diff --git "a/checkpoint-1176/trainer_state.json" "b/checkpoint-1176/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-1176/trainer_state.json"
@@ -0,0 +1,11214 @@
+{
+  "best_metric": 0.8602466236054023,
+  "best_model_checkpoint": "/tmp/logs/binary_classification_model_v3.1.5_Junction_batch1/checkpoint-1176",
+  "epoch": 196.0,
+  "eval_steps": 500,
+  "global_step": 1176,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.16666666666666666,
+      "grad_norm": 301205.78125,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.701,
+      "step": 1
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 336849.3125,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.7097,
+      "step": 2
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 127303.109375,
+      "learning_rate": 2.5e-05,
+      "loss": 0.6876,
+      "step": 3
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 234841.890625,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.6984,
+      "step": 4
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 540732.1875,
+      "learning_rate": 4.1666666666666665e-05,
+      "loss": 0.7388,
+      "step": 5
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 565904.6875,
+      "learning_rate": 5e-05,
+      "loss": 0.7434,
+      "step": 6
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.6405995438253502,
+      "eval_best_threshold": 0.4179820716381073,
+      "eval_f1": 0.7283920216695395,
+      "eval_loss": 0.7047091722488403,
+      "eval_pr_auc": 0.6283352362755803,
+      "eval_precision": 0.6258992805755396,
+      "eval_recall": 0.8710247349823321,
+      "eval_roc_auc": 0.6278743001377173,
+      "eval_runtime": 2.4479,
+      "eval_samples_per_second": 1253.729,
+      "eval_steps_per_second": 2.451,
+      "step": 6
+    },
+    {
+      "epoch": 1.1666666666666667,
+      "grad_norm": 371967.21875,
+      "learning_rate": 5.833333333333333e-05,
+      "loss": 0.7082,
+      "step": 7
+    },
+    {
+      "epoch": 1.3333333333333333,
+      "grad_norm": 43612.97265625,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.6722,
+      "step": 8
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 397443.65625,
+      "learning_rate": 7.5e-05,
+      "loss": 0.7188,
+      "step": 9
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 525898.0,
+      "learning_rate": 8.333333333333333e-05,
+      "loss": 0.7528,
+      "step": 10
+    },
+    {
+      "epoch": 1.8333333333333335,
+      "grad_norm": 535364.5,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 0.7645,
+      "step": 11
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 310039.46875,
+      "learning_rate": 0.0001,
+      "loss": 0.7025,
+      "step": 12
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.6298468556533072,
+      "eval_best_threshold": 0.4796014130115509,
+      "eval_f1": 0.7262650602409638,
+      "eval_loss": 0.6673709154129028,
+      "eval_pr_auc": 0.6426689105421621,
+      "eval_precision": 0.6146003262642741,
+      "eval_recall": 0.8875147232037691,
+      "eval_roc_auc": 0.6411069271868306,
+      "eval_runtime": 2.5096,
+      "eval_samples_per_second": 1222.919,
+      "eval_steps_per_second": 2.391,
+      "step": 12
+    },
+    {
+      "epoch": 2.1666666666666665,
+      "grad_norm": 79163.5078125,
+      "learning_rate": 0.00010833333333333334,
+      "loss": 0.666,
+      "step": 13
+    },
+    {
+      "epoch": 2.3333333333333335,
+      "grad_norm": 238443.5625,
+      "learning_rate": 0.00011666666666666667,
+      "loss": 0.6792,
+      "step": 14
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 89458.0234375,
+      "learning_rate": 0.000125,
+      "loss": 0.6711,
+      "step": 15
+    },
+    {
+      "epoch": 2.6666666666666665,
+      "grad_norm": 56144.84765625,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 0.6595,
+      "step": 16
+    },
+    {
+      "epoch": 2.8333333333333335,
+      "grad_norm": 152777.34375,
+      "learning_rate": 0.00014166666666666668,
+      "loss": 0.6647,
+      "step": 17
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 20701.99609375,
+      "learning_rate": 0.00015,
+      "loss": 0.6524,
+      "step": 18
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.6888237210817856,
+      "eval_best_threshold": 0.4486621618270874,
+      "eval_f1": 0.7586555471316654,
+      "eval_loss": 0.6511502265930176,
+      "eval_pr_auc": 0.6680483472768871,
+      "eval_precision": 0.6644532979194334,
+      "eval_recall": 0.8839811542991755,
+      "eval_roc_auc": 0.683494934186957,
+      "eval_runtime": 2.5522,
+      "eval_samples_per_second": 1202.471,
+      "eval_steps_per_second": 2.351,
+      "step": 18
+    },
+    {
+      "epoch": 3.1666666666666665,
+      "grad_norm": 139181.78125,
+      "learning_rate": 0.00015833333333333332,
+      "loss": 0.6525,
+      "step": 19
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 117104.03125,
+      "learning_rate": 0.00016666666666666666,
+      "loss": 0.6531,
+      "step": 20
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 93101.1171875,
+      "learning_rate": 0.000175,
+      "loss": 0.6425,
+      "step": 21
+    },
+    {
+      "epoch": 3.6666666666666665,
+      "grad_norm": 467410.875,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 0.68,
+      "step": 22
+    },
+    {
+      "epoch": 3.8333333333333335,
+      "grad_norm": 552058.625,
+      "learning_rate": 0.00019166666666666667,
+      "loss": 0.6909,
+      "step": 23
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 75542.5625,
+      "learning_rate": 0.0002,
+      "loss": 0.644,
+      "step": 24
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.6940371456500489,
+      "eval_best_threshold": 0.6427821516990662,
+      "eval_f1": 0.7617356001014971,
+      "eval_loss": 0.730407178401947,
+      "eval_pr_auc": 0.6693287247885799,
+      "eval_precision": 0.669193045028979,
+      "eval_recall": 0.8839811542991755,
+      "eval_roc_auc": 0.6848871414346822,
+      "eval_runtime": 2.8443,
+      "eval_samples_per_second": 1079.0,
+      "eval_steps_per_second": 2.109,
+      "step": 24
+    },
+    {
+      "epoch": 4.166666666666667,
+      "grad_norm": 683070.9375,
+      "learning_rate": 0.00020833333333333335,
+      "loss": 0.7293,
+      "step": 25
+    },
+    {
+      "epoch": 4.333333333333333,
+      "grad_norm": 651965.375,
+      "learning_rate": 0.00021666666666666668,
+      "loss": 0.7538,
+      "step": 26
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 410760.625,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 0.6788,
+      "step": 27
+    },
+    {
+      "epoch": 4.666666666666667,
+      "grad_norm": 15163.0390625,
+      "learning_rate": 0.00023333333333333333,
+      "loss": 0.6325,
+      "step": 28
+    },
+    {
+      "epoch": 4.833333333333333,
+      "grad_norm": 611296.9375,
+      "learning_rate": 0.00024166666666666667,
+      "loss": 0.7411,
+      "step": 29
+    },
+    {
+      "epoch": 5.0,
+      "grad_norm": 758565.25,
+      "learning_rate": 0.00025,
+      "loss": 0.825,
+      "step": 30
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.6979472140762464,
+      "eval_best_threshold": 0.23702159523963928,
+      "eval_f1": 0.7587822014051522,
+      "eval_loss": 0.7560154795646667,
+      "eval_pr_auc": 0.668429405093208,
+      "eval_precision": 0.6797202797202797,
+      "eval_recall": 0.8586572438162544,
+      "eval_roc_auc": 0.6841386313670608,
+      "eval_runtime": 2.6473,
+      "eval_samples_per_second": 1159.289,
+      "eval_steps_per_second": 2.266,
+      "step": 30
+    },
+    {
+      "epoch": 5.166666666666667,
+      "grad_norm": 616902.5625,
+      "learning_rate": 0.00025833333333333334,
+      "loss": 0.7771,
+      "step": 31
+    },
+    {
+      "epoch": 5.333333333333333,
+      "grad_norm": 274811.5625,
+      "learning_rate": 0.0002666666666666667,
+      "loss": 0.665,
+      "step": 32
+    },
+    {
+      "epoch": 5.5,
+      "grad_norm": 98389.8046875,
+      "learning_rate": 0.000275,
+      "loss": 0.6466,
+      "step": 33
+    },
+    {
+      "epoch": 5.666666666666667,
+      "grad_norm": 208001.3125,
+      "learning_rate": 0.00028333333333333335,
+      "loss": 0.668,
+      "step": 34
+    },
+    {
+      "epoch": 5.833333333333333,
+      "grad_norm": 151922.03125,
+      "learning_rate": 0.0002916666666666667,
+      "loss": 0.65,
+      "step": 35
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 16863.15234375,
+      "learning_rate": 0.0003,
+      "loss": 0.6409,
+      "step": 36
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.6976213750407299,
+      "eval_best_threshold": 0.4489557445049286,
+      "eval_f1": 0.759958613554061,
+      "eval_loss": 0.634763240814209,
+      "eval_pr_auc": 0.6705264493262868,
+      "eval_precision": 0.6775830258302583,
+      "eval_recall": 0.8651354534746761,
+      "eval_roc_auc": 0.6867211521857353,
+      "eval_runtime": 2.6393,
+      "eval_samples_per_second": 1162.804,
+      "eval_steps_per_second": 2.273,
+      "step": 36
+    },
+    {
+      "epoch": 6.166666666666667,
+      "grad_norm": 97510.015625,
+      "learning_rate": 0.00030833333333333337,
+      "loss": 0.6443,
+      "step": 37
+    },
+    {
+      "epoch": 6.333333333333333,
+      "grad_norm": 27098.69140625,
+      "learning_rate": 0.00031666666666666665,
+      "loss": 0.6299,
+      "step": 38
+    },
+    {
+      "epoch": 6.5,
+      "grad_norm": 123789.1640625,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.6373,
+      "step": 39
+    },
+    {
+      "epoch": 6.666666666666667,
+      "grad_norm": 149780.671875,
+      "learning_rate": 0.0003333333333333333,
+      "loss": 0.629,
+      "step": 40
+    },
+    {
+      "epoch": 6.833333333333333,
+      "grad_norm": 91179.4140625,
+      "learning_rate": 0.00034166666666666666,
+      "loss": 0.6251,
+      "step": 41
+    },
+    {
+      "epoch": 7.0,
+      "grad_norm": 95404.40625,
+      "learning_rate": 0.00035,
+      "loss": 0.6469,
+      "step": 42
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.696969696969697,
+      "eval_best_threshold": 0.36696115136146545,
+      "eval_f1": 0.7593167701863354,
+      "eval_loss": 0.6393815279006958,
+      "eval_pr_auc": 0.6808847453038372,
+      "eval_precision": 0.6772853185595568,
+      "eval_recall": 0.8639575971731449,
+      "eval_roc_auc": 0.6971218982473051,
+      "eval_runtime": 2.6949,
+      "eval_samples_per_second": 1138.828,
+      "eval_steps_per_second": 2.226,
+      "step": 42
+    },
+    {
+      "epoch": 7.166666666666667,
+      "grad_norm": 155450.640625,
+      "learning_rate": 0.00035833333333333333,
+      "loss": 0.6309,
+      "step": 43
+    },
+    {
+      "epoch": 7.333333333333333,
+      "grad_norm": 117154.0625,
+      "learning_rate": 0.00036666666666666667,
+      "loss": 0.6391,
+      "step": 44
+    },
+    {
+      "epoch": 7.5,
+      "grad_norm": 18065.970703125,
+      "learning_rate": 0.000375,
+      "loss": 0.617,
+      "step": 45
+    },
+    {
+      "epoch": 7.666666666666667,
+      "grad_norm": 33959.37890625,
+      "learning_rate": 0.00038333333333333334,
+      "loss": 0.6206,
+      "step": 46
+    },
+    {
+      "epoch": 7.833333333333333,
+      "grad_norm": 19164.263671875,
+      "learning_rate": 0.0003916666666666667,
+      "loss": 0.6177,
+      "step": 47
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 100220.8515625,
+      "learning_rate": 0.0004,
+      "loss": 0.6098,
+      "step": 48
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.68198110133594,
+      "eval_best_threshold": 0.328480988740921,
+      "eval_f1": 0.7661715380929564,
+      "eval_loss": 0.6046918034553528,
+      "eval_pr_auc": 0.6817425061021674,
+      "eval_precision": 0.6457996768982229,
+      "eval_recall": 0.941696113074205,
+      "eval_roc_auc": 0.701265658572878,
+      "eval_runtime": 2.6753,
+      "eval_samples_per_second": 1147.147,
+      "eval_steps_per_second": 2.243,
+      "step": 48
+    },
+    {
+      "epoch": 8.166666666666666,
+      "grad_norm": 16806.453125,
+      "learning_rate": 0.00040833333333333336,
+      "loss": 0.6062,
+      "step": 49
+    },
+    {
+      "epoch": 8.333333333333334,
+      "grad_norm": 124528.859375,
+      "learning_rate": 0.0004166666666666667,
+      "loss": 0.6257,
+      "step": 50
+    },
+    {
+      "epoch": 8.5,
+      "grad_norm": 153063.265625,
+      "learning_rate": 0.000425,
+      "loss": 0.61,
+      "step": 51
+    },
+    {
+      "epoch": 8.666666666666666,
+      "grad_norm": 78803.90625,
+      "learning_rate": 0.00043333333333333337,
+      "loss": 0.621,
+      "step": 52
+    },
+    {
+      "epoch": 8.833333333333334,
+      "grad_norm": 91850.375,
+      "learning_rate": 0.00044166666666666665,
+      "loss": 0.6079,
+      "step": 53
+    },
+    {
+      "epoch": 9.0,
+      "grad_norm": 169883.046875,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.6332,
+      "step": 54
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.6862170087976539,
+      "eval_best_threshold": 0.37674590945243835,
+      "eval_f1": 0.7696723271944511,
+      "eval_loss": 0.6066368818283081,
+      "eval_pr_auc": 0.674455454194375,
+      "eval_precision": 0.6480064438179621,
+      "eval_recall": 0.947585394581861,
+      "eval_roc_auc": 0.701175880320865,
+      "eval_runtime": 2.755,
+      "eval_samples_per_second": 1113.956,
+      "eval_steps_per_second": 2.178,
+      "step": 54
+    },
+    {
+      "epoch": 9.166666666666666,
+      "grad_norm": 117117.21875,
+      "learning_rate": 0.0004583333333333333,
+      "loss": 0.6193,
+      "step": 55
+    },
+    {
+      "epoch": 9.333333333333334,
+      "grad_norm": 38909.0,
+      "learning_rate": 0.00046666666666666666,
+      "loss": 0.6097,
+      "step": 56
+    },
+    {
+      "epoch": 9.5,
+      "grad_norm": 69082.0234375,
+      "learning_rate": 0.000475,
+      "loss": 0.5995,
+      "step": 57
+    },
+    {
+      "epoch": 9.666666666666666,
+      "grad_norm": 11782.515625,
+      "learning_rate": 0.00048333333333333334,
+      "loss": 0.6056,
+      "step": 58
+    },
+    {
+      "epoch": 9.833333333333334,
+      "grad_norm": 299948.0625,
+      "learning_rate": 0.0004916666666666666,
+      "loss": 0.6576,
+      "step": 59
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 461681.4375,
+      "learning_rate": 0.0005,
+      "loss": 0.7297,
+      "step": 60
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.671228413163897,
+      "eval_best_threshold": 0.13465377688407898,
+      "eval_f1": 0.7627557018575123,
+      "eval_loss": 0.6625670790672302,
+      "eval_pr_auc": 0.6769399620865038,
+      "eval_precision": 0.6348336594911937,
+      "eval_recall": 0.9552414605418139,
+      "eval_roc_auc": 0.7029843321915601,
+      "eval_runtime": 3.0432,
+      "eval_samples_per_second": 1008.479,
+      "eval_steps_per_second": 1.972,
+      "step": 60
+    },
+    {
+      "epoch": 10.166666666666666,
+      "grad_norm": 323015.5,
+      "learning_rate": 0.0005083333333333333,
+      "loss": 0.6792,
+      "step": 61
+    },
+    {
+      "epoch": 10.333333333333334,
+      "grad_norm": 41346.72265625,
+      "learning_rate": 0.0005166666666666667,
+      "loss": 0.5971,
+      "step": 62
+    },
+    {
+      "epoch": 10.5,
+      "grad_norm": 165735.25,
+      "learning_rate": 0.0005250000000000001,
+      "loss": 0.624,
+      "step": 63
+    },
+    {
+      "epoch": 10.666666666666666,
+      "grad_norm": 185497.203125,
+      "learning_rate": 0.0005333333333333334,
+      "loss": 0.6514,
+      "step": 64
+    },
+    {
+      "epoch": 10.833333333333334,
+      "grad_norm": 153241.40625,
+      "learning_rate": 0.0005416666666666666,
+      "loss": 0.6562,
+      "step": 65
+    },
+    {
+      "epoch": 11.0,
+      "grad_norm": 57239.83203125,
+      "learning_rate": 0.00055,
+      "loss": 0.6358,
+      "step": 66
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.6992505702183122,
+      "eval_best_threshold": 0.5167019963264465,
+      "eval_f1": 0.7631511419040288,
+      "eval_loss": 0.6300772428512573,
+      "eval_pr_auc": 0.6678537551436834,
