step": 57600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.3252169573646546e-09,
+      "loss": 1.4532,
+      "step": 57610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.227872922530686e-09,
+      "loss": 1.4045,
+      "step": 57620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.131282827783469e-09,
+      "loss": 1.4446,
+      "step": 57630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.035446702325198e-09,
+      "loss": 1.3509,
+      "step": 57640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.940364575129642e-09,
+      "loss": 1.2871,
+      "step": 57650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.846036474943251e-09,
+      "loss": 1.492,
+      "step": 57660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.752462430284045e-09,
+      "loss": 1.3284,
+      "step": 57670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.659642469441895e-09,
+      "loss": 1.4046,
+      "step": 57680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.567576620479354e-09,
+      "loss": 1.3889,
+      "step": 57690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.476264911230544e-09,
+      "loss": 1.3508,
+      "step": 57700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.38570736930144e-09,
+      "loss": 1.4574,
+      "step": 57710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.2959040220704194e-09,
+      "loss": 1.4103,
+      "step": 57720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.20685489668743e-09,
+      "loss": 1.4207,
+      "step": 57730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.118560020075103e-09,
+      "loss": 1.4028,
+      "step": 57740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 5.031019418927085e-09,
+      "loss": 1.3455,
+      "step": 57750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.9442331197097074e-09,
+      "loss": 1.4711,
+      "step": 57760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.858201148660868e-09,
+      "loss": 1.3553,
+      "step": 57770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.772923531790874e-09,
+      "loss": 1.4352,
+      "step": 57780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6884002948810455e-09,
+      "loss": 1.387,
+      "step": 57790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.604631463485665e-09,
+      "loss": 1.4359,
+      "step": 57800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.521617062930306e-09,
+      "loss": 1.4816,
+      "step": 57810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.43935711831267e-09,
+      "loss": 1.3121,
+      "step": 57820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.35785165450231e-09,
+      "loss": 1.412,
+      "step": 57830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.277100696140901e-09,
+      "loss": 1.4221,
+      "step": 57840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.197104267641694e-09,
+      "loss": 1.3573,
+      "step": 57850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.117862393190064e-09,
+      "loss": 1.4403,
+      "step": 57860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.039375096742959e-09,
+      "loss": 1.3488,
+      "step": 57870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.961642402029175e-09,
+      "loss": 1.4661,
+      "step": 57880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.884664332550192e-09,
+      "loss": 1.4631,
+      "step": 57890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.808440911577949e-09,
+      "loss": 1.4069,
+      "step": 57900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.732972162157622e-09,
+      "loss": 1.4587,
+      "step": 57910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6582581071054057e-09,
+      "loss": 1.3847,
+      "step": 57920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5842987690096198e-09,
+      "loss": 1.4324,
+      "step": 57930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.511094170230156e-09,
+      "loss": 1.3583,
+      "step": 57940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4386443328993103e-09,
+      "loss": 1.3674,
+      "step": 57950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3669492789201173e-09,
+      "loss": 1.4236,
+      "step": 57960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.296009029968572e-09,
+      "loss": 1.3254,
+      "step": 57970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2258236074919624e-09,
+      "loss": 1.4862,
+      "step": 57980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1563930327091486e-09,
+      "loss": 1.4167,
+      "step": 57990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0877173266113947e-09,
+      "loss": 1.465,
+      "step": 58000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0197965099609817e-09,
+      "loss": 1.4684,
+      "step": 58010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9526306032923167e-09,
+      "loss": 1.3673,
+      "step": 58020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.886219626911935e-09,
+      "loss": 1.493,
+      "step": 58030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8205636008976634e-09,
+      "loss": 1.4275,
+      "step": 58040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7556625450994578e-09,
+      "loss": 1.3532,
+      "step": 58050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.69151647913829e-09,
+      "loss": 1.491,
+      "step": 58060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.62812542240809e-09,
+      "loss": 1.3681,
+      "step": 58070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.565489394072973e-09,
+      "loss": 1.4243,
+      "step": 58080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.503608413070291e-09,
+      "loss": 1.4008,
+      "step": 58090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4424824981084116e-09,
+      "loss": 1.3921,
+      "step": 58100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.382111667667275e-09,
+      "loss": 1.4524,
+      "step": 58110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.32249593999867e-09,
+      "loss": 1.4743,
+      "step": 58120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.263635333126235e-09,
+      "loss": 1.5287,
+      "step": 58130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2055298648457346e-09,
+      "loss": 1.4236,
+      "step": 58140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.148179552723395e-09,
+      "loss": 1.2725,
+      "step": 58150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.091584414098402e-09,
+      "loss": 1.3702,
+      "step": 58160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0357444660806802e-09,
+      "loss": 1.4433,
+      "step": 58170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.980659725552836e-09,
+      "loss": 1.4171,
+      "step": 58180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9263302091682147e-09,
+      "loss": 1.3991,
+      "step": 58190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8727559333522883e-09,
+      "loss": 1.4246,
+      "step": 58200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8199369143021007e-09,
+      "loss": 1.5009,
+      "step": 58210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7678731679865447e-09,
+      "loss": 1.4233,
+      "step": 58220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7165647101458071e-09,
+      "loss": 1.3676,
+      "step": 58230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6660115562919243e-09,
+      "loss": 1.3658,
+      "step": 58240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6162137217087814e-09,
+      "loss": 1.3769,
+      "step": 58250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5671712214515578e-09,
+      "loss": 1.4095,
+      "step": 58260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5188840703475593e-09,