+      "eval_precision": 0.6762164620281946,
+      "eval_recall": 0.875736160188457,
+      "eval_roc_auc": 0.6964240763793849,
+      "eval_runtime": 2.8763,
+      "eval_samples_per_second": 1067.013,
+      "eval_steps_per_second": 2.086,
+      "step": 66
+    },
+    {
+      "epoch": 11.166666666666666,
+      "grad_norm": 76762.9453125,
+      "learning_rate": 0.0005583333333333333,
+      "loss": 0.6305,
+      "step": 67
+    },
+    {
+      "epoch": 11.333333333333334,
+      "grad_norm": 20210.671875,
+      "learning_rate": 0.0005666666666666667,
+      "loss": 0.6153,
+      "step": 68
+    },
+    {
+      "epoch": 11.5,
+      "grad_norm": 59593.27734375,
+      "learning_rate": 0.000575,
+      "loss": 0.6065,
+      "step": 69
+    },
+    {
+      "epoch": 11.666666666666666,
+      "grad_norm": 11676.9892578125,
+      "learning_rate": 0.0005833333333333334,
+      "loss": 0.6214,
+      "step": 70
+    },
+    {
+      "epoch": 11.833333333333334,
+      "grad_norm": 198962.015625,
+      "learning_rate": 0.0005916666666666667,
+      "loss": 0.634,
+      "step": 71
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 289025.59375,
+      "learning_rate": 0.0006,
+      "loss": 0.6703,
+      "step": 72
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.6682958618442489,
+      "eval_best_threshold": 0.1743956208229065,
+      "eval_f1": 0.7651130595293032,
+      "eval_loss": 0.6442840695381165,
+      "eval_pr_auc": 0.6679981832738308,
+      "eval_precision": 0.6289833080424886,
+      "eval_recall": 0.9764428739693758,
+      "eval_roc_auc": 0.6914338660749034,
+      "eval_runtime": 3.0962,
+      "eval_samples_per_second": 991.223,
+      "eval_steps_per_second": 1.938,
+      "step": 72
+    },
+    {
+      "epoch": 12.166666666666666,
+      "grad_norm": 264987.90625,
+      "learning_rate": 0.0006083333333333333,
+      "loss": 0.6585,
+      "step": 73
+    },
+    {
+      "epoch": 12.333333333333334,
+      "grad_norm": 23771.486328125,
+      "learning_rate": 0.0006166666666666667,
+      "loss": 0.5975,
+      "step": 74
+    },
+    {
+      "epoch": 12.5,
+      "grad_norm": 299908.5,
+      "learning_rate": 0.000625,
+      "loss": 0.6949,
+      "step": 75
+    },
+    {
+      "epoch": 12.666666666666666,
+      "grad_norm": 361077.90625,
+      "learning_rate": 0.0006333333333333333,
+      "loss": 0.7435,
+      "step": 76
+    },
+    {
+      "epoch": 12.833333333333334,
+      "grad_norm": 336040.4375,
+      "learning_rate": 0.0006416666666666667,
+      "loss": 0.7378,
+      "step": 77
+    },
+    {
+      "epoch": 13.0,
+      "grad_norm": 272416.1875,
+      "learning_rate": 0.0006500000000000001,
+      "loss": 0.6772,
+      "step": 78
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.6855653307266211,
+      "eval_best_threshold": 0.5437505841255188,
+      "eval_f1": 0.7697446910045336,
+      "eval_loss": 0.6109954714775085,
+      "eval_pr_auc": 0.6833836331698967,
+      "eval_precision": 0.6470116325711993,
+      "eval_recall": 0.9499411071849234,
+      "eval_roc_auc": 0.7098822229610673,
+      "eval_runtime": 3.0708,
+      "eval_samples_per_second": 999.429,
+      "eval_steps_per_second": 1.954,
+      "step": 78
+    },
+    {
+      "epoch": 13.166666666666666,
+      "grad_norm": 138968.046875,
+      "learning_rate": 0.0006583333333333334,
+      "loss": 0.6236,
+      "step": 79
+    },
+    {
+      "epoch": 13.333333333333334,
+      "grad_norm": 133513.9375,
+      "learning_rate": 0.0006666666666666666,
+      "loss": 0.6264,
+      "step": 80
+    },
+    {
+      "epoch": 13.5,
+      "grad_norm": 265727.65625,
+      "learning_rate": 0.000675,
+      "loss": 0.6646,
+      "step": 81
+    },
+    {
+      "epoch": 13.666666666666666,
+      "grad_norm": 326814.34375,
+      "learning_rate": 0.0006833333333333333,
+      "loss": 0.688,
+      "step": 82
+    },
+    {
+      "epoch": 13.833333333333334,
+      "grad_norm": 237995.25,
+      "learning_rate": 0.0006916666666666667,
+      "loss": 0.6591,
+      "step": 83
+    },
+    {
+      "epoch": 14.0,
+      "grad_norm": 130573.796875,
+      "learning_rate": 0.0007,
+      "loss": 0.6188,
+      "step": 84
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.696969696969697,
+      "eval_best_threshold": 0.5145177245140076,
+      "eval_f1": 0.7700296735905044,
+      "eval_loss": 0.6013315916061401,
+      "eval_pr_auc": 0.6843879067612049,
+      "eval_precision": 0.6636828644501279,
+      "eval_recall": 0.9169611307420494,
+      "eval_roc_auc": 0.7103317585626545,
+      "eval_runtime": 3.0866,
+      "eval_samples_per_second": 994.308,
+      "eval_steps_per_second": 1.944,
+      "step": 84
+    },
+    {
+      "epoch": 14.166666666666666,
+      "grad_norm": 38961.95703125,
+      "learning_rate": 0.0007083333333333334,
+      "loss": 0.6077,
+      "step": 85
+    },
+    {
+      "epoch": 14.333333333333334,
+      "grad_norm": 114087.984375,
+      "learning_rate": 0.0007166666666666667,
+      "loss": 0.6092,
+      "step": 86
+    },
+    {
+      "epoch": 14.5,
+      "grad_norm": 85208.2265625,
+      "learning_rate": 0.000725,
+      "loss": 0.6075,
+      "step": 87
+    },
+    {
+      "epoch": 14.666666666666666,
+      "grad_norm": 6441.580078125,
+      "learning_rate": 0.0007333333333333333,
+      "loss": 0.5931,
+      "step": 88
+    },
+    {
+      "epoch": 14.833333333333334,
+      "grad_norm": 107248.140625,
+      "learning_rate": 0.0007416666666666667,
+      "loss": 0.6044,
+      "step": 89
+    },
+    {
+      "epoch": 15.0,
+      "grad_norm": 164448.296875,
+      "learning_rate": 0.00075,
+      "loss": 0.6059,
+      "step": 90
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.6881720430107527,
+      "eval_best_threshold": 0.324313223361969,
+      "eval_f1": 0.7685610640870617,
+      "eval_loss": 0.5854882001876831,
+      "eval_pr_auc": 0.6890732725571252,
+      "eval_precision": 0.652031185884284,
+      "eval_recall": 0.9358068315665489,
+      "eval_roc_auc": 0.7133827156675507,
+      "eval_runtime": 3.0825,
+      "eval_samples_per_second": 995.637,
+      "eval_steps_per_second": 1.947,
+      "step": 90
+    },
+    {
+      "epoch": 15.166666666666666,
+      "grad_norm": 34315.7265625,
+      "learning_rate": 0.0007583333333333333,
+      "loss": 0.5898,
+      "step": 91
+    },
+    {
+      "epoch": 15.333333333333334,
+      "grad_norm": 98205.3828125,
+      "learning_rate": 0.0007666666666666667,
+      "loss": 0.5922,
+      "step": 92
+    },
+    {
+      "epoch": 15.5,
+      "grad_norm": 131727.921875,
+      "learning_rate": 0.0007750000000000001,
+      "loss": 0.6055,
+      "step": 93
+    },
+    {
+      "epoch": 15.666666666666666,
+      "grad_norm": 64623.3515625,
+      "learning_rate": 0.0007833333333333334,
+      "loss": 0.5943,
+      "step": 94
+    },
+    {
+      "epoch": 15.833333333333334,
+      "grad_norm": 412366.34375,
+      "learning_rate": 0.0007916666666666666,
+      "loss": 0.6883,
+      "step": 95
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 377649.59375,
+      "learning_rate": 0.0008,
+      "loss": 0.6986,
+      "step": 96
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.7015314434669273,
+      "eval_best_threshold": 0.3653980791568756,
+      "eval_f1": 0.7690368129097327,
+      "eval_loss": 0.6090668439865112,
+      "eval_pr_auc": 0.6800774346982651,
+      "eval_precision": 0.6723985890652557,
+      "eval_recall": 0.8981154299175501,
+      "eval_roc_auc": 0.7118375846986931,
+      "eval_runtime": 2.9665,
+      "eval_samples_per_second": 1034.567,
+      "eval_steps_per_second": 2.023,
+      "step": 96
+    },
+    {
+      "epoch": 16.166666666666668,
+      "grad_norm": 162461.421875,
+      "learning_rate": 0.0008083333333333333,
+      "loss": 0.6176,
+      "step": 97
+    },
+    {
+      "epoch": 16.333333333333332,
+      "grad_norm": 40319.34765625,
+      "learning_rate": 0.0008166666666666667,
+      "loss": 0.6023,
+      "step": 98
+    },
+    {
+      "epoch": 16.5,
+      "grad_norm": 56993.296875,
+      "learning_rate": 0.000825,
+      "loss": 0.6056,
+      "step": 99
+    },
+    {
+      "epoch": 16.666666666666668,
+      "grad_norm": 33296.796875,
+      "learning_rate": 0.0008333333333333334,
+      "loss": 0.6037,
+      "step": 100
+    },
+    {
+      "epoch": 16.833333333333332,
+      "grad_norm": 137979.578125,
+      "learning_rate": 0.0008416666666666667,
+      "loss": 0.598,
+      "step": 101
+    },
+    {
+      "epoch": 17.0,
+      "grad_norm": 180135.84375,
+      "learning_rate": 0.00085,
+      "loss": 0.6112,
+      "step": 102
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.6950146627565983,
+      "eval_best_threshold": 0.2957038879394531,
+      "eval_f1": 0.7717073170731708,
+      "eval_loss": 0.582582950592041,
+      "eval_pr_auc": 0.6880525636282292,
+      "eval_precision": 0.6586178184845962,
+      "eval_recall": 0.9316843345111896,
+      "eval_roc_auc": 0.7159605113150667,
+      "eval_runtime": 3.1459,
+      "eval_samples_per_second": 975.563,
+      "eval_steps_per_second": 1.907,
+      "step": 102
+    },
+    {
+      "epoch": 17.166666666666668,
+      "grad_norm": 58858.67578125,
+      "learning_rate": 0.0008583333333333333,
+      "loss": 0.5902,
+      "step": 103
+    },
+    {
+      "epoch": 17.333333333333332,
+      "grad_norm": 116587.671875,
+      "learning_rate": 0.0008666666666666667,
+      "loss": 0.605,
+      "step": 104
+    },
+    {
+      "epoch": 17.5,
+      "grad_norm": 176699.15625,
+      "learning_rate": 0.000875,
+      "loss": 0.6269,
+      "step": 105
+    },
+    {
+      "epoch": 17.666666666666668,
+      "grad_norm": 132956.703125,
+      "learning_rate": 0.0008833333333333333,
+      "loss": 0.6156,
+      "step": 106
+    },
+    {
+      "epoch": 17.833333333333332,
+      "grad_norm": 26539.21875,
+      "learning_rate": 0.0008916666666666667,
+      "loss": 0.584,
+      "step": 107
+    },
+    {
+      "epoch": 18.0,
+      "grad_norm": 279153.15625,
+      "learning_rate": 0.0009000000000000001,
+      "loss": 0.6648,
+      "step": 108
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.6702508960573477,
+      "eval_best_threshold": 0.08056089282035828,
+      "eval_f1": 0.7634408602150538,
+      "eval_loss": 0.753210186958313,
+      "eval_pr_auc": 0.6648919074173523,
+      "eval_precision": 0.6329457364341086,
+      "eval_recall": 0.9617196702002355,
+      "eval_roc_auc": 0.6946051002638364,
+      "eval_runtime": 3.1237,
+      "eval_samples_per_second": 982.501,
+      "eval_steps_per_second": 1.921,
+      "step": 108
+    },
+    {
+      "epoch": 18.166666666666668,
+      "grad_norm": 493320.5625,
+      "learning_rate": 0.0009083333333333334,
+      "loss": 0.776,
+      "step": 109
+    },
+    {
+      "epoch": 18.333333333333332,
+      "grad_norm": 441418.5625,
+      "learning_rate": 0.0009166666666666666,
+      "loss": 0.7381,
+      "step": 110
+    },
+    {
+      "epoch": 18.5,
+      "grad_norm": 206490.640625,
+      "learning_rate": 0.000925,
+      "loss": 0.6299,
+      "step": 111
+    },
+    {
+      "epoch": 18.666666666666668,
+      "grad_norm": 41816.65625,
+      "learning_rate": 0.0009333333333333333,
+      "loss": 0.6002,
+      "step": 112
+    },
+    {
+      "epoch": 18.833333333333332,
+      "grad_norm": 163746.375,
+      "learning_rate": 0.0009416666666666667,
+      "loss": 0.6387,
+      "step": 113
+    },
+    {
+      "epoch": 19.0,
+      "grad_norm": 222226.390625,
+      "learning_rate": 0.00095,
+      "loss": 0.6646,
+      "step": 114
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.703486477680026,
+      "eval_best_threshold": 0.7201119661331177,
+      "eval_f1": 0.7642487046632125,
+      "eval_loss": 0.657902717590332,
+      "eval_pr_auc": 0.6878116517095747,
+      "eval_precision": 0.6822386679000925,
+      "eval_recall": 0.8686690223792697,
+      "eval_roc_auc": 0.7094689852651981,
+      "eval_runtime": 3.1363,
+      "eval_samples_per_second": 978.534,
+      "eval_steps_per_second": 1.913,
+      "step": 114
+    },
+    {
+      "epoch": 19.166666666666668,
+      "grad_norm": 224572.46875,
+      "learning_rate": 0.0009583333333333334,
+      "loss": 0.6642,
+      "step": 115
+    },
+    {
+      "epoch": 19.333333333333332,
+      "grad_norm": 196608.109375,
+      "learning_rate": 0.0009666666666666667,
+      "loss": 0.6549,
+      "step": 116
+    },
+    {
+      "epoch": 19.5,
+      "grad_norm": 117076.5234375,
+      "learning_rate": 0.000975,
+      "loss": 0.6154,
+      "step": 117
+    },
+    {
+      "epoch": 19.666666666666668,
+      "grad_norm": 34586.9453125,
+      "learning_rate": 0.0009833333333333332,
+      "loss": 0.5969,
+      "step": 118
+    },
+    {
+      "epoch": 19.833333333333332,
+      "grad_norm": 125128.671875,
+      "learning_rate": 0.0009916666666666667,
+      "loss": 0.6043,
+      "step": 119
+    },
+    {
+      "epoch": 20.0,
+      "grad_norm": 230583.84375,
+      "learning_rate": 0.001,
+      "loss": 0.639,
+      "step": 120
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.6901270772238515,
+      "eval_best_threshold": 0.285312294960022,
+      "eval_f1": 0.7669688801764274,
+      "eval_loss": 0.6092373132705688,
+      "eval_pr_auc": 0.6895824261804095,
+      "eval_precision": 0.6567352077213596,
+      "eval_recall": 0.9216725559481743,
+      "eval_roc_auc": 0.7143187291179651,
+      "eval_runtime": 3.0536,
+      "eval_samples_per_second": 1005.043,
+      "eval_steps_per_second": 1.965,
+      "step": 120
+    },
+    {
+      "epoch": 20.166666666666668,
+      "grad_norm": 177096.625,
+      "learning_rate": 0.0009999978846027431,
+      "loss": 0.6194,
+      "step": 121
+    },
+    {
+      "epoch": 20.333333333333332,
+      "grad_norm": 22809.671875,
+      "learning_rate": 0.0009999915384288722,
+      "loss": 0.5856,
+      "step": 122
+    },
+    {
+      "epoch": 20.5,
+      "grad_norm": 84596.2734375,
+      "learning_rate": 0.0009999809615320856,
+      "loss": 0.5983,
+      "step": 123
+    },
+    {
+      "epoch": 20.666666666666668,
+      "grad_norm": 132173.640625,
+      "learning_rate": 0.0009999661540018813,
+      "loss": 0.6117,
+      "step": 124
+    },
+    {
+      "epoch": 20.833333333333332,
+      "grad_norm": 133689.75,
+      "learning_rate": 0.0009999471159635538,
+      "loss": 0.6065,
+      "step": 125
+    },
+    {
+      "epoch": 21.0,
+      "grad_norm": 14212.5078125,
+      "learning_rate": 0.0009999238475781956,
+      "loss": 0.5791,
+      "step": 126
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.6699250570218312,
+      "eval_best_threshold": 0.131568044424057,
+      "eval_f1": 0.7661048256753636,
+      "eval_loss": 0.6092211604118347,
+      "eval_pr_auc": 0.671715494700898,
+      "eval_precision": 0.6300797569312572,
+      "eval_recall": 0.9770318021201413,
+      "eval_roc_auc": 0.7061214162798469,
+      "eval_runtime": 2.9753,
+      "eval_samples_per_second": 1031.483,
+      "eval_steps_per_second": 2.017,
+      "step": 126
+    },
+    {
+      "epoch": 21.166666666666668,
+      "grad_norm": 180716.0,
+      "learning_rate": 0.0009998963490426942,
+      "loss": 0.6252,
+      "step": 127
+    },
+    {
+      "epoch": 21.333333333333332,
+      "grad_norm": 180470.5625,
+      "learning_rate": 0.000999864620589731,
+      "loss": 0.6253,
+      "step": 128
+    },
+    {
+      "epoch": 21.5,
+      "grad_norm": 25203.005859375,
+      "learning_rate": 0.0009998286624877785,
+      "loss": 0.583,
+      "step": 129
+    },
+    {
+      "epoch": 21.666666666666668,
+      "grad_norm": 142969.25,
+      "learning_rate": 0.0009997884750411004,
+      "loss": 0.6237,
+      "step": 130
+    },
+    {
+      "epoch": 21.833333333333332,
+      "grad_norm": 161729.53125,
+      "learning_rate": 0.0009997440585897456,
+      "loss": 0.6415,
+      "step": 131
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 105764.96875,
+      "learning_rate": 0.0009996954135095479,
+      "loss": 0.6369,
+      "step": 132
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.7070707070707071,
+      "eval_best_threshold": 0.6199614405632019,
+      "eval_f1": 0.7642276422764228,
+      "eval_loss": 0.6236046552658081,
+      "eval_pr_auc": 0.6916097943168671,
+      "eval_precision": 0.6888888888888889,
+      "eval_recall": 0.8580683156654888,
+      "eval_roc_auc": 0.707266196383268,
+      "eval_runtime": 3.0077,
+      "eval_samples_per_second": 1020.391,
+      "eval_steps_per_second": 1.995,
+      "step": 132
+    },
+    {
+      "epoch": 22.166666666666668,
+      "grad_norm": 35118.234375,
+      "learning_rate": 0.0009996425402121222,
+      "loss": 0.6214,
+      "step": 133
+    },
+    {
+      "epoch": 22.333333333333332,
+      "grad_norm": 71758.453125,
+      "learning_rate": 0.0009995854391448608,
+      "loss": 0.6363,
+      "step": 134
+    },
+    {
+      "epoch": 22.5,
+      "grad_norm": 171227.109375,
+      "learning_rate": 0.000999524110790929,
+      "loss": 0.6473,
+      "step": 135
+    },
+    {
+      "epoch": 22.666666666666668,
+      "grad_norm": 192111.375,
+      "learning_rate": 0.0009994585556692623,
+      "loss": 0.645,
+      "step": 136
+    },
+    {
+      "epoch": 22.833333333333332,
+      "grad_norm": 133878.125,
+      "learning_rate": 0.0009993887743345612,
+      "loss": 0.6269,
+      "step": 137
+    },
+    {
+      "epoch": 23.0,
+      "grad_norm": 106632.1953125,
+      "learning_rate": 0.0009993147673772868,
+      "loss": 0.6224,
+      "step": 138
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.7044639947865754,
+      "eval_best_threshold": 0.5462677478790283,
+      "eval_f1": 0.7639864689045017,
+      "eval_loss": 0.5997253060340881,
+      "eval_pr_auc": 0.6833916293267107,
+      "eval_precision": 0.6843822843822844,
+      "eval_recall": 0.8645465253239105,
+      "eval_roc_auc": 0.709266017685886,
+      "eval_runtime": 3.0311,
+      "eval_samples_per_second": 1012.516,
+      "eval_steps_per_second": 1.98,
+      "step": 138
+    },
+    {
+      "epoch": 23.166666666666668,
+      "grad_norm": 15841.970703125,
+      "learning_rate": 0.0009992365354236556,
+      "loss": 0.6118,
+      "step": 139
+    },
+    {
+      "epoch": 23.333333333333332,
+      "grad_norm": 19277.970703125,
+      "learning_rate": 0.0009991540791356342,
+      "loss": 0.5759,
+      "step": 140
+    },
+    {
+      "epoch": 23.5,
+      "grad_norm": 255906.0,
+      "learning_rate": 0.0009990673992109334,
+      "loss": 0.6571,
+      "step": 141
+    },
+    {
+      "epoch": 23.666666666666668,
+      "grad_norm": 456702.65625,
+      "learning_rate": 0.0009989764963830037,
+      "loss": 0.7798,
+      "step": 142
+    },
+    {
+      "epoch": 23.833333333333332,
+      "grad_norm": 116625.515625,
+      "learning_rate": 0.0009988813714210273,
+      "loss": 0.6315,
+      "step": 143
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 246554.75,
+      "learning_rate": 0.0009987820251299122,
+      "loss": 0.6794,
+      "step": 144
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.6979472140762464,
+      "eval_best_threshold": 0.7202028036117554,
+      "eval_f1": 0.7644218551461245,
+      "eval_loss": 0.7107158899307251,
+      "eval_pr_auc": 0.6772575213237118,
+      "eval_precision": 0.6723290120697363,
+      "eval_recall": 0.8857479387514723,
+      "eval_roc_auc": 0.7083542744327861,
+      "eval_runtime": 3.2437,
+      "eval_samples_per_second": 946.132,
+      "eval_steps_per_second": 1.85,
+      "step": 144
+    },
+    {
+      "epoch": 24.166666666666668,
+      "grad_norm": 318765.8125,
+      "learning_rate": 0.0009986784583502861,
+      "loss": 0.7435,
+      "step": 145
+    },
+    {
+      "epoch": 24.333333333333332,
+      "grad_norm": 248666.8125,
+      "learning_rate": 0.0009985706719584887,
+      "loss": 0.673,
+      "step": 146
+    },
+    {
+      "epoch": 24.5,
+      "grad_norm": 124225.7421875,
+      "learning_rate": 0.000998458666866564,
+      "loss": 0.605,
+      "step": 147
+    },
+    {
+      "epoch": 24.666666666666668,
+      "grad_norm": 93734.8203125,
+      "learning_rate": 0.0009983424440222531,
+      "loss": 0.6139,
+      "step": 148
+    },
+    {
+      "epoch": 24.833333333333332,
+      "grad_norm": 167052.6875,
+      "learning_rate": 0.0009982220044089858,
+      "loss": 0.6204,
+      "step": 149
+    },
+    {
+      "epoch": 25.0,
+      "grad_norm": 135724.3125,
+      "learning_rate": 0.0009980973490458728,
+      "loss": 0.6234,
+      "step": 150
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.7021831215379603,
+      "eval_best_threshold": 0.5155451893806458,
+      "eval_f1": 0.7701207243460765,
+      "eval_loss": 0.6027764678001404,
+      "eval_pr_auc": 0.6704919528444424,
+      "eval_precision": 0.6720807726075505,
+      "eval_recall": 0.9016489988221437,
+      "eval_roc_auc": 0.7086511010937482,
+      "eval_runtime": 3.2106,
+      "eval_samples_per_second": 955.91,
+      "eval_steps_per_second": 1.869,
+      "step": 150
+    },
+    {
+      "epoch": 25.166666666666668,
+      "grad_norm": 20833.037109375,
+      "learning_rate": 0.0009979684789876964,
+      "loss": 0.6018,
+      "step": 151
+    },
+    {
+      "epoch": 25.333333333333332,
+      "grad_norm": 67574.53125,
+      "learning_rate": 0.0009978353953249022,
+      "loss": 0.6104,
+      "step": 152
+    },
+    {
+      "epoch": 25.5,
+      "grad_norm": 76152.453125,
+      "learning_rate": 0.0009976980991835895,
+      "loss": 0.6053,
+      "step": 153
+    },
+    {
+      "epoch": 25.666666666666668,
+      "grad_norm": 51570.14453125,
+      "learning_rate": 0.0009975565917255016,
+      "loss": 0.606,
+      "step": 154
+    },
+    {
+      "epoch": 25.833333333333332,
+      "grad_norm": 66497.03125,
+      "learning_rate": 0.0009974108741480166,
+      "loss": 0.5844,
+      "step": 155
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 30464.404296875,
+      "learning_rate": 0.0009972609476841367,
+      "loss": 0.5894,
+      "step": 156
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.6950146627565983,
+      "eval_best_threshold": 0.5118791460990906,
+      "eval_f1": 0.7697983275946877,
+      "eval_loss": 0.5994439125061035,
+      "eval_pr_auc": 0.6800822574078459,
+      "eval_precision": 0.6608952702702703,
+      "eval_recall": 0.9216725559481743,
+      "eval_roc_auc": 0.7118126701598568,
+      "eval_runtime": 3.261,
+      "eval_samples_per_second": 941.128,
+      "eval_steps_per_second": 1.84,
+      "step": 156
+    },
+    {
+      "epoch": 26.166666666666668,
+      "grad_norm": 126097.4765625,
+      "learning_rate": 0.0009971068136024781,
+      "loss": 0.6058,
+      "step": 157
+    },
+    {
+      "epoch": 26.333333333333332,
+      "grad_norm": 124055.9609375,
+      "learning_rate": 0.00099694847320726,
+      "loss": 0.5917,
+      "step": 158
+    },
+    {
+      "epoch": 26.5,
+      "grad_norm": 77981.765625,
+      "learning_rate": 0.0009967859278382939,
+      "loss": 0.6046,
+      "step": 159
+    },
+    {
+      "epoch": 26.666666666666668,
+      "grad_norm": 100627.078125,
+      "learning_rate": 0.0009966191788709714,
+      "loss": 0.6002,
+      "step": 160
+    },
+    {
+      "epoch": 26.833333333333332,
+      "grad_norm": 156980.953125,
+      "learning_rate": 0.0009964482277162545,
+      "loss": 0.6138,
+      "step": 161
+    },
+    {
+      "epoch": 27.0,
+      "grad_norm": 35392.578125,
+      "learning_rate": 0.0009962730758206612,
+      "loss": 0.5852,
+      "step": 162
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.6972955360052134,
+      "eval_best_threshold": 0.5760672092437744,
+      "eval_f1": 0.769421692727724,
+      "eval_loss": 0.6142106652259827,
+      "eval_pr_auc": 0.6808733029108327,
+      "eval_precision": 0.6649506649506649,
+      "eval_recall": 0.9128386336866903,
+      "eval_roc_auc": 0.7136039395899754,
+      "eval_runtime": 3.2326,
+      "eval_samples_per_second": 949.392,
+      "eval_steps_per_second": 1.856,
+      "step": 162
+    },
+    {
+      "epoch": 27.166666666666668,
+      "grad_norm": 168890.375,
+      "learning_rate": 0.0009960937246662545,
+      "loss": 0.6207,
+      "step": 163
+    },
+    {
+      "epoch": 27.333333333333332,
+      "grad_norm": 213891.765625,
+      "learning_rate": 0.0009959101757706307,
+      "loss": 0.6425,
+      "step": 164
+    },
+    {
+      "epoch": 27.5,
+      "grad_norm": 175997.421875,
+      "learning_rate": 0.0009957224306869053,
+      "loss": 0.6338,
+      "step": 165
+    },
+    {
+      "epoch": 27.666666666666668,
+      "grad_norm": 158954.6875,
+      "learning_rate": 0.0009955304910036994,
+      "loss": 0.6388,
+      "step": 166
+    },
+    {
+      "epoch": 27.833333333333332,
+      "grad_norm": 18399.896484375,
+      "learning_rate": 0.000995334358345128,
+      "loss": 0.5959,
+      "step": 167
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 158990.375,
+      "learning_rate": 0.0009951340343707852,
+      "loss": 0.6161,
+      "step": 168
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.7012056044314109,
+      "eval_best_threshold": 0.2593577802181244,
+      "eval_f1": 0.7741935483870968,
+      "eval_loss": 0.6534614562988281,
+      "eval_pr_auc": 0.6770361224581476,
+      "eval_precision": 0.6652560304697418,
+      "eval_recall": 0.9257950530035336,
+      "eval_roc_auc": 0.7161636936748859,
+      "eval_runtime": 3.2488,
+      "eval_samples_per_second": 944.661,
+      "eval_steps_per_second": 1.847,
+      "step": 168
+    },
+    {
+      "epoch": 28.166666666666668,
+      "grad_norm": 273902.53125,
+      "learning_rate": 0.00099492952077573,
+      "loss": 0.6574,
+      "step": 169
+    },
+    {
+      "epoch": 28.333333333333332,
+      "grad_norm": 311438.375,
+      "learning_rate": 0.0009947208192904722,
+      "loss": 0.6913,
+      "step": 170
+    },
+    {
+      "epoch": 28.5,
+      "grad_norm": 273603.5625,
+      "learning_rate": 0.0009945079316809585,
+      "loss": 0.6621,
+      "step": 171
+    },
+    {
+      "epoch": 28.666666666666668,
+      "grad_norm": 171142.90625,
+      "learning_rate": 0.0009942908597485559,
+      "loss": 0.6109,
+      "step": 172
+    },
+    {
+      "epoch": 28.833333333333332,
+      "grad_norm": 21905.29296875,
+      "learning_rate": 0.000994069605330038,
+      "loss": 0.5994,
+      "step": 173
+    },
+    {
+      "epoch": 29.0,
+      "grad_norm": 28439.1640625,
+      "learning_rate": 0.0009938441702975688,
+      "loss": 0.577,
+      "step": 174
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.6982730531117628,
+      "eval_best_threshold": 0.4422946870326996,
+      "eval_f1": 0.7714708785784797,
+      "eval_loss": 0.5790307521820068,
+      "eval_pr_auc": 0.6856352502299288,
+      "eval_precision": 0.6639762107051826,
+      "eval_recall": 0.9204946996466431,
+      "eval_roc_auc": 0.7194105306023562,
+      "eval_runtime": 3.1704,
+      "eval_samples_per_second": 968.024,
+      "eval_steps_per_second": 1.893,
+      "step": 174
+    },
+    {
+      "epoch": 29.166666666666668,
+      "grad_norm": 64802.87890625,
+      "learning_rate": 0.000993614556558687,
+      "loss": 0.5879,
+      "step": 175
+    },
+    {
+      "epoch": 29.333333333333332,
+      "grad_norm": 96375.390625,
+      "learning_rate": 0.0009933807660562897,
+      "loss": 0.6068,
+      "step": 176
+    },
+    {
+      "epoch": 29.5,
+      "grad_norm": 114612.1171875,
+      "learning_rate": 0.0009931428007686157,
+      "loss": 0.605,
+      "step": 177
+    },
+    {
+      "epoch": 29.666666666666668,
+      "grad_norm": 76551.109375,
+      "learning_rate": 0.0009929006627092298,
+      "loss": 0.6021,
+      "step": 178
+    },
+    {
+      "epoch": 29.833333333333332,
+      "grad_norm": 81290.6875,
+      "learning_rate": 0.0009926543539270048,
+      "loss": 0.5932,
+      "step": 179
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 42667.765625,
+      "learning_rate": 0.000992403876506104,
+      "loss": 0.5767,
+      "step": 180
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.6898012381883349,
+      "eval_best_threshold": 0.3204585909843445,
+      "eval_f1": 0.7729007633587787,
+      "eval_loss": 0.5811671018600464,
+      "eval_pr_auc": 0.6880019352858036,
+      "eval_precision": 0.6495589414595028,
+      "eval_recall": 0.9540636042402827,
+      "eval_roc_auc": 0.7201586111089633,
+      "eval_runtime": 3.1797,
+      "eval_samples_per_second": 965.188,
+      "eval_steps_per_second": 1.887,
+      "step": 180
+    },
+    {
+      "epoch": 30.166666666666668,
+      "grad_norm": 75468.046875,
+      "learning_rate": 0.0009921492325659649,
+      "loss": 0.592,
+      "step": 181
+    },
+    {
+      "epoch": 30.333333333333332,
+      "grad_norm": 130032.859375,
+      "learning_rate": 0.0009918904242612794,
+      "loss": 0.6024,
+      "step": 182
+    },
+    {
+      "epoch": 30.5,
+      "grad_norm": 42846.8671875,
+      "learning_rate": 0.0009916274537819774,
+      "loss": 0.5912,
+      "step": 183
+    },
+    {
+      "epoch": 30.666666666666668,
+      "grad_norm": 27902.64453125,
+      "learning_rate": 0.0009913603233532068,
+      "loss": 0.5837,
+      "step": 184
+    },
+    {
+      "epoch": 30.833333333333332,
+      "grad_norm": 64019.296875,
+      "learning_rate": 0.0009910890352353154,
+      "loss": 0.5779,
+      "step": 185
+    },
+    {
+      "epoch": 31.0,
+      "grad_norm": 30909.447265625,
+      "learning_rate": 0.000990813591723832,
+      "loss": 0.5862,
+      "step": 186
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.6966438579341805,
+      "eval_best_threshold": 0.35055145621299744,
+      "eval_f1": 0.7737545565006075,
+      "eval_loss": 0.5751112699508667,
+      "eval_pr_auc": 0.6866912253137908,
+      "eval_precision": 0.6586677699627638,
+      "eval_recall": 0.9375736160188457,
+      "eval_roc_auc": 0.7222192152951212,
+      "eval_runtime": 3.1318,
+      "eval_samples_per_second": 979.942,
+      "eval_steps_per_second": 1.916,
+      "step": 186
+    },
+    {
+      "epoch": 31.166666666666668,
+      "grad_norm": 64181.921875,
+      "learning_rate": 0.0009905339951494464,
+      "loss": 0.57,
+      "step": 187
+    },
+    {
+      "epoch": 31.333333333333332,
+      "grad_norm": 125966.375,
+      "learning_rate": 0.0009902502478779896,
+      "loss": 0.5932,
+      "step": 188
+    },
+    {
+      "epoch": 31.5,
+      "grad_norm": 17534.990234375,
+      "learning_rate": 0.0009899623523104148,
+      "loss": 0.5786,
+      "step": 189
+    },
+    {
+      "epoch": 31.666666666666668,
+      "grad_norm": 118685.2109375,
+      "learning_rate": 0.000989670310882776,
+      "loss": 0.6049,
+      "step": 190
+    },
+    {
+      "epoch": 31.833333333333332,
+      "grad_norm": 118189.421875,
+      "learning_rate": 0.000989374126066207,
+      "loss": 0.617,
+      "step": 191
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 65363.68359375,
+      "learning_rate": 0.0009890738003669028,
+      "loss": 0.6002,
+      "step": 192
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.7057673509286413,
+      "eval_best_threshold": 0.5547782182693481,
+      "eval_f1": 0.7664856477889838,
+      "eval_loss": 0.5928676724433899,
+      "eval_pr_auc": 0.6915146830214178,
+      "eval_precision": 0.6832641770401107,
+      "eval_recall": 0.872791519434629,
+      "eval_roc_auc": 0.7202294886763422,
+      "eval_runtime": 3.1162,
+      "eval_samples_per_second": 984.839,
+      "eval_steps_per_second": 1.925,
+      "step": 192
+    },
+    {
+      "epoch": 32.166666666666664,
+      "grad_norm": 27484.95703125,
+      "learning_rate": 0.0009887693363260957,
+      "loss": 0.5969,
+      "step": 193
+    },
+    {
+      "epoch": 32.333333333333336,
+      "grad_norm": 122337.7109375,
+      "learning_rate": 0.0009884607365200355,
+      "loss": 0.5997,
+      "step": 194
+    },
+    {
+      "epoch": 32.5,
+      "grad_norm": 214062.09375,
+      "learning_rate": 0.0009881480035599667,
+      "loss": 0.6365,
+      "step": 195
+    },
+    {
+      "epoch": 32.666666666666664,
+      "grad_norm": 273055.5625,