+      "loss": 1.3931,
+      "step": 58270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4713522829948313e-09,
+      "loss": 1.4091,
+      "step": 58280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4245758737638226e-09,
+      "loss": 1.4757,
+      "step": 58290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.378554856796832e-09,
+      "loss": 1.4069,
+      "step": 58300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3332892460068969e-09,
+      "loss": 1.4694,
+      "step": 58310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2887790550791812e-09,
+      "loss": 1.3321,
+      "step": 58320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2450242974704207e-09,
+      "loss": 1.4775,
+      "step": 58330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2020249864092003e-09,
+      "loss": 1.431,
+      "step": 58340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1597811348951217e-09,
+      "loss": 1.3483,
+      "step": 58350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1182927557001898e-09,
+      "loss": 1.5254,
+      "step": 58360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.07755986136715e-09,
+      "loss": 1.3918,
+      "step": 58370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0375824642105959e-09,
+      "loss": 1.4512,
+      "step": 58380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.983605763175252e-10,
+      "loss": 1.4092,
+      "step": 58390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.598942095451202e-10,
+      "loss": 1.3896,
+      "step": 58400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.221833755235221e-10,
+      "loss": 1.4506,
+      "step": 58410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.852280856533335e-10,
+      "loss": 1.3747,
+      "step": 58420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.490283511075614e-10,
+      "loss": 1.4792,
+      "step": 58430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.135841828302293e-10,
+      "loss": 1.3837,
+      "step": 58440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.788955915372099e-10,
+      "loss": 1.3576,
+      "step": 58450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.449625877156696e-10,
+      "loss": 1.4455,
+      "step": 58460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.117851816249021e-10,
+      "loss": 1.435,
+      "step": 58470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.793633832952174e-10,
+      "loss": 1.4739,
+      "step": 58480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.476972025287742e-10,
+      "loss": 1.4741,
+      "step": 58490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.167866488993035e-10,
+      "loss": 1.3566,
+      "step": 58500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.866317317515524e-10,
+      "loss": 1.4521,
+      "step": 58510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.572324602023948e-10,
+      "loss": 1.4098,
+      "step": 58520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.285888431399988e-10,
+      "loss": 1.4618,
+      "step": 58530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.00700889224659e-10,
+      "loss": 1.4404,
+      "step": 58540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.735686068871314e-10,
+      "loss": 1.3031,
+      "step": 58550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.471920043308542e-10,
+      "loss": 1.4591,
+      "step": 58560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.215710895297265e-10,
+      "loss": 1.4331,
+      "step": 58570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9670587022977436e-10,
+      "loss": 1.4406,
+      "step": 58580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7259635394887304e-10,
+      "loss": 1.3909,
+      "step": 58590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.492425479756367e-10,
+      "loss": 1.439,
+      "step": 58600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2664445937080624e-10,
+      "loss": 1.5251,
+      "step": 58610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.04802094966139e-10,
+      "loss": 1.3497,
+      "step": 58620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.837154613655191e-10,
+      "loss": 1.4515,
+      "step": 58630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6338456494412466e-10,
+      "loss": 1.328,
+      "step": 58640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.43809411848428e-10,
+      "loss": 1.447,
+      "step": 58650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2499000799647287e-10,
+      "loss": 1.4212,
+      "step": 58660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0692635907815228e-10,
+      "loss": 1.4083,
+      "step": 58670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8961847055437577e-10,
+      "loss": 1.5401,
+      "step": 58680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7306634765790197e-10,
+      "loss": 1.3215,
+      "step": 58690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.572699953930612e-10,
+      "loss": 1.4257,
+      "step": 58700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4222941853547777e-10,
+      "loss": 1.4124,
+      "step": 58710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2794462163234765e-10,
+      "loss": 1.4234,
+      "step": 58720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1441560900243842e-10,
+      "loss": 1.4504,
+      "step": 58730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0164238473581167e-10,
+      "loss": 1.4112,
+      "step": 58740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.962495269410065e-11,
+      "loss": 1.5037,
+      "step": 58750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.836331651078777e-11,
+      "loss": 1.4763,
+      "step": 58760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.785747959064948e-11,
+      "loss": 1.3698,
+      "step": 58770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.81074451097563e-11,
+      "loss": 1.42,
+      "step": 58780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.9113216015750366e-11,
+      "loss": 1.4119,
+      "step": 58790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.087479502784541e-11,
+      "loss": 1.4089,
+      "step": 58800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3392184637104364e-11,
+      "loss": 1.5342,
+      "step": 58810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6665387105329066e-11,
+      "loss": 1.3669,
+      "step": 58820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0694404466170547e-11,
+      "loss": 1.4288,
+      "step": 58830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.547923852540656e-11,
+      "loss": 1.3959,
+      "step": 58840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1019890858998683e-11,
+      "loss": 1.4089,
+      "step": 58850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.31636281559034e-12,
+      "loss": 1.4896,
+      "step": 58860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.368655515119002e-12,
+      "loss": 1.425,
+      "step": 58870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1767698479835398e-12,
+      "loss": 1.4598,
+      "step": 58880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.407064775422079e-13,
+      "loss": 1.3851,
+      "step": 58890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.046583761465385e-14,
+      "loss": 1.4119,
+      "step": 58900
+    },
+    {
+      "epoch": 1.0,
+      "step": 58904,
+      "total_flos": 9441864279523328.0,
+      "train_loss": 1.4414458279212694,
+      "train_runtime": 115932.9754,
+      "train_samples_per_second": 16.259,
+      "train_steps_per_second": 0.508
+    }
+  ],
+  "max_steps": 58904,
+  "num_train_epochs": 1,
+  "total_flos": 9441864279523328.0,
+  "trial_name": null,
+  "trial_params": null
+}