+      "learning_rate": 0.0009878311400921072,
+      "loss": 0.6572,
+      "step": 196
+    },
+    {
+      "epoch": 32.833333333333336,
+      "grad_norm": 230877.578125,
+      "learning_rate": 0.0009875101487976253,
+      "loss": 0.6328,
+      "step": 197
+    },
+    {
+      "epoch": 33.0,
+      "grad_norm": 112179.0390625,
+      "learning_rate": 0.0009871850323926177,
+      "loss": 0.6035,
+      "step": 198
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.7106549364613881,
+      "eval_best_threshold": 0.47324299812316895,
+      "eval_f1": 0.7764350453172205,
+      "eval_loss": 0.5827791094779968,
+      "eval_pr_auc": 0.6930905376712406,
+      "eval_precision": 0.6781002638522428,
+      "eval_recall": 0.9081272084805654,
+      "eval_roc_auc": 0.725550890522939,
+      "eval_runtime": 3.3007,
+      "eval_samples_per_second": 929.815,
+      "eval_steps_per_second": 1.818,
+      "step": 198
+    },
+    {
+      "epoch": 33.166666666666664,
+      "grad_norm": 21957.048828125,
+      "learning_rate": 0.0009868557936280856,
+      "loss": 0.5878,
+      "step": 199
+    },
+    {
+      "epoch": 33.333333333333336,
+      "grad_norm": 58557.421875,
+      "learning_rate": 0.0009865224352899118,
+      "loss": 0.578,
+      "step": 200
+    },
+    {
+      "epoch": 33.5,
+      "grad_norm": 114058.5078125,
+      "learning_rate": 0.0009861849601988384,
+      "loss": 0.5892,
+      "step": 201
+    },
+    {
+      "epoch": 33.666666666666664,
+      "grad_norm": 107046.25,
+      "learning_rate": 0.0009858433712104401,
+      "loss": 0.5904,
+      "step": 202
+    },
+    {
+      "epoch": 33.833333333333336,
+      "grad_norm": 23158.916015625,
+      "learning_rate": 0.0009854976712151031,
+      "loss": 0.5933,
+      "step": 203
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 66935.171875,
+      "learning_rate": 0.0009851478631379982,
+      "loss": 0.5734,
+      "step": 204
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.7103290974258716,
+      "eval_best_threshold": 0.4419545829296112,
+      "eval_f1": 0.7754483455418035,
+      "eval_loss": 0.5724775791168213,
+      "eval_pr_auc": 0.6953394030992055,
+      "eval_precision": 0.6789031402034498,
+      "eval_recall": 0.9040047114252061,
+      "eval_roc_auc": 0.7296596416258367,
+      "eval_runtime": 3.4261,
+      "eval_samples_per_second": 895.774,
+      "eval_steps_per_second": 1.751,
+      "step": 204
+    },
+    {
+      "epoch": 34.166666666666664,
+      "grad_norm": 45089.19921875,
+      "learning_rate": 0.000984793949939058,
+      "loss": 0.584,
+      "step": 205
+    },
+    {
+      "epoch": 34.333333333333336,
+      "grad_norm": 26124.52734375,
+      "learning_rate": 0.0009844359346129503,
+      "loss": 0.5852,
+      "step": 206
+    },
+    {
+      "epoch": 34.5,
+      "grad_norm": 71812.40625,
+      "learning_rate": 0.0009840738201890538,
+      "loss": 0.5733,
+      "step": 207
+    },
+    {
+      "epoch": 34.666666666666664,
+      "grad_norm": 78225.8125,
+      "learning_rate": 0.000983707609731432,
+      "loss": 0.5812,
+      "step": 208
+    },
+    {
+      "epoch": 34.833333333333336,
+      "grad_norm": 52104.34765625,
+      "learning_rate": 0.0009833373063388071,
+      "loss": 0.5843,
+      "step": 209
+    },
+    {
+      "epoch": 35.0,
+      "grad_norm": 59323.8984375,
+      "learning_rate": 0.0009829629131445341,
+      "loss": 0.5729,
+      "step": 210
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.7103290974258716,
+      "eval_best_threshold": 0.5349152684211731,
+      "eval_f1": 0.7753348496335608,
+      "eval_loss": 0.5791059732437134,
+      "eval_pr_auc": 0.697594188551865,
+      "eval_precision": 0.6790615316511731,
+      "eval_recall": 0.9034157832744405,
+      "eval_roc_auc": 0.7347727063804417,
+      "eval_runtime": 3.3061,
+      "eval_samples_per_second": 928.288,
+      "eval_steps_per_second": 1.815,
+      "step": 210
+    },
+    {
+      "epoch": 35.166666666666664,
+      "grad_norm": 68259.4609375,
+      "learning_rate": 0.0009825844333165748,
+      "loss": 0.5711,
+      "step": 211
+    },
+    {
+      "epoch": 35.333333333333336,
+      "grad_norm": 9691.2099609375,
+      "learning_rate": 0.0009822018700574695,
+      "loss": 0.5765,
+      "step": 212
+    },
+    {
+      "epoch": 35.5,
+      "grad_norm": 84118.2265625,
+      "learning_rate": 0.0009818152266043115,
+      "loss": 0.5978,
+      "step": 213
+    },
+    {
+      "epoch": 35.666666666666664,
+      "grad_norm": 39651.7265625,
+      "learning_rate": 0.0009814245062287187,
+      "loss": 0.5697,
+      "step": 214
+    },
+    {
+      "epoch": 35.833333333333336,
+      "grad_norm": 131176.8125,
+      "learning_rate": 0.0009810297122368067,
+      "loss": 0.5914,
+      "step": 215
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 187397.1875,
+      "learning_rate": 0.0009806308479691594,
+      "loss": 0.6202,
+      "step": 216
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.704138155751059,
+      "eval_best_threshold": 0.6704061031341553,
+      "eval_f1": 0.7657378740970072,
+      "eval_loss": 0.612671434879303,
+      "eval_pr_auc": 0.7066488157065188,
+      "eval_precision": 0.6813590449954087,
+      "eval_recall": 0.8739693757361602,
+      "eval_roc_auc": 0.7290279721541368,
+      "eval_runtime": 3.3102,
+      "eval_samples_per_second": 927.121,
+      "eval_steps_per_second": 1.813,
+      "step": 216
+    },
+    {
+      "epoch": 36.166666666666664,
+      "grad_norm": 144801.84375,
+      "learning_rate": 0.0009802279168008028,
+      "loss": 0.6077,
+      "step": 217
+    },
+    {
+      "epoch": 36.333333333333336,
+      "grad_norm": 70672.8828125,
+      "learning_rate": 0.0009798209221411748,
+      "loss": 0.5993,
+      "step": 218
+    },
+    {
+      "epoch": 36.5,
+      "grad_norm": 117014.921875,
+      "learning_rate": 0.0009794098674340967,
+      "loss": 0.5838,
+      "step": 219
+    },
+    {
+      "epoch": 36.666666666666664,
+      "grad_norm": 179098.9375,
+      "learning_rate": 0.0009789947561577445,
+      "loss": 0.5913,
+      "step": 220
+    },
+    {
+      "epoch": 36.833333333333336,
+      "grad_norm": 133181.796875,
+      "learning_rate": 0.0009785755918246195,
+      "loss": 0.5817,
+      "step": 221
+    },
+    {
+      "epoch": 37.0,
+      "grad_norm": 16482.361328125,
+      "learning_rate": 0.0009781523779815178,
+      "loss": 0.5722,
+      "step": 222
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.7044639947865754,
+      "eval_best_threshold": 0.47125041484832764,
+      "eval_f1": 0.7745463584389759,
+      "eval_loss": 0.5698202252388,
+      "eval_pr_auc": 0.7141467386682018,
+      "eval_precision": 0.6701075268817205,
+      "eval_recall": 0.917550058892815,
+      "eval_roc_auc": 0.7431517235276582,
+      "eval_runtime": 3.3671,
+      "eval_samples_per_second": 911.473,
+      "eval_steps_per_second": 1.782,
+      "step": 222
+    },
+    {
+      "epoch": 37.166666666666664,
+      "grad_norm": 28902.521484375,
+      "learning_rate": 0.000977725118209501,
+      "loss": 0.5777,
+      "step": 223
+    },
+    {
+      "epoch": 37.333333333333336,
+      "grad_norm": 24806.671875,
+      "learning_rate": 0.000977293816123866,
+      "loss": 0.5778,
+      "step": 224
+    },
+    {
+      "epoch": 37.5,
+      "grad_norm": 65626.3515625,
+      "learning_rate": 0.0009768584753741135,
+      "loss": 0.5654,
+      "step": 225
+    },
+    {
+      "epoch": 37.666666666666664,
+      "grad_norm": 87890.7265625,
+      "learning_rate": 0.0009764190996439181,
+      "loss": 0.5841,
+      "step": 226
+    },
+    {
+      "epoch": 37.833333333333336,
+      "grad_norm": 78071.140625,
+      "learning_rate": 0.0009759756926510965,
+      "loss": 0.5676,
+      "step": 227
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 43882.6015625,
+      "learning_rate": 0.0009755282581475768,
+      "loss": 0.5653,
+      "step": 228
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.6937113066145324,
+      "eval_best_threshold": 0.318327933549881,
+      "eval_f1": 0.7702834799608993,
+      "eval_loss": 0.579859733581543,
+      "eval_pr_auc": 0.7055837739628394,
+      "eval_precision": 0.658312447786132,
+      "eval_recall": 0.928150765606596,
+      "eval_roc_auc": 0.7329030420651919,
+      "eval_runtime": 3.3237,
+      "eval_samples_per_second": 923.359,
+      "eval_steps_per_second": 1.805,
+      "step": 228
+    },
+    {
+      "epoch": 38.166666666666664,
+      "grad_norm": 60592.203125,
+      "learning_rate": 0.0009750767999193656,
+      "loss": 0.5774,
+      "step": 229
+    },
+    {
+      "epoch": 38.333333333333336,
+      "grad_norm": 68110.4375,
+      "learning_rate": 0.000974621321786517,
+      "loss": 0.5741,
+      "step": 230
+    },
+    {
+      "epoch": 38.5,
+      "grad_norm": 45131.796875,
+      "learning_rate": 0.0009741618276030996,
+      "loss": 0.5557,
+      "step": 231
+    },
+    {
+      "epoch": 38.666666666666664,
+      "grad_norm": 37979.95703125,
+      "learning_rate": 0.0009736983212571646,
+      "loss": 0.5668,
+      "step": 232
+    },
+    {
+      "epoch": 38.833333333333336,
+      "grad_norm": 72676.8046875,
+      "learning_rate": 0.0009732308066707122,
+      "loss": 0.5715,
+      "step": 233
+    },
+    {
+      "epoch": 39.0,
+      "grad_norm": 110384.1328125,
+      "learning_rate": 0.0009727592877996585,
+      "loss": 0.5794,
+      "step": 234
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.7194525904203324,
+      "eval_best_threshold": 0.49165481328964233,
+      "eval_f1": 0.7757228444907528,
+      "eval_loss": 0.5648224949836731,
+      "eval_pr_auc": 0.7175625146397417,
+      "eval_precision": 0.6954694068192433,
+      "eval_recall": 0.8769140164899882,
+      "eval_roc_auc": 0.7463987752356358,
+      "eval_runtime": 3.3315,
+      "eval_samples_per_second": 921.213,
+      "eval_steps_per_second": 1.801,
+      "step": 234
+    },
+    {
+      "epoch": 39.166666666666664,
+      "grad_norm": 33855.453125,
+      "learning_rate": 0.0009722837686338024,
+      "loss": 0.5784,
+      "step": 235
+    },
+    {
+      "epoch": 39.333333333333336,
+      "grad_norm": 82675.9375,
+      "learning_rate": 0.0009718042531967918,
+      "loss": 0.5696,
+      "step": 236
+    },
+    {
+      "epoch": 39.5,
+      "grad_norm": 103150.46875,
+      "learning_rate": 0.0009713207455460893,
+      "loss": 0.5773,
+      "step": 237
+    },
+    {
+      "epoch": 39.666666666666664,
+      "grad_norm": 67929.5625,
+      "learning_rate": 0.0009708332497729377,
+      "loss": 0.5606,
+      "step": 238
+    },
+    {
+      "epoch": 39.833333333333336,
+      "grad_norm": 111831.5390625,
+      "learning_rate": 0.000970341770002326,
+      "loss": 0.5724,
+      "step": 239
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 159352.671875,
+      "learning_rate": 0.0009698463103929542,
+      "loss": 0.5975,
+      "step": 240
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.7080482241772564,
+      "eval_best_threshold": 0.3221179246902466,
+      "eval_f1": 0.7775571002979146,
+      "eval_loss": 0.5620133876800537,
+      "eval_pr_auc": 0.7292111688740706,
+      "eval_precision": 0.6721030042918454,
+      "eval_recall": 0.9222614840989399,
+      "eval_roc_auc": 0.7542481436520763,
+      "eval_runtime": 3.3798,
+      "eval_samples_per_second": 908.043,
+      "eval_steps_per_second": 1.775,
+      "step": 240
+    },
+    {
+      "epoch": 40.166666666666664,
+      "grad_norm": 64492.37890625,
+      "learning_rate": 0.0009693468751371977,
+      "loss": 0.5548,
+      "step": 241
+    },
+    {
+      "epoch": 40.333333333333336,
+      "grad_norm": 53474.79296875,
+      "learning_rate": 0.0009688434684610725,
+      "loss": 0.5691,
+      "step": 242
+    },
+    {
+      "epoch": 40.5,
+      "grad_norm": 55020.44921875,
+      "learning_rate": 0.0009683360946241988,
+      "loss": 0.5637,
+      "step": 243
+    },
+    {
+      "epoch": 40.666666666666664,
+      "grad_norm": 32489.10546875,
+      "learning_rate": 0.0009678247579197657,
+      "loss": 0.5637,
+      "step": 244
+    },
+    {
+      "epoch": 40.833333333333336,
+      "grad_norm": 170300.09375,
+      "learning_rate": 0.0009673094626744943,
+      "loss": 0.5844,
+      "step": 245
+    },
+    {
+      "epoch": 41.0,
+      "grad_norm": 191009.390625,
+      "learning_rate": 0.0009667902132486009,
+      "loss": 0.6063,
+      "step": 246
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.7204301075268817,
+      "eval_best_threshold": 0.370286226272583,
+      "eval_f1": 0.7804503582395087,
+      "eval_loss": 0.5586404204368591,
+      "eval_pr_auc": 0.7353634379200227,
+      "eval_precision": 0.6900452488687783,
+      "eval_recall": 0.8981154299175501,
+      "eval_roc_auc": 0.760805822098165,
+      "eval_runtime": 3.1116,
+      "eval_samples_per_second": 986.318,
+      "eval_steps_per_second": 1.928,
+      "step": 246
+    },
+    {
+      "epoch": 41.166666666666664,
+      "grad_norm": 94776.4140625,
+      "learning_rate": 0.000966267014035761,
+      "loss": 0.5513,
+      "step": 247
+    },
+    {
+      "epoch": 41.333333333333336,
+      "grad_norm": 90176.515625,
+      "learning_rate": 0.0009657398694630713,
+      "loss": 0.5744,
+      "step": 248
+    },
+    {
+      "epoch": 41.5,
+      "grad_norm": 115559.1484375,
+      "learning_rate": 0.0009652087839910124,
+      "loss": 0.5936,
+      "step": 249
+    },
+    {
+      "epoch": 41.666666666666664,
+      "grad_norm": 89386.3125,
+      "learning_rate": 0.0009646737621134112,
+      "loss": 0.5794,
+      "step": 250
+    },
+    {
+      "epoch": 41.833333333333336,
+      "grad_norm": 16934.8046875,
+      "learning_rate": 0.0009641348083574033,
+      "loss": 0.5606,
+      "step": 251
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 160944.21875,
+      "learning_rate": 0.0009635919272833937,
+      "loss": 0.5878,
+      "step": 252
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.6917562724014337,
+      "eval_best_threshold": 0.2776191234588623,
+      "eval_f1": 0.7723772858517806,
+      "eval_loss": 0.5870246291160583,
+      "eval_pr_auc": 0.7297108243263718,
+      "eval_precision": 0.6529698942229455,
+      "eval_recall": 0.9452296819787986,
+      "eval_roc_auc": 0.7465766134956043,
+      "eval_runtime": 3.1963,
+      "eval_samples_per_second": 960.178,
+      "eval_steps_per_second": 1.877,
+      "step": 252
+    },
+    {
+      "epoch": 42.166666666666664,
+      "grad_norm": 165402.4375,
+      "learning_rate": 0.0009630451234850189,
+      "loss": 0.5937,
+      "step": 253
+    },
+    {
+      "epoch": 42.333333333333336,
+      "grad_norm": 66799.2421875,
+      "learning_rate": 0.000962494401589108,
+      "loss": 0.5639,
+      "step": 254
+    },
+    {
+      "epoch": 42.5,
+      "grad_norm": 130646.4921875,
+      "learning_rate": 0.0009619397662556434,
+      "loss": 0.5808,
+      "step": 255
+    },
+    {
+      "epoch": 42.666666666666664,
+      "grad_norm": 183843.8125,
+      "learning_rate": 0.0009613812221777212,
+      "loss": 0.6015,
+      "step": 256
+    },
+    {
+      "epoch": 42.833333333333336,
+      "grad_norm": 191116.703125,
+      "learning_rate": 0.000960818774081512,
+      "loss": 0.6168,
+      "step": 257
+    },
+    {
+      "epoch": 43.0,
+      "grad_norm": 149275.234375,
+      "learning_rate": 0.0009602524267262203,
+      "loss": 0.5984,
+      "step": 258
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.7129358097100033,
+      "eval_best_threshold": 0.5024541616439819,
+      "eval_f1": 0.7796949237309327,
+      "eval_loss": 0.5609505772590637,
+      "eval_pr_auc": 0.7453360679888218,
+      "eval_precision": 0.6775315080399826,
+      "eval_recall": 0.9181389870435807,
+      "eval_roc_auc": 0.7664687249512233,
+      "eval_runtime": 3.0926,
+      "eval_samples_per_second": 992.366,
+      "eval_steps_per_second": 1.94,
+      "step": 258
+    },
+    {
+      "epoch": 43.166666666666664,
+      "grad_norm": 43646.16015625,
+      "learning_rate": 0.0009596821849040446,
+      "loss": 0.5642,
+      "step": 259
+    },
+    {
+      "epoch": 43.333333333333336,
+      "grad_norm": 160944.84375,
+      "learning_rate": 0.000959108053440137,
+      "loss": 0.571,
+      "step": 260
+    },
+    {
+      "epoch": 43.5,
+      "grad_norm": 258851.640625,
+      "learning_rate": 0.000958530037192562,
+      "loss": 0.6145,
+      "step": 261
+    },
+    {
+      "epoch": 43.666666666666664,
+      "grad_norm": 305888.8125,
+      "learning_rate": 0.0009579481410522556,
+      "loss": 0.6399,
+      "step": 262
+    },
+    {
+      "epoch": 43.833333333333336,
+      "grad_norm": 204104.984375,
+      "learning_rate": 0.0009573623699429836,
+      "loss": 0.5947,
+      "step": 263
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 62314.6171875,
+      "learning_rate": 0.0009567727288213005,
+      "loss": 0.5938,
+      "step": 264
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.7197784294558488,
+      "eval_best_threshold": 0.6008195877075195,
+      "eval_f1": 0.7718832891246684,
+      "eval_loss": 0.5699706077575684,
+      "eval_pr_auc": 0.7407351609366284,
+      "eval_precision": 0.7022200772200772,
+      "eval_recall": 0.8568904593639576,
+      "eval_roc_auc": 0.760154607600309,
+      "eval_runtime": 3.2186,
+      "eval_samples_per_second": 953.521,
+      "eval_steps_per_second": 1.864,
+      "step": 264
+    },
+    {
+      "epoch": 44.166666666666664,
+      "grad_norm": 50153.59765625,
+      "learning_rate": 0.0009561792226765072,
+      "loss": 0.5787,
+      "step": 265
+    },
+    {
+      "epoch": 44.333333333333336,
+      "grad_norm": 68213.96875,
+      "learning_rate": 0.0009555818565306084,
+      "loss": 0.5667,
+      "step": 266
+    },
+    {
+      "epoch": 44.5,
+      "grad_norm": 25824.265625,
+      "learning_rate": 0.0009549806354382716,
+      "loss": 0.5499,
+      "step": 267
+    },
+    {
+      "epoch": 44.666666666666664,
+      "grad_norm": 157802.890625,
+      "learning_rate": 0.0009543755644867822,
+      "loss": 0.5828,
+      "step": 268
+    },
+    {
+      "epoch": 44.833333333333336,
+      "grad_norm": 143645.234375,
+      "learning_rate": 0.0009537666487960018,
+      "loss": 0.5936,
+      "step": 269
+    },
+    {
+      "epoch": 45.0,
+      "grad_norm": 43802.171875,
+      "learning_rate": 0.0009531538935183251,
+      "loss": 0.5778,
+      "step": 270
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.7217334636689475,
+      "eval_best_threshold": 0.587894856929779,
+      "eval_f1": 0.7795560144553433,
+      "eval_loss": 0.5726057887077332,
+      "eval_pr_auc": 0.7474146922079,
+      "eval_precision": 0.6939338235294118,
+      "eval_recall": 0.889281507656066,
+      "eval_roc_auc": 0.7676195189088462,
+      "eval_runtime": 3.1802,
+      "eval_samples_per_second": 965.029,
+      "eval_steps_per_second": 1.887,
+      "step": 270
+    },
+    {
+      "epoch": 45.166666666666664,
+      "grad_norm": 97220.7578125,
+      "learning_rate": 0.0009525373038386351,
+      "loss": 0.5589,
+      "step": 271
+    },
+    {
+      "epoch": 45.333333333333336,
+      "grad_norm": 141003.765625,
+      "learning_rate": 0.0009519168849742603,
+      "loss": 0.5977,
+      "step": 272
+    },
+    {
+      "epoch": 45.5,
+      "grad_norm": 139000.796875,
+      "learning_rate": 0.0009512926421749304,
+      "loss": 0.6014,
+      "step": 273
+    },
+    {
+      "epoch": 45.666666666666664,
+      "grad_norm": 118896.6171875,
+      "learning_rate": 0.0009506645807227311,
+      "loss": 0.5868,
+      "step": 274
+    },
+    {
+      "epoch": 45.833333333333336,
+      "grad_norm": 49140.4765625,
+      "learning_rate": 0.0009500327059320606,
+      "loss": 0.5658,
+      "step": 275
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 99207.765625,
+      "learning_rate": 0.0009493970231495835,
+      "loss": 0.5769,
+      "step": 276
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.7064190289996741,
+      "eval_best_threshold": 0.2906385064125061,
+      "eval_f1": 0.7784607819031227,
+      "eval_loss": 0.5690336227416992,
+      "eval_pr_auc": 0.7417411845176535,
+      "eval_precision": 0.6682144364710848,
+      "eval_recall": 0.9322732626619552,
+      "eval_roc_auc": 0.7648114785576027,
+      "eval_runtime": 3.1389,
+      "eval_samples_per_second": 977.733,
+      "eval_steps_per_second": 1.912,
+      "step": 276
+    },
+    {
+      "epoch": 46.166666666666664,
+      "grad_norm": 127385.4296875,
+      "learning_rate": 0.0009487575377541864,
+      "loss": 0.5773,
+      "step": 277
+    },
+    {
+      "epoch": 46.333333333333336,
+      "grad_norm": 54752.7265625,
+      "learning_rate": 0.0009481142551569317,
+      "loss": 0.5695,
+      "step": 278
+    },
+    {
+      "epoch": 46.5,
+      "grad_norm": 55140.91015625,
+      "learning_rate": 0.0009474671808010125,
+      "loss": 0.5552,
+      "step": 279
+    },
+    {
+      "epoch": 46.666666666666664,
+      "grad_norm": 34584.78515625,
+      "learning_rate": 0.0009468163201617061,
+      "loss": 0.5581,
+      "step": 280
+    },
+    {
+      "epoch": 46.833333333333336,
+      "grad_norm": 62221.57421875,
+      "learning_rate": 0.0009461616787463277,
+      "loss": 0.555,
+      "step": 281
+    },
+    {
+      "epoch": 47.0,
+      "grad_norm": 38826.92578125,
+      "learning_rate": 0.0009455032620941839,
+      "loss": 0.5488,
+      "step": 282
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.7073965461062235,
+      "eval_best_threshold": 0.489240437746048,
+      "eval_f1": 0.7765057242409159,
+      "eval_loss": 0.5562921166419983,
+      "eval_pr_auc": 0.758716026312086,
+      "eval_precision": 0.6724137931034483,
+      "eval_recall": 0.9187279151943463,
+      "eval_roc_auc": 0.7723453344089541,
+      "eval_runtime": 3.1347,
+      "eval_samples_per_second": 979.053,
+      "eval_steps_per_second": 1.914,
+      "step": 282
+    },
+    {
+      "epoch": 47.166666666666664,
+      "grad_norm": 73640.1484375,
+      "learning_rate": 0.0009448410757765259,
+      "loss": 0.5463,
+      "step": 283
+    },
+    {
+      "epoch": 47.333333333333336,
+      "grad_norm": 91469.265625,
+      "learning_rate": 0.0009441751253965021,
+      "loss": 0.5645,
+      "step": 284
+    },
+    {
+      "epoch": 47.5,
+      "grad_norm": 42532.26171875,
+      "learning_rate": 0.0009435054165891108,
+      "loss": 0.5533,
+      "step": 285
+    },
+    {
+      "epoch": 47.666666666666664,
+      "grad_norm": 124230.6328125,
+      "learning_rate": 0.0009428319550211531,
+      "loss": 0.5645,
+      "step": 286
+    },
+    {
+      "epoch": 47.833333333333336,
+      "grad_norm": 112197.359375,
+      "learning_rate": 0.0009421547463911835,
+      "loss": 0.5687,
+      "step": 287
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 22061.705078125,
+      "learning_rate": 0.0009414737964294635,
+      "loss": 0.5414,
+      "step": 288
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.7174975562072337,
+      "eval_best_threshold": 0.4483521580696106,
+      "eval_f1": 0.782543265613243,
+      "eval_loss": 0.5482359528541565,
+      "eval_pr_auc": 0.7655134822312841,
+      "eval_precision": 0.6815203145478375,
+      "eval_recall": 0.9187279151943463,
+      "eval_roc_auc": 0.7805329821242479,
+      "eval_runtime": 3.1627,
+      "eval_samples_per_second": 970.359,
+      "eval_steps_per_second": 1.897,
+      "step": 288
+    },
+    {
+      "epoch": 48.166666666666664,
+      "grad_norm": 65311.2890625,
+      "learning_rate": 0.0009407891108979117,
+      "loss": 0.5508,
+      "step": 289
+    },
+    {
+      "epoch": 48.333333333333336,
+      "grad_norm": 15003.859375,
+      "learning_rate": 0.0009401006955900555,
+      "loss": 0.5266,
+      "step": 290
+    },
+    {
+      "epoch": 48.5,
+      "grad_norm": 134944.453125,
+      "learning_rate": 0.0009394085563309827,
+      "loss": 0.5614,
+      "step": 291
+    },
+    {
+      "epoch": 48.666666666666664,
+      "grad_norm": 112159.2421875,
+      "learning_rate": 0.0009387126989772909,
+      "loss": 0.5566,
+      "step": 292
+    },
+    {
+      "epoch": 48.833333333333336,
+      "grad_norm": 104069.140625,
+      "learning_rate": 0.0009380131294170393,
+      "loss": 0.5535,
+      "step": 293
+    },
+    {
+      "epoch": 49.0,
+      "grad_norm": 139390.40625,
+      "learning_rate": 0.0009373098535696979,
+      "loss": 0.5785,
+      "step": 294
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.7165200391006843,
+      "eval_best_threshold": 0.40806540846824646,
+      "eval_f1": 0.7819548872180451,
+      "eval_loss": 0.5490941405296326,
+      "eval_pr_auc": 0.7576025952229788,
+      "eval_precision": 0.680628272251309,
+      "eval_recall": 0.9187279151943463,
+      "eval_roc_auc": 0.7739138764530975,
+      "eval_runtime": 3.1794,
+      "eval_samples_per_second": 965.263,
+      "eval_steps_per_second": 1.887,
+      "step": 294
+    },
+    {
+      "epoch": 49.166666666666664,
+      "grad_norm": 26307.974609375,
+      "learning_rate": 0.0009366028773860979,
+      "loss": 0.5608,
+      "step": 295
+    },
+    {
+      "epoch": 49.333333333333336,
+      "grad_norm": 60580.23828125,
+      "learning_rate": 0.0009358922068483812,
+      "loss": 0.5353,
+      "step": 296
+    },
+    {
+      "epoch": 49.5,
+      "grad_norm": 55138.63671875,
+      "learning_rate": 0.0009351778479699498,
+      "loss": 0.543,
+      "step": 297
+    },
+    {
+      "epoch": 49.666666666666664,
+      "grad_norm": 41980.36328125,
+      "learning_rate": 0.0009344598067954151,
+      "loss": 0.5526,
+      "step": 298
+    },
+    {
+      "epoch": 49.833333333333336,
+      "grad_norm": 28825.138671875,
+      "learning_rate": 0.0009337380894005462,
+      "loss": 0.5639,
+      "step": 299
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 36231.70703125,
+      "learning_rate": 0.0009330127018922195,
+      "loss": 0.5266,
+      "step": 300
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.7347670250896058,
+      "eval_best_threshold": 0.47259435057640076,
+      "eval_f1": 0.7824692677712454,
+      "eval_loss": 0.5478315949440002,
+      "eval_pr_auc": 0.7759952567944263,
+      "eval_precision": 0.7162426614481409,
+      "eval_recall": 0.8621908127208481,
+      "eval_roc_auc": 0.7818392771690897,
+      "eval_runtime": 3.0541,
+      "eval_samples_per_second": 1004.894,
+      "eval_steps_per_second": 1.965,
+      "step": 300
+    },
+    {
+      "epoch": 50.166666666666664,
+      "grad_norm": 90094.9375,
+      "learning_rate": 0.0009322836504083654,
+      "loss": 0.5464,
+      "step": 301
+    },
+    {
+      "epoch": 50.333333333333336,
+      "grad_norm": 46082.94140625,
+      "learning_rate": 0.0009315509411179181,
+      "loss": 0.5416,
+      "step": 302
+    },
+    {
+      "epoch": 50.5,
+      "grad_norm": 34948.2890625,
+      "learning_rate": 0.0009308145802207629,
+      "loss": 0.5505,
+      "step": 303
+    },
+    {
+      "epoch": 50.666666666666664,
+      "grad_norm": 35781.1640625,
+      "learning_rate": 0.0009300745739476828,
+      "loss": 0.5344,
+      "step": 304
+    },
+    {
+      "epoch": 50.833333333333336,
+      "grad_norm": 35878.0703125,
+      "learning_rate": 0.0009293309285603067,
+      "loss": 0.5303,
+      "step": 305
+    },
+    {
+      "epoch": 51.0,
+      "grad_norm": 172840.984375,
+      "learning_rate": 0.0009285836503510562,
+      "loss": 0.5748,
+      "step": 306
+    },
+    {
+      "epoch": 51.0,
+      "eval_accuracy": 0.718475073313783,
+      "eval_best_threshold": 0.49718165397644043,
+      "eval_f1": 0.7808219178082192,
+      "eval_loss": 0.537997841835022,
+      "eval_pr_auc": 0.7953585398041019,
+      "eval_precision": 0.6858288770053476,
+      "eval_recall": 0.9063604240282686,
+      "eval_roc_auc": 0.7944567728455583,
+      "eval_runtime": 3.0929,
+      "eval_samples_per_second": 992.283,
+      "eval_steps_per_second": 1.94,
+      "step": 306
+    },
+    {
+      "epoch": 51.166666666666664,
+      "grad_norm": 57544.5859375,
+      "learning_rate": 0.0009278327456430926,
+      "loss": 0.5362,
+      "step": 307
+    },
+    {
+      "epoch": 51.333333333333336,
+      "grad_norm": 315130.65625,
+      "learning_rate": 0.0009270782207902628,
+      "loss": 0.5983,
+      "step": 308
+    },
+    {
+      "epoch": 51.5,
+      "grad_norm": 210392.46875,
+      "learning_rate": 0.0009263200821770461,
+      "loss": 0.5589,
+      "step": 309
+    },
+    {
+      "epoch": 51.666666666666664,
+      "grad_norm": 60350.46484375,
+      "learning_rate": 0.0009255583362184998,
+      "loss": 0.5427,
+      "step": 310
+    },
+    {
+      "epoch": 51.833333333333336,
+      "grad_norm": 47631.484375,
+      "learning_rate": 0.0009247929893602054,
+      "loss": 0.5454,
+      "step": 311
+    },
+    {
+      "epoch": 52.0,
+      "grad_norm": 87656.984375,
+      "learning_rate": 0.0009240240480782129,
+      "loss": 0.543,
+      "step": 312
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.7344411860540893,
+      "eval_best_threshold": 0.44728729128837585,
+      "eval_f1": 0.7871506920867067,
+      "eval_loss": 0.5385949015617371,
+      "eval_pr_auc": 0.7933372180577872,
+      "eval_precision": 0.7071797278273111,
+      "eval_recall": 0.8875147232037691,
+      "eval_roc_auc": 0.7921294112694474,
+      "eval_runtime": 3.3335,
+      "eval_samples_per_second": 920.652,
+      "eval_steps_per_second": 1.8,
+      "step": 312
+    },
+    {
+      "epoch": 52.166666666666664,
+      "grad_norm": 114437.171875,
+      "learning_rate": 0.0009232515188789881,
+      "loss": 0.5307,
+      "step": 313
+    },
+    {
+      "epoch": 52.333333333333336,
+      "grad_norm": 104620.28125,
+      "learning_rate": 0.0009224754082993551,
+      "loss": 0.5373,
+      "step": 314
+    },
+    {
+      "epoch": 52.5,
+      "grad_norm": 112161.15625,
+      "learning_rate": 0.0009216957229064429,
+      "loss": 0.5456,
+      "step": 315
+    },
+    {
+      "epoch": 52.666666666666664,
+      "grad_norm": 19849.81640625,
+      "learning_rate": 0.0009209124692976287,
+      "loss": 0.5372,
+      "step": 316
+    },
+    {
+      "epoch": 52.833333333333336,
+      "grad_norm": 77253.65625,
+      "learning_rate": 0.0009201256541004829,
+      "loss": 0.533,
+      "step": 317
+    },
+    {
+      "epoch": 53.0,
+      "grad_norm": 85743.8828125,
+      "learning_rate": 0.0009193352839727121,
+      "loss": 0.552,
+      "step": 318
+    },
+    {
+      "epoch": 53.0,
+      "eval_accuracy": 0.7344411860540893,
+      "eval_best_threshold": 0.3785835802555084,
+      "eval_f1": 0.788586251621271,
+      "eval_loss": 0.523644208908081,
+      "eval_pr_auc": 0.8115773421762836,
+      "eval_precision": 0.7046824292999536,
+      "eval_recall": 0.8951707891637221,
+      "eval_roc_auc": 0.8043590133498972,
+      "eval_runtime": 3.1206,
+      "eval_samples_per_second": 983.45,
+      "eval_steps_per_second": 1.923,
+      "step": 318
+    },
+    {
+      "epoch": 53.166666666666664,
+      "grad_norm": 61792.125,
+      "learning_rate": 0.0009185413656021035,
+      "loss": 0.5262,
+      "step": 319
+    },
+    {
+      "epoch": 53.333333333333336,
+      "grad_norm": 78983.8359375,
+      "learning_rate": 0.0009177439057064682,
+      "loss": 0.5294,
+      "step": 320
+    },
+    {
+      "epoch": 53.5,
+      "grad_norm": 161168.796875,
+      "learning_rate": 0.0009169429110335841,
+      "loss": 0.5393,
+      "step": 321
+    },
+    {
+      "epoch": 53.666666666666664,
+      "grad_norm": 30353.265625,
+      "learning_rate": 0.000916138388361139,
+      "loss": 0.5081,
+      "step": 322
+    },
+    {
+      "epoch": 53.833333333333336,
+      "grad_norm": 146950.171875,
+      "learning_rate": 0.0009153303444966727,
+      "loss": 0.5313,
+      "step": 323
+    },
+    {
+      "epoch": 54.0,
+      "grad_norm": 53159.03515625,
+      "learning_rate": 0.0009145187862775209,
+      "loss": 0.5135,
+      "step": 324
+    },
+    {
+      "epoch": 54.0,
+      "eval_accuracy": 0.7416096448354513,
+      "eval_best_threshold": 0.44953474402427673,
+      "eval_f1": 0.7889273356401384,
+      "eval_loss": 0.5188906192779541,
+      "eval_pr_auc": 0.8162425693807065,
+      "eval_precision": 0.7197668771248179,
+      "eval_recall": 0.872791519434629,
+      "eval_roc_auc": 0.8099746644913697,
+      "eval_runtime": 3.3884,
+      "eval_samples_per_second": 905.737,
+      "eval_steps_per_second": 1.771,
+      "step": 324
+    },
+    {
+      "epoch": 54.166666666666664,
+      "grad_norm": 39070.1484375,
+      "learning_rate": 0.0009137037205707553,
+      "loss": 0.5175,
+      "step": 325
+    },
+    {
+      "epoch": 54.333333333333336,
+      "grad_norm": 245568.0,
+      "learning_rate": 0.0009128851542731271,
+      "loss": 0.5452,
+      "step": 326
+    },
+    {
+      "epoch": 54.5,
+      "grad_norm": 52586.328125,
+      "learning_rate": 0.0009120630943110077,
+      "loss": 0.5086,
+      "step": 327
+    },
+    {
+      "epoch": 54.666666666666664,
+      "grad_norm": 335065.6875,
+      "learning_rate": 0.0009112375476403312,
+      "loss": 0.544,
+      "step": 328
+    },
+    {
+      "epoch": 54.833333333333336,
+      "grad_norm": 30743.716796875,
+      "learning_rate": 0.0009104085212465336,
+      "loss": 0.5215,
+      "step": 329
+    },
+    {
+      "epoch": 55.0,
+      "grad_norm": 85467.7734375,
+      "learning_rate": 0.0009095760221444959,
+      "loss": 0.5252,
+      "step": 330
+    },
+    {
+      "epoch": 55.0,
+      "eval_accuracy": 0.7403062886933854,
+      "eval_best_threshold": 0.5788118839263916,
+      "eval_f1": 0.7887622581500132,
+      "eval_loss": 0.5344216823577881,
+      "eval_pr_auc": 0.8178882682425423,
+      "eval_precision": 0.7171084337349397,
+      "eval_recall": 0.8763250883392226,
+      "eval_roc_auc": 0.8111254584489925,
+      "eval_runtime": 3.4301,
+      "eval_samples_per_second": 894.717,
+      "eval_steps_per_second": 1.749,
+      "step": 330
+    },
+    {
+      "epoch": 55.166666666666664,
+      "grad_norm": 142337.96875,
+      "learning_rate": 0.0009087400573784833,
+      "loss": 0.5131,
+      "step": 331
+    },
+    {
+      "epoch": 55.333333333333336,
+      "grad_norm": 36073.97265625,
+      "learning_rate": 0.0009079006340220861,
+      "loss": 0.5315,
+      "step": 332
+    },
+    {
+      "epoch": 55.5,
+      "grad_norm": 161002.625,
+      "learning_rate": 0.0009070577591781596,
+      "loss": 0.5176,
+      "step": 333
+    },
+    {
+      "epoch": 55.666666666666664,
+      "grad_norm": 71573.75,
+      "learning_rate": 0.0009062114399787647,
+      "loss": 0.5202,
+      "step": 334
+    },
+    {
+      "epoch": 55.833333333333336,
+      "grad_norm": 163313.109375,
+      "learning_rate": 0.0009053616835851062,
+      "loss": 0.5409,
+      "step": 335
+    },
+    {
+      "epoch": 56.0,
+      "grad_norm": 71877.1328125,
+      "learning_rate": 0.0009045084971874737,
+      "loss": 0.524,
+      "step": 336
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.7396546106223526,
+      "eval_best_threshold": 0.34169289469718933,
+      "eval_f1": 0.7896814951302974,
+      "eval_loss": 0.5214141607284546,
+      "eval_pr_auc": 0.8224051034771116,
+      "eval_precision": 0.7139457401237506,
+      "eval_recall": 0.8833922261484098,
+      "eval_roc_auc": 0.8150621703656167,
+      "eval_runtime": 3.4488,
+      "eval_samples_per_second": 889.878,
+      "eval_steps_per_second": 1.74,
+      "step": 336
+    },
+    {
+      "epoch": 56.166666666666664,
+      "grad_norm": 132228.5,
+      "learning_rate": 0.00090365188800518,
+      "loss": 0.5285,
+      "step": 337
+    },
+    {
+      "epoch": 56.333333333333336,
+      "grad_norm": 79436.7265625,
+      "learning_rate": 0.0009027918632864998,
+      "loss": 0.5134,
+      "step": 338
+    },
+    {
+      "epoch": 56.5,
+      "grad_norm": 140446.203125,
+      "learning_rate": 0.0009019284303086086,
+      "loss": 0.5139,
+      "step": 339
+    },
+    {
+      "epoch": 56.666666666666664,
+      "grad_norm": 65026.02734375,
+      "learning_rate": 0.0009010615963775219,
+      "loss": 0.5218,
+      "step": 340
+    },
+    {
+      "epoch": 56.833333333333336,
+      "grad_norm": 141111.953125,
+      "learning_rate": 0.0009001913688280319,
+      "loss": 0.5012,
+      "step": 341
+    },
+    {
+      "epoch": 57.0,
+      "grad_norm": 92607.140625,
+      "learning_rate": 0.0008993177550236464,
+      "loss": 0.5093,
+      "step": 342
+    },
+    {
+      "epoch": 57.0,
+      "eval_accuracy": 0.7337895079830564,
+      "eval_best_threshold": 0.38702327013015747,
+      "eval_f1": 0.790566521404768,
+      "eval_loss": 0.5087202787399292,
+      "eval_pr_auc": 0.8277830168523511,
+      "eval_precision": 0.6999546073536087,
+      "eval_recall": 0.9081272084805654,
+      "eval_roc_auc": 0.820387223480836,
+      "eval_runtime": 3.3861,
+      "eval_samples_per_second": 906.364,
+      "eval_steps_per_second": 1.772,
+      "step": 342
+    },
+    {
+      "epoch": 57.166666666666664,
+      "grad_norm": 94696.0234375,
+      "learning_rate": 0.0008984407623565266,
+      "loss": 0.5158,
+      "step": 343
+    },
+    {
+      "epoch": 57.333333333333336,
+      "grad_norm": 248622.8125,
+      "learning_rate": 0.0008975603982474239,
+      "loss": 0.5549,
+      "step": 344
+    },
+    {
+      "epoch": 57.5,
+      "grad_norm": 45057.03515625,
+      "learning_rate": 0.0008966766701456176,
+      "loss": 0.4958,
+      "step": 345
+    },
+    {
+      "epoch": 57.666666666666664,
+      "grad_norm": 71083.1796875,
+      "learning_rate": 0.0008957895855288517,
+      "loss": 0.497,
+      "step": 346
+    },
+    {
+      "epoch": 57.833333333333336,
+      "grad_norm": 191695.15625,
+      "learning_rate": 0.0008948991519032716,
+      "loss": 0.5265,
+      "step": 347
+    },
+    {
+      "epoch": 58.0,
+      "grad_norm": 41673.71484375,
+      "learning_rate": 0.0008940053768033609,
+      "loss": 0.491,
+      "step": 348
+    },
+    {
+      "epoch": 58.0,
+      "eval_accuracy": 0.762137504072988,
+      "eval_best_threshold": 0.42669785022735596,
+      "eval_f1": 0.7985651214128036,
+      "eval_loss": 0.5082907676696777,
+      "eval_pr_auc": 0.8291562365284104,
+      "eval_precision": 0.7512980269989615,
+      "eval_recall": 0.8521790341578327,
+      "eval_roc_auc": 0.8227326266195524,
+      "eval_runtime": 3.5884,
+      "eval_samples_per_second": 855.265,
+      "eval_steps_per_second": 1.672,
+      "step": 348
+    },
+    {
+      "epoch": 58.166666666666664,
+      "grad_norm": 123645.5390625,
+      "learning_rate": 0.0008931082677918772,
+      "loss": 0.5067,
+      "step": 349
+    },
+    {
+      "epoch": 58.333333333333336,
+      "grad_norm": 155054.109375,
+      "learning_rate": 0.0008922078324597878,
+      "loss": 0.5377,
+      "step": 350
+    },
+    {
+      "epoch": 58.5,
+      "grad_norm": 90397.453125,
+      "learning_rate": 0.0008913040784262069,
+      "loss": 0.5253,
+      "step": 351
+    },
+    {
+      "epoch": 58.666666666666664,
+      "grad_norm": 75122.953125,
+      "learning_rate": 0.0008903970133383297,
+      "loss": 0.4941,
+      "step": 352
+    },
+    {
+      "epoch": 58.833333333333336,
+      "grad_norm": 174997.9375,
+      "learning_rate": 0.0008894866448713679,
+      "loss": 0.4985,
+      "step": 353
+    },
+    {
+      "epoch": 59.0,
+      "grad_norm": 73979.5,
+      "learning_rate": 0.0008885729807284854,
+      "loss": 0.5067,
+      "step": 354
+    },
+    {
+      "epoch": 59.0,
+      "eval_accuracy": 0.7442163571195829,
+      "eval_best_threshold": 0.3636360764503479,
+      "eval_f1": 0.7960509223174851,
+      "eval_loss": 0.5114080309867859,
+      "eval_pr_auc": 0.828087073890647,
+      "eval_precision": 0.7122268712226871,
+      "eval_recall": 0.9022379269729093,
+      "eval_roc_auc": 0.8213150752719767,
+      "eval_runtime": 3.5187,
+      "eval_samples_per_second": 872.201,
+      "eval_steps_per_second": 1.705,
+      "step": 354
+    },
+    {
+      "epoch": 59.166666666666664,
+      "grad_norm": 72822.265625,
+      "learning_rate": 0.0008876560286407329,
+      "loss": 0.5072,
+      "step": 355
+    },
+    {
+      "epoch": 59.333333333333336,
+      "grad_norm": 122237.328125,
+      "learning_rate": 0.000886735796366982,
+      "loss": 0.5217,
+      "step": 356
+    },
+    {
+      "epoch": 59.5,
+      "grad_norm": 88669.0234375,
+      "learning_rate": 0.00088581229169386,
+      "loss": 0.5115,
+      "step": 357
+    },
+    {
+      "epoch": 59.666666666666664,
+      "grad_norm": 44189.984375,
+      "learning_rate": 0.0008848855224356839,
+      "loss": 0.4921,
+      "step": 358
+    },
+    {
+      "epoch": 59.833333333333336,
+      "grad_norm": 222048.953125,
+      "learning_rate": 0.0008839554964343943,
+      "loss": 0.4887,
+      "step": 359
+    },
+    {
+      "epoch": 60.0,
+      "grad_norm": 40720.640625,
+      "learning_rate": 0.000883022221559489,
+      "loss": 0.5042,
+      "step": 360
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.7386770935158032,
+      "eval_best_threshold": 0.32580840587615967,
+      "eval_f1": 0.7916883116883117,
+      "eval_loss": 0.516082227230072,
+      "eval_pr_auc": 0.8290447998055172,
+      "eval_precision": 0.70817843866171,
+      "eval_recall": 0.8975265017667845,
+      "eval_roc_auc": 0.8196621244885001,
+      "eval_runtime": 3.559,
+      "eval_samples_per_second": 862.32,
+      "eval_steps_per_second": 1.686,
+      "step": 360
+    },
+    {
+      "epoch": 60.166666666666664,
+      "grad_norm": 236312.984375,
+      "learning_rate": 0.0008820857057079565,
+      "loss": 0.5054,
+      "step": 361
+    },
+    {
+      "epoch": 60.333333333333336,
+      "grad_norm": 268437.3125,
+      "learning_rate": 0.0008811459568042091,
+      "loss": 0.519,
+      "step": 362
+    },
+    {
+      "epoch": 60.5,
+      "grad_norm": 90306.2734375,
+      "learning_rate": 0.0008802029828000156,
+      "loss": 0.5026,
+      "step": 363
+    },
+    {
+      "epoch": 60.666666666666664,
+      "grad_norm": 734178.375,
+      "learning_rate": 0.0008792567916744345,
+      "loss": 0.6645,
+      "step": 364
+    },
+    {
+      "epoch": 60.833333333333336,
+      "grad_norm": 626324.9375,
+      "learning_rate": 0.0008783073914337465,
+      "loss": 0.6744,
+      "step": 365
+    },
+    {
+      "epoch": 61.0,
+      "grad_norm": 299050.0,
+      "learning_rate": 0.000877354790111386,
+      "loss": 0.557,
+      "step": 366
+    },
+    {
+      "epoch": 61.0,
+      "eval_accuracy": 0.7419354838709677,
+      "eval_best_threshold": 0.6245605945587158,
+      "eval_f1": 0.7858301784748513,
+      "eval_loss": 0.542291522026062,
+      "eval_pr_auc": 0.8285012823959621,
+      "eval_precision": 0.7265,
+      "eval_recall": 0.8557126030624264,
+      "eval_roc_auc": 0.815912486393655,
+      "eval_runtime": 3.3978,
+      "eval_samples_per_second": 903.228,
+      "eval_steps_per_second": 1.766,
+      "step": 366
+    },
+    {
+      "epoch": 61.166666666666664,
+      "grad_norm": 184496.296875,
+      "learning_rate": 0.0008763989957678742,
+      "loss": 0.5311,
+      "step": 367
+    },
+    {
+      "epoch": 61.333333333333336,
+      "grad_norm": 247417.09375,
+      "learning_rate": 0.0008754400164907496,
+      "loss": 0.5739,
+      "step": 368
+    },
+    {
+      "epoch": 61.5,
+      "grad_norm": 127733.125,
+      "learning_rate": 0.0008744778603945012,
+      "loss": 0.5525,
+      "step": 369
+    },
+    {
+      "epoch": 61.666666666666664,
+      "grad_norm": 96245.8984375,
+      "learning_rate": 0.000873512535620498,
+      "loss": 0.5395,
+      "step": 370
+    },
+    {
+      "epoch": 61.833333333333336,
+      "grad_norm": 138518.046875,
+      "learning_rate": 0.000872544050336922,
+      "loss": 0.5446,
+      "step": 371
+    },
+    {
+      "epoch": 62.0,
+      "grad_norm": 48077.8515625,
+      "learning_rate": 0.0008715724127386971,
+      "loss": 0.5098,
+      "step": 372
+    },
+    {
+      "epoch": 62.0,
+      "eval_accuracy": 0.7347670250896058,
+      "eval_best_threshold": 0.5379018783569336,
+      "eval_f1": 0.7909604519774012,
+      "eval_loss": 0.5349071025848389,
+      "eval_pr_auc": 0.8189569305770087,
+      "eval_precision": 0.7012750455373407,
+      "eval_recall": 0.9069493521790342,
+      "eval_roc_auc": 0.8158420383872905,
+      "eval_runtime": 3.4752,
+      "eval_samples_per_second": 883.106,
+      "eval_steps_per_second": 1.727,
+      "step": 372
+    },
+    {
+      "epoch": 62.166666666666664,
+      "grad_norm": 115830.6328125,
+      "learning_rate": 0.0008705976310474219,
+      "loss": 0.5403,
+      "step": 373
+    },
+    {
+      "epoch": 62.333333333333336,
+      "grad_norm": 141982.296875,
+      "learning_rate": 0.000869619713511298,
+      "loss": 0.52,
+      "step": 374
+    },
+    {
+      "epoch": 62.5,
+      "grad_norm": 49035.359375,
+      "learning_rate": 0.000868638668405062,
+      "loss": 0.518,
+      "step": 375
+    },
+    {
+      "epoch": 62.666666666666664,
+      "grad_norm": 322461.53125,
+      "learning_rate": 0.0008676545040299144,
+      "loss": 0.5501,
+      "step": 376
+    },
+    {
+      "epoch": 62.833333333333336,
+      "grad_norm": 462798.125,
+      "learning_rate": 0.0008666672287134493,
+      "loss": 0.6246,
+      "step": 377
+    },
+    {
+      "epoch": 63.0,
+      "grad_norm": 299769.09375,
+      "learning_rate": 0.0008656768508095852,
+      "loss": 0.582,
+      "step": 378
+    },
+    {
+      "epoch": 63.0,
+      "eval_accuracy": 0.7240143369175627,
+      "eval_best_threshold": 0.30305054783821106,
+      "eval_f1": 0.7872393870886711,
+      "eval_loss": 0.5427862405776978,
+      "eval_pr_auc": 0.8257574777514395,
+      "eval_precision": 0.6863775733683749,
+      "eval_recall": 0.9228504122497055,
+      "eval_roc_auc": 0.8145997479335967,
+      "eval_runtime": 3.5458,
+      "eval_samples_per_second": 865.527,
+      "eval_steps_per_second": 1.692,
+      "step": 378
+    },
+    {
+      "epoch": 63.166666666666664,
+      "grad_norm": 148813.078125,
+      "learning_rate": 0.0008646833786984927,
+      "loss": 0.5409,
+      "step": 379
+    },
+    {
+      "epoch": 63.333333333333336,
+      "grad_norm": 225467.84375,
+      "learning_rate": 0.0008636868207865244,
+      "loss": 0.5438,
+      "step": 380
+    },
+    {
+      "epoch": 63.5,
+      "grad_norm": 68444.2109375,
+      "learning_rate": 0.0008626871855061438,
+      "loss": 0.5249,
+      "step": 381
+    },
+    {
+      "epoch": 63.666666666666664,
+      "grad_norm": 412893.6875,
+      "learning_rate": 0.000861684481315854,
+      "loss": 0.6003,
+      "step": 382
+    },
+    {
+      "epoch": 63.833333333333336,
+      "grad_norm": 368489.28125,
+      "learning_rate": 0.0008606787167001256,
+      "loss": 0.6114,
+      "step": 383
+    },
+    {
+      "epoch": 64.0,
+      "grad_norm": 166563.171875,
+      "learning_rate": 0.0008596699001693256,
+      "loss": 0.5418,
+      "step": 384
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.7432388400130335,
+      "eval_best_threshold": 0.4921227991580963,
+      "eval_f1": 0.7971163748712667,
+      "eval_loss": 0.5194808840751648,
+      "eval_pr_auc": 0.8246771013143861,
+      "eval_precision": 0.7081427264409881,
+      "eval_recall": 0.911660777385159,
+      "eval_roc_auc": 0.8198530643594086,
+      "eval_runtime": 3.5419,
+      "eval_samples_per_second": 866.482,
+      "eval_steps_per_second": 1.694,
+      "step": 384
+    },
+    {
+      "epoch": 64.16666666666667,
+      "grad_norm": 87660.3984375,
+      "learning_rate": 0.0008586580402596447,
+      "loss": 0.5204,
+      "step": 385
+    },
+    {
+      "epoch": 64.33333333333333,
+      "grad_norm": 101163.578125,
+      "learning_rate": 0.0008576431455330258,
+      "loss": 0.5452,
+      "step": 386
+    },
+    {
+      "epoch": 64.5,
+      "grad_norm": 25221.27734375,
+      "learning_rate": 0.0008566252245770908,
+      "loss": 0.4953,
+      "step": 387
+    },
+    {
+      "epoch": 64.66666666666667,
+      "grad_norm": 135303.609375,
+      "learning_rate": 0.0008556042860050685,
+      "loss": 0.5405,
+      "step": 388
+    },
+    {
+      "epoch": 64.83333333333333,
+      "grad_norm": 210229.96875,
+      "learning_rate": 0.0008545803384557219,
+      "loss": 0.5689,
+      "step": 389
+    },
+    {
+      "epoch": 65.0,
+      "grad_norm": 44948.34375,
+      "learning_rate": 0.0008535533905932737,
+      "loss": 0.5223,
+      "step": 390
+    },
+    {
+      "epoch": 65.0,
+      "eval_accuracy": 0.7409579667644184,
+      "eval_best_threshold": 0.556387722492218,
+      "eval_f1": 0.7936672722553854,
+      "eval_loss": 0.5385985374450684,
+      "eval_pr_auc": 0.822251513124422,
+      "eval_precision": 0.7095127610208817,
+      "eval_recall": 0.9004711425206124,
+      "eval_roc_auc": 0.8151740710098722,
+      "eval_runtime": 3.5537,
+      "eval_samples_per_second": 863.598,
+      "eval_steps_per_second": 1.688,
+      "step": 390
+    },
+    {
+      "epoch": 65.16666666666667,
+      "grad_norm": 142769.75,
+      "learning_rate": 0.0008525234511073352,
+      "loss": 0.5446,
+      "step": 391
+    },
+    {
+      "epoch": 65.33333333333333,
+      "grad_norm": 170959.328125,
+      "learning_rate": 0.000851490528712831,
+      "loss": 0.5357,
+      "step": 392
+    },
+    {
+      "epoch": 65.5,
+      "grad_norm": 125007.1171875,
+      "learning_rate": 0.0008504546321499254,
+      "loss": 0.535,
+      "step": 393
+    },
+    {
+      "epoch": 65.66666666666667,
+      "grad_norm": 106462.4296875,
+      "learning_rate": 0.00084941577018395,
+      "loss": 0.5239,
+      "step": 394
+    },
+    {
+      "epoch": 65.83333333333333,
+      "grad_norm": 168138.53125,
+      "learning_rate": 0.0008483739516053274,
+      "loss": 0.5587,
+      "step": 395
+    },
+    {
+      "epoch": 66.0,
+      "grad_norm": 133317.84375,
+      "learning_rate": 0.0008473291852294987,
+      "loss": 0.5255,
+      "step": 396
+    },
+    {
+      "epoch": 66.0,
+      "eval_accuracy": 0.7386770935158032,
+      "eval_best_threshold": 0.4510115087032318,
+      "eval_f1": 0.7912545549193128,
+      "eval_loss": 0.5226884484291077,
+      "eval_pr_auc": 0.8253034319203902,
+      "eval_precision": 0.7089552238805971,
+      "eval_recall": 0.8951707891637221,
+      "eval_roc_auc": 0.8134717207097379,
+      "eval_runtime": 3.9669,
+      "eval_samples_per_second": 773.652,
+      "eval_steps_per_second": 1.513,
+      "step": 396
+    },
+    {
+      "epoch": 66.16666666666667,
+      "grad_norm": 80206.7265625,
+      "learning_rate": 0.0008462814798968471,
+      "loss": 0.527,
+      "step": 397
+    },
+    {
+      "epoch": 66.33333333333333,
+      "grad_norm": 80147.8046875,
+      "learning_rate": 0.0008452308444726248,
+      "loss": 0.5161,
+      "step": 398
+    },
+    {
+      "epoch": 66.5,
+      "grad_norm": 39128.91796875,
+      "learning_rate": 0.000844177287846877,
+      "loss": 0.5153,
+      "step": 399
+    },
+    {
+      "epoch": 66.66666666666667,
+      "grad_norm": 37543.1328125,
+      "learning_rate": 0.0008431208189343669,
+      "loss": 0.4919,
+      "step": 400
+    },
+    {
+      "epoch": 66.83333333333333,
+      "grad_norm": 75717.34375,
+      "learning_rate": 0.0008420614466745,
+      "loss": 0.5031,
+      "step": 401
+    },
+    {
+      "epoch": 67.0,
+      "grad_norm": 33691.29296875,
+      "learning_rate": 0.0008409991800312493,
+      "loss": 0.4967,
+      "step": 402
+    },
+    {
+      "epoch": 67.0,
+      "eval_accuracy": 0.7461713913326816,
+      "eval_best_threshold": 0.33050093054771423,
+      "eval_f1": 0.7949460384311661,
+      "eval_loss": 0.513331949710846,
+      "eval_pr_auc": 0.8367566848805189,
+      "eval_precision": 0.7187053783912423,
+      "eval_recall": 0.889281507656066,
+      "eval_roc_auc": 0.8243490647168032,
+      "eval_runtime": 3.7958,
+      "eval_samples_per_second": 808.525,
+      "eval_steps_per_second": 1.581,
+      "step": 402
+    },
+    {
+      "epoch": 67.16666666666667,
+      "grad_norm": 148904.53125,
+      "learning_rate": 0.0008399340279930785,
+      "loss": 0.5028,
+      "step": 403
+    },
+    {
+      "epoch": 67.33333333333333,
+      "grad_norm": 65357.32421875,
+      "learning_rate": 0.0008388659995728663,
+      "loss": 0.5002,
+      "step": 404
+    },
+    {
+      "epoch": 67.5,
+      "grad_norm": 106315.421875,
+      "learning_rate": 0.0008377951038078302,
+      "loss": 0.4969,
+      "step": 405
+    },
+    {
+      "epoch": 67.66666666666667,
+      "grad_norm": 45195.2890625,
+      "learning_rate": 0.0008367213497594501,
+      "loss": 0.5188,
+      "step": 406
+    },
+    {
+      "epoch": 67.83333333333333,
+      "grad_norm": 90925.59375,
+      "learning_rate": 0.0008356447465133919,
+      "loss": 0.5067,
+      "step": 407
+    },
+    {
+      "epoch": 68.0,
+      "grad_norm": 217360.078125,
+      "learning_rate": 0.0008345653031794292,
+      "loss": 0.512,
+      "step": 408
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.7435646790485501,
+      "eval_best_threshold": 0.4942082166671753,
+      "eval_f1": 0.7937090432503277,
+      "eval_loss": 0.5158348679542542,
+      "eval_pr_auc": 0.8390055356862849,
+      "eval_precision": 0.7151629664619745,
+      "eval_recall": 0.8916372202591284,
+      "eval_roc_auc": 0.8256519232735298,
+      "eval_runtime": 3.5256,
+      "eval_samples_per_second": 870.495,
+      "eval_steps_per_second": 1.702,
+      "step": 408
+    },
+    {
+      "epoch": 68.16666666666667,
+      "grad_norm": 104175.828125,
+      "learning_rate": 0.000833483028891368,
+      "loss": 0.4955,
+      "step": 409
+    },
+    {
+      "epoch": 68.33333333333333,
+      "grad_norm": 124932.0390625,
+      "learning_rate": 0.0008323979328069688,
+      "loss": 0.5162,
+      "step": 410
+    },
+    {
+      "epoch": 68.5,
+      "grad_norm": 109886.3125,
+      "learning_rate": 0.0008313100241078688,
+      "loss": 0.5087,
+      "step": 411
+    },
+    {
+      "epoch": 68.66666666666667,
+      "grad_norm": 156512.953125,
+      "learning_rate": 0.0008302193119995038,
+      "loss": 0.5231,
+      "step": 412
+    },
+    {
+      "epoch": 68.83333333333333,
+      "grad_norm": 134718.625,
+      "learning_rate": 0.000829125805711032,
+      "loss": 0.5115,
+      "step": 413
+    },
+    {
+      "epoch": 69.0,
+      "grad_norm": 152846.578125,
+      "learning_rate": 0.0008280295144952537,
+      "loss": 0.5167,
+      "step": 414
+    },
+    {
+      "epoch": 69.0,
+      "eval_accuracy": 0.7481264255457803,
+      "eval_best_threshold": 0.3464486598968506,
+      "eval_f1": 0.7973787680209699,
+      "eval_loss": 0.5154203772544861,
+      "eval_pr_auc": 0.8414014550914709,
+      "eval_precision": 0.7184695323571091,
+      "eval_recall": 0.8957597173144877,
+      "eval_roc_auc": 0.8285329889972242,
+      "eval_runtime": 3.5935,
+      "eval_samples_per_second": 854.049,
+      "eval_steps_per_second": 1.67,
+      "step": 414
+    },
+    {
+      "epoch": 69.16666666666667,
+      "grad_norm": 170052.90625,
+      "learning_rate": 0.0008269304476285349,
+      "loss": 0.5021,
+      "step": 415
+    },
+    {
+      "epoch": 69.33333333333333,
+      "grad_norm": 86306.34375,
+      "learning_rate": 0.0008258286144107276,
+      "loss": 0.522,
+      "step": 416
+    },
+    {
+      "epoch": 69.5,
+      "grad_norm": 55669.80859375,
+      "learning_rate": 0.0008247240241650918,
+      "loss": 0.483,
+      "step": 417
+    },
+    {
+      "epoch": 69.66666666666667,
+      "grad_norm": 203016.5625,
+      "learning_rate": 0.0008236166862382162,
+      "loss": 0.5094,
+      "step": 418
+    },
+    {
+      "epoch": 69.83333333333333,
+      "grad_norm": 104809.640625,
+      "learning_rate": 0.0008225066099999392,
+      "loss": 0.5041,
+      "step": 419
+    },
+    {
+      "epoch": 70.0,
+      "grad_norm": 225442.765625,
+      "learning_rate": 0.0008213938048432696,
+      "loss": 0.5173,
+      "step": 420
+    },
+    {
+      "epoch": 70.0,
+      "eval_accuracy": 0.7543173672205931,
+      "eval_best_threshold": 0.5091540217399597,
+      "eval_f1": 0.7990405117270789,
+      "eval_loss": 0.5147111415863037,
+      "eval_pr_auc": 0.845698763041466,
+      "eval_precision": 0.7297955209347614,
+      "eval_recall": 0.8828032979976443,
+      "eval_roc_auc": 0.8304823368806482,
+      "eval_runtime": 3.5604,
+      "eval_samples_per_second": 861.975,
+      "eval_steps_per_second": 1.685,
+      "step": 420
+    },
+    {
+      "epoch": 70.16666666666667,
+      "grad_norm": 152348.796875,
+      "learning_rate": 0.0008202782801843076,
+      "loss": 0.4905,
+      "step": 421
+    },
+    {
+      "epoch": 70.33333333333333,
+      "grad_norm": 145317.3125,
+      "learning_rate": 0.0008191600454621642,
+      "loss": 0.4952,
+      "step": 422
+    },
+    {
+      "epoch": 70.5,
+      "grad_norm": 123521.515625,
+      "learning_rate": 0.000818039110138882,
+      "loss": 0.5022,
+      "step": 423
+    },
+    {
+      "epoch": 70.66666666666667,
+      "grad_norm": 187403.15625,
+      "learning_rate": 0.0008169154836993551,
+      "loss": 0.5047,
+      "step": 424
+    },
+    {
+      "epoch": 70.83333333333333,
+      "grad_norm": 131991.8125,
+      "learning_rate": 0.0008157891756512488,
+      "loss": 0.4971,
+      "step": 425
+    },
+    {
+      "epoch": 71.0,
+      "grad_norm": 182392.90625,
+      "learning_rate": 0.0008146601955249188,
+      "loss": 0.5092,
+      "step": 426
+    },
+    {
+      "epoch": 71.0,
+      "eval_accuracy": 0.7491039426523297,
+      "eval_best_threshold": 0.3482086956501007,
+      "eval_f1": 0.7979002624671916,
+      "eval_loss": 0.5007196068763733,
+      "eval_pr_auc": 0.8512863437213596,
+      "eval_precision": 0.7196969696969697,
+      "eval_recall": 0.8951707891637221,
+      "eval_roc_auc": 0.8366776806110762,
+      "eval_runtime": 3.5517,
+      "eval_samples_per_second": 864.102,
+      "eval_steps_per_second": 1.689,
+      "step": 426
+    },
+    {
+      "epoch": 71.16666666666667,
+      "grad_norm": 201647.609375,
+      "learning_rate": 0.0008135285528733309,
+      "loss": 0.5154,
+      "step": 427
+    },
+    {
+      "epoch": 71.33333333333333,
+      "grad_norm": 121528.09375,
+      "learning_rate": 0.00081239425727198,
+      "loss": 0.5022,
+      "step": 428
+    },
+    {
+      "epoch": 71.5,
+      "grad_norm": 85086.0703125,
+      "learning_rate": 0.0008112573183188099,
+      "loss": 0.4703,
+      "step": 429
+    },
+    {
+      "epoch": 71.66666666666667,
+      "grad_norm": 160130.96875,
+      "learning_rate": 0.00081011774563413,
+      "loss": 0.483,
+      "step": 430
+    },
+    {
+      "epoch": 71.83333333333333,
+      "grad_norm": 174087.8125,
+      "learning_rate": 0.0008089755488605366,
+      "loss": 0.4995,
+      "step": 431
+    },
+    {
+      "epoch": 72.0,
+      "grad_norm": 93436.890625,
+      "learning_rate": 0.0008078307376628291,
+      "loss": 0.4824,
+      "step": 432
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.7647442163571195,
+      "eval_best_threshold": 0.5070761442184448,
+      "eval_f1": 0.8005524861878454,
+      "eval_loss": 0.48814845085144043,
+      "eval_pr_auc": 0.8545800658620809,
+      "eval_precision": 0.7539021852237253,
+      "eval_recall": 0.8533568904593639,
+      "eval_roc_auc": 0.8396495984893199,
+      "eval_runtime": 3.6564,
+      "eval_samples_per_second": 839.359,
+      "eval_steps_per_second": 1.641,
+      "step": 432
+    },
+    {
+      "epoch": 72.16666666666667,
+      "grad_norm": 51381.62890625,
+      "learning_rate": 0.0008066833217279297,
+      "loss": 0.4834,
+      "step": 433
+    },
+    {
+      "epoch": 72.33333333333333,
+      "grad_norm": 209116.90625,
+      "learning_rate": 0.0008055333107648,
+      "loss": 0.5018,
+      "step": 434
+    },
+    {
+      "epoch": 72.5,
+      "grad_norm": 237707.515625,
+      "learning_rate": 0.0008043807145043603,
+      "loss": 0.4991,
+      "step": 435
+    },
+    {
+      "epoch": 72.66666666666667,
+      "grad_norm": 95548.7109375,
+      "learning_rate": 0.0008032255426994069,
+      "loss": 0.4799,
+      "step": 436
+    },
+    {
+      "epoch": 72.83333333333333,
+      "grad_norm": 117247.875,
+      "learning_rate": 0.0008020678051245282,
+      "loss": 0.4845,
+      "step": 437
+    },
+    {
+      "epoch": 73.0,
+      "grad_norm": 128688.5703125,
+      "learning_rate": 0.0008009075115760243,
+      "loss": 0.4959,
+      "step": 438
+    },
+    {
+      "epoch": 73.0,
+      "eval_accuracy": 0.761485826001955,
+      "eval_best_threshold": 0.37738415598869324,
+      "eval_f1": 0.8025889967637541,
+      "eval_loss": 0.4879034757614136,
+      "eval_pr_auc": 0.8562801951827024,
+      "eval_precision": 0.7402985074626866,
+      "eval_recall": 0.8763250883392226,
+      "eval_roc_auc": 0.8414073621603138,
+      "eval_runtime": 3.7092,
+      "eval_samples_per_second": 827.405,
+      "eval_steps_per_second": 1.618,
+      "step": 438
+    },
+    {
+      "epoch": 73.16666666666667,
+      "grad_norm": 154160.890625,
+      "learning_rate": 0.000799744671871822,
+      "loss": 0.5086,
+      "step": 439
+    },
+    {
+      "epoch": 73.33333333333333,
+      "grad_norm": 119829.8125,
+      "learning_rate": 0.0007985792958513931,
+      "loss": 0.4827,
+      "step": 440
+    },
+    {
+      "epoch": 73.5,
+      "grad_norm": 152021.484375,
+      "learning_rate": 0.0007974113933756707,
+      "loss": 0.4863,
+      "step": 441
+    },
+    {
+      "epoch": 73.66666666666667,
+      "grad_norm": 209087.796875,
+      "learning_rate": 0.0007962409743269654,
+      "loss": 0.4906,
+      "step": 442
+    },
+    {
+      "epoch": 73.83333333333333,
+      "grad_norm": 216866.65625,
+      "learning_rate": 0.0007950680486088822,
+      "loss": 0.51,
+      "step": 443
+    },
+    {
+      "epoch": 74.0,
+      "grad_norm": 57889.4296875,
+      "learning_rate": 0.0007938926261462366,
+      "loss": 0.4732,
+      "step": 444
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy": 0.7552948843271424,
+      "eval_best_threshold": 0.4220488965511322,
+      "eval_f1": 0.8007429026266915,
+      "eval_loss": 0.4835204780101776,
+      "eval_pr_auc": 0.8598558329755022,
+      "eval_precision": 0.7286335103814582,
+      "eval_recall": 0.8886925795053003,
+      "eval_roc_auc": 0.8418996390828357,
+      "eval_runtime": 3.7001,
+      "eval_samples_per_second": 829.428,
+      "eval_steps_per_second": 1.622,
+      "step": 444
+    },
+    {
+      "epoch": 74.16666666666667,
+      "grad_norm": 53130.546875,
+      "learning_rate": 0.0007927147168849704,
+      "loss": 0.4673,
+      "step": 445
+    },
+    {
+      "epoch": 74.33333333333333,
+      "grad_norm": 162182.0,
+      "learning_rate": 0.0007915343307920673,
+      "loss": 0.497,
+      "step": 446
+    },
+    {
+      "epoch": 74.5,
+      "grad_norm": 89758.8203125,
+      "learning_rate": 0.0007903514778554699,
+      "loss": 0.4919,
+      "step": 447
+    },
+    {
+      "epoch": 74.66666666666667,
+      "grad_norm": 248663.484375,
+      "learning_rate": 0.0007891661680839932,
+      "loss": 0.5119,
+      "step": 448
+    },
+    {
+      "epoch": 74.83333333333333,
+      "grad_norm": 246315.734375,
+      "learning_rate": 0.0007879784115072417,
+      "loss": 0.5121,
+      "step": 449
+    },
+    {
+      "epoch": 75.0,
+      "grad_norm": 38542.20703125,
+      "learning_rate": 0.0007867882181755231,
+      "loss": 0.4878,
+      "step": 450
+    },
+    {
+      "epoch": 75.0,
+      "eval_accuracy": 0.7673509286412512,
+      "eval_best_threshold": 0.31659069657325745,
+      "eval_f1": 0.8055555555555556,
+      "eval_loss": 0.529030442237854,
+      "eval_pr_auc": 0.856213564735403,
+      "eval_precision": 0.7492401215805471,
+      "eval_recall": 0.8710247349823321,
+      "eval_roc_auc": 0.8407879351775246,
+      "eval_runtime": 4.2878,
+      "eval_samples_per_second": 715.756,
+      "eval_steps_per_second": 1.399,
+      "step": 450
+    },
+    {
+      "epoch": 75.16666666666667,
+      "grad_norm": 333584.46875,
+      "learning_rate": 0.0007855955981597644,
+      "loss": 0.5461,
+      "step": 451
+    },
+    {
+      "epoch": 75.33333333333333,
+      "grad_norm": 322359.9375,
+      "learning_rate": 0.0007844005615514259,
+      "loss": 0.5358,
+      "step": 452
+    },
+    {
+      "epoch": 75.5,
+      "grad_norm": 84971.671875,
+      "learning_rate": 0.0007832031184624164,
+      "loss": 0.4801,
+      "step": 453
+    },
+    {
+      "epoch": 75.66666666666667,
+      "grad_norm": 296875.28125,
+      "learning_rate": 0.0007820032790250074,
+      "loss": 0.5571,
+      "step": 454
+    },
+    {
+      "epoch": 75.83333333333333,
+      "grad_norm": 333367.90625,
+      "learning_rate": 0.0007808010533917464,
+      "loss": 0.5737,
+      "step": 455
+    },
+    {
+      "epoch": 76.0,
+      "grad_norm": 226147.171875,
+      "learning_rate": 0.0007795964517353734,
+      "loss": 0.5162,
+      "step": 456
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.7422613229064842,
+      "eval_best_threshold": 0.37861621379852295,
+      "eval_f1": 0.792769190463715,
+      "eval_loss": 0.509261965751648,
+      "eval_pr_auc": 0.8386148647298124,
+      "eval_precision": 0.7140160453043889,
+      "eval_recall": 0.8910482921083628,
+      "eval_roc_auc": 0.8225797028984201,
+      "eval_runtime": 3.7448,
+      "eval_samples_per_second": 819.535,
+      "eval_steps_per_second": 1.602,
+      "step": 456
+    },
+    {
+      "epoch": 76.16666666666667,
+      "grad_norm": 73060.7265625,
+      "learning_rate": 0.0007783894842487322,
+      "loss": 0.5189,
+      "step": 457
+    },
+    {
+      "epoch": 76.33333333333333,
+      "grad_norm": 191211.65625,
+      "learning_rate": 0.0007771801611446858,
+      "loss": 0.527,
+      "step": 458
+    },
+    {
+      "epoch": 76.5,
+      "grad_norm": 145793.390625,
+      "learning_rate": 0.0007759684926560291,
+      "loss": 0.5262,
+      "step": 459
+    },
+    {
+      "epoch": 76.66666666666667,
+      "grad_norm": 39650.92578125,
+      "learning_rate": 0.000774754489035403,
+      "loss": 0.4805,
+      "step": 460
+    },
+    {
+      "epoch": 76.83333333333333,
+      "grad_norm": 93844.328125,
+      "learning_rate": 0.0007735381605552072,
+      "loss": 0.4957,
+      "step": 461
+    },
+    {
+      "epoch": 77.0,
+      "grad_norm": 129532.5234375,
+      "learning_rate": 0.0007723195175075137,
+      "loss": 0.5023,
+      "step": 462
+    },
+    {
+      "epoch": 77.0,
+      "eval_accuracy": 0.747148908439231,
+      "eval_best_threshold": 0.3997952342033386,
+      "eval_f1": 0.7988595127008813,
+      "eval_loss": 0.49405384063720703,
+      "eval_pr_auc": 0.8504424681690863,
+      "eval_precision": 0.7134259259259259,
+      "eval_recall": 0.9075382803297998,
+      "eval_roc_auc": 0.8348483950311818,
+      "eval_runtime": 3.7383,
+      "eval_samples_per_second": 820.962,
+      "eval_steps_per_second": 1.605,
+      "step": 462
+    },
+    {
+      "epoch": 77.16666666666667,
+      "grad_norm": 30520.681640625,
+      "learning_rate": 0.0007710985702039786,
+      "loss": 0.5047,
+      "step": 463
+    },
+    {
+      "epoch": 77.33333333333333,
+      "grad_norm": 122456.078125,
+      "learning_rate": 0.0007698753289757565,
+      "loss": 0.4829,
+      "step": 464
+    },
+    {
+      "epoch": 77.5,
+      "grad_norm": 161532.8125,
+      "learning_rate": 0.000768649804173412,
+      "loss": 0.4899,
+      "step": 465
+    },
+    {
+      "epoch": 77.66666666666667,
+      "grad_norm": 46629.90234375,
+      "learning_rate": 0.0007674220061668323,
+      "loss": 0.4808,
+      "step": 466
+    },
+    {
+      "epoch": 77.83333333333333,
+      "grad_norm": 169030.65625,
+      "learning_rate": 0.0007661919453451394,
+      "loss": 0.491,
+      "step": 467
+    },
+    {
+      "epoch": 78.0,
+      "grad_norm": 80774.3828125,
+      "learning_rate": 0.0007649596321166025,
+      "loss": 0.5117,
+      "step": 468
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy": 0.7432388400130335,
+      "eval_best_threshold": 0.40479913353919983,
+      "eval_f1": 0.7895299145299145,
+      "eval_loss": 0.512270450592041,
+      "eval_pr_auc": 0.8395600491799751,
+      "eval_precision": 0.7223851417399805,
+      "eval_recall": 0.8704358068315665,
+      "eval_roc_auc": 0.8215491860248338,
+      "eval_runtime": 3.7435,
+      "eval_samples_per_second": 819.822,
+      "eval_steps_per_second": 1.603,
+      "step": 468
+    },
+    {
+      "epoch": 78.16666666666667,
+      "grad_norm": 150559.34375,
+      "learning_rate": 0.00076372507690855,
+      "loss": 0.516,
+      "step": 469
+    },
+    {
+      "epoch": 78.33333333333333,
+      "grad_norm": 95044.8203125,
+      "learning_rate": 0.00076248829016728,
+      "loss": 0.4937,
+      "step": 470
+    },
+    {
+      "epoch": 78.5,
+      "grad_norm": 100775.890625,
+      "learning_rate": 0.0007612492823579744,
+      "loss": 0.4884,
+      "step": 471
+    },
+    {
+      "epoch": 78.66666666666667,
+      "grad_norm": 106388.2109375,
+      "learning_rate": 0.0007600080639646077,
+      "loss": 0.4644,
+      "step": 472
+    },
+    {
+      "epoch": 78.83333333333333,
+      "grad_norm": 151662.96875,
+      "learning_rate": 0.00075876464548986,
+      "loss": 0.4927,
+      "step": 473
+    },
+    {
+      "epoch": 79.0,
+      "grad_norm": 163484.265625,
+      "learning_rate": 0.0007575190374550272,
+      "loss": 0.5001,
+      "step": 474
+    },
+    {
+      "epoch": 79.0,
+      "eval_accuracy": 0.7536656891495601,
+      "eval_best_threshold": 0.38404473662376404,
+      "eval_f1": 0.8009478672985783,
+      "eval_loss": 0.4811509847640991,
+      "eval_pr_auc": 0.861499823224796,
+      "eval_precision": 0.7242857142857143,
+      "eval_recall": 0.8957597173144877,
+      "eval_roc_auc": 0.8440641970344825,
+      "eval_runtime": 3.7485,
+      "eval_samples_per_second": 818.72,
+      "eval_steps_per_second": 1.601,
+      "step": 474
+    },
+    {
+      "epoch": 79.16666666666667,
+      "grad_norm": 38699.5625,
+      "learning_rate": 0.0007562712503999327,
+      "loss": 0.4759,
+      "step": 475
+    },
+    {
+      "epoch": 79.33333333333333,
+      "grad_norm": 138076.015625,
+      "learning_rate": 0.0007550212948828377,
+      "loss": 0.4878,
+      "step": 476
+    },
+    {
+      "epoch": 79.5,
+      "grad_norm": 175757.359375,
+      "learning_rate": 0.0007537691814803521,
+      "loss": 0.5021,
+      "step": 477
+    },
+    {
+      "epoch": 79.66666666666667,
+      "grad_norm": 64115.30859375,
+      "learning_rate": 0.000752514920787345,
+      "loss": 0.4995,
+      "step": 478
+    },
+    {
+      "epoch": 79.83333333333333,
+      "grad_norm": 122277.4765625,
+      "learning_rate": 0.000751258523416855,
+      "loss": 0.4964,
+      "step": 479
+    },
+    {
+      "epoch": 80.0,
+      "grad_norm": 67262.625,
+      "learning_rate": 0.00075,
+      "loss": 0.4762,
+      "step": 480
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.7530140110785272,
+      "eval_best_threshold": 0.48448118567466736,
+      "eval_f1": 0.7952458130740141,
+      "eval_loss": 0.5035498738288879,
+      "eval_pr_auc": 0.8543361186664507,
+      "eval_precision": 0.7345309381237525,
+      "eval_recall": 0.866902237926973,
+      "eval_roc_auc": 0.8367859299867093,
+      "eval_runtime": 3.7951,
+      "eval_samples_per_second": 808.683,
+      "eval_steps_per_second": 1.581,
+      "step": 480
+    },
+    {
+      "epoch": 80.16666666666667,
+      "grad_norm": 128038.4921875,
+      "learning_rate": 0.0007487393611858883,
+      "loss": 0.5011,
+      "step": 481
+    },
+    {
+      "epoch": 80.33333333333333,
+      "grad_norm": 115982.5390625,
+      "learning_rate": 0.0007474766176415271,
+      "loss": 0.5074,
+      "step": 482
+    },
+    {
+      "epoch": 80.5,
+      "grad_norm": 346951.09375,
+      "learning_rate": 0.0007462117800517336,
+      "loss": 0.5196,
+      "step": 483
+    },
+    {
+      "epoch": 80.66666666666667,
+      "grad_norm": 445729.40625,
+      "learning_rate": 0.0007449448591190436,
+      "loss": 0.5751,
+      "step": 484
+    },
+    {
+      "epoch": 80.83333333333333,
+      "grad_norm": 270755.5,
+      "learning_rate": 0.0007436758655636212,
+      "loss": 0.5385,
+      "step": 485
+    },
+    {
+      "epoch": 81.0,
+      "grad_norm": 146638.015625,
+      "learning_rate": 0.0007424048101231686,
+      "loss": 0.4997,
+      "step": 486
+    },
+    {
+      "epoch": 81.0,
+      "eval_accuracy": 0.757901596611274,
+      "eval_best_threshold": 0.5885984897613525,
+      "eval_f1": 0.8016021361815754,
+      "eval_loss": 0.5125724673271179,
+      "eval_pr_auc": 0.8595454646304805,
+      "eval_precision": 0.7332681973619931,
+      "eval_recall": 0.8839811542991755,
+      "eval_roc_auc": 0.841192152092091,
+      "eval_runtime": 3.8042,
+      "eval_samples_per_second": 806.738,
+      "eval_steps_per_second": 1.577,
+      "step": 486
+    },
+    {
+      "epoch": 81.16666666666667,
+      "grad_norm": 233428.828125,
+      "learning_rate": 0.0007411317035528344,
+      "loss": 0.5256,
+      "step": 487
+    },
+    {
+      "epoch": 81.33333333333333,
+      "grad_norm": 86020.21875,
+      "learning_rate": 0.0007398565566251232,
+      "loss": 0.4879,
+      "step": 488
+    },
+    {
+      "epoch": 81.5,
+      "grad_norm": 195027.78125,
+      "learning_rate": 0.0007385793801298042,
+      "loss": 0.5046,
+      "step": 489
+    },
+    {
+      "epoch": 81.66666666666667,
+      "grad_norm": 217321.640625,
+      "learning_rate": 0.0007373001848738202,
+      "loss": 0.5145,
+      "step": 490
+    },
+    {
+      "epoch": 81.83333333333333,
+      "grad_norm": 123365.1796875,
+      "learning_rate": 0.0007360189816811956,
+      "loss": 0.4879,
+      "step": 491
+    },
+    {
+      "epoch": 82.0,
+      "grad_norm": 115342.3515625,
+      "learning_rate": 0.0007347357813929454,
+      "loss": 0.4934,
+      "step": 492
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy": 0.7468230694037146,
+      "eval_best_threshold": 0.4948383569717407,
+      "eval_f1": 0.7964369924024103,
+      "eval_loss": 0.5086027979850769,
+      "eval_pr_auc": 0.8558498971534145,
+      "eval_precision": 0.7173194903256253,
+      "eval_recall": 0.8951707891637221,
+      "eval_roc_auc": 0.8383057168557166,
+      "eval_runtime": 3.7413,
+      "eval_samples_per_second": 820.293,
+      "eval_steps_per_second": 1.604,
+      "step": 492
+    },
+    {
+      "epoch": 82.16666666666667,
+      "grad_norm": 158514.203125,
+      "learning_rate": 0.000733450594866983,
+      "loss": 0.4833,
+      "step": 493
+    },
+    {
+      "epoch": 82.33333333333333,
+      "grad_norm": 76403.3828125,
+      "learning_rate": 0.0007321634329780285,
+      "loss": 0.4802,
+      "step": 494
+    },
+    {
+      "epoch": 82.5,
+      "grad_norm": 213661.546875,
+      "learning_rate": 0.000730874306617517,
+      "loss": 0.5218,
+      "step": 495
+    },
+    {
+      "epoch": 82.66666666666667,
+      "grad_norm": 266354.0625,
+      "learning_rate": 0.0007295832266935059,
+      "loss": 0.5181,
+      "step": 496
+    },
+    {
+      "epoch": 82.83333333333333,
+      "grad_norm": 134915.609375,
+      "learning_rate": 0.0007282902041305828,
+      "loss": 0.4869,
+      "step": 497
+    },
+    {
+      "epoch": 83.0,
+      "grad_norm": 109487.90625,
+      "learning_rate": 0.0007269952498697733,
+      "loss": 0.4924,
+      "step": 498
+    },
+    {
+      "epoch": 83.0,
+      "eval_accuracy": 0.7556207233626588,
+      "eval_best_threshold": 0.4894412159919739,
+      "eval_f1": 0.8012718600953895,
+      "eval_loss": 0.49321600794792175,
+      "eval_pr_auc": 0.8601618845148933,
+      "eval_precision": 0.7283236994219653,
+      "eval_recall": 0.8904593639575972,
+      "eval_roc_auc": 0.8421629599846733,
+      "eval_runtime": 3.7702,
+      "eval_samples_per_second": 814.019,
+      "eval_steps_per_second": 1.591,
+      "step": 498
+    },
+    {
+      "epoch": 83.16666666666667,
+      "grad_norm": 106140.2109375,
+      "learning_rate": 0.0007256983748684484,
+      "loss": 0.4818,
+      "step": 499
+    },
+    {
+      "epoch": 83.33333333333333,
+      "grad_norm": 32685.55078125,
+      "learning_rate": 0.0007243995901002312,
+      "loss": 0.4807,
+      "step": 500
+    },
+    {
+      "epoch": 83.5,
+      "grad_norm": 118984.3671875,
+      "learning_rate": 0.0007230989065549044,
+      "loss": 0.4849,
+      "step": 501
+    },
+    {
+      "epoch": 83.66666666666667,
+      "grad_norm": 121551.421875,
+      "learning_rate": 0.0007217963352383181,
+      "loss": 0.494,
+      "step": 502
+    },
+    {
+      "epoch": 83.83333333333333,
+      "grad_norm": 90743.734375,
+      "learning_rate": 0.000720491887172295,
+      "loss": 0.4627,
+      "step": 503
+    },
+    {
+      "epoch": 84.0,
+      "grad_norm": 113933.1171875,
+      "learning_rate": 0.0007191855733945387,
+      "loss": 0.4873,
+      "step": 504
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.7706093189964157,
+      "eval_best_threshold": 0.4577232599258423,
+      "eval_f1": 0.8032420346562326,
+      "eval_loss": 0.4764363467693329,
+      "eval_pr_auc": 0.8592588506486144,
+      "eval_precision": 0.7643617021276595,
+      "eval_recall": 0.8462897526501767,
+      "eval_roc_auc": 0.8475118537361928,
+      "eval_runtime": 3.5755,
+      "eval_samples_per_second": 858.343,
+      "eval_steps_per_second": 1.678,
+      "step": 504
+    },
+    {
+      "epoch": 84.16666666666667,
+      "grad_norm": 47223.42578125,
+      "learning_rate": 0.0007178774049585397,
+      "loss": 0.4676,
+      "step": 505
+    },
+    {
+      "epoch": 84.33333333333333,
+      "grad_norm": 329249.5,
+      "learning_rate": 0.0007165673929334815,
+      "loss": 0.5294,
+      "step": 506
+    },
+    {
+      "epoch": 84.5,
+      "grad_norm": 365006.0625,
+      "learning_rate": 0.0007152555484041476,
+      "loss": 0.5173,
+      "step": 507
+    },
+    {
+      "epoch": 84.66666666666667,
+      "grad_norm": 209159.03125,
+      "learning_rate": 0.0007139418824708272,
+      "loss": 0.4985,
+      "step": 508
+    },
+    {
+      "epoch": 84.83333333333333,
+      "grad_norm": 80609.34375,
+      "learning_rate": 0.0007126264062492216,
+      "loss": 0.47,
+      "step": 509
+    },
+    {
+      "epoch": 85.0,
+      "grad_norm": 154371.296875,
+      "learning_rate": 0.0007113091308703497,
+      "loss": 0.4881,
+      "step": 510
+    },
+    {
+      "epoch": 85.0,
+      "eval_accuracy": 0.7735418703160639,
+      "eval_best_threshold": 0.5390059351921082,
+      "eval_f1": 0.8057031031590719,
+      "eval_loss": 0.4798082411289215,
+      "eval_pr_auc": 0.8604935935331044,
+      "eval_precision": 0.766897285790314,
+      "eval_recall": 0.8486454652532391,
+      "eval_roc_auc": 0.847249391956384,
+      "eval_runtime": 3.4838,
+      "eval_samples_per_second": 880.946,
+      "eval_steps_per_second": 1.722,
+      "step": 510
+    },
+    {
+      "epoch": 85.16666666666667,
+      "grad_norm": 72014.5390625,
+      "learning_rate": 0.0007099900674804548,
+      "loss": 0.4731,
+      "step": 511
+    },
+    {
+      "epoch": 85.33333333333333,
+      "grad_norm": 187983.53125,
+      "learning_rate": 0.000708669227240909,
+      "loss": 0.4994,
+      "step": 512
+    },
+    {
+      "epoch": 85.5,
+      "grad_norm": 226755.40625,
+      "learning_rate": 0.0007073466213281196,
+      "loss": 0.4852,
+      "step": 513
+    },
+    {
+      "epoch": 85.66666666666667,
+      "grad_norm": 62913.12890625,
+      "learning_rate": 0.0007060222609334342,
+      "loss": 0.4722,
+      "step": 514
+    },
+    {
+      "epoch": 85.83333333333333,
+      "grad_norm": 155044.375,
+      "learning_rate": 0.0007046961572630462,
+      "loss": 0.4641,
+      "step": 515
+    },
+    {
+      "epoch": 86.0,
+      "grad_norm": 197888.65625,
+      "learning_rate": 0.0007033683215379002,
+      "loss": 0.5081,
+      "step": 516
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy": 0.7725643532095146,
+      "eval_best_threshold": 0.45184916257858276,
+      "eval_f1": 0.8087671232876712,
+      "eval_loss": 0.46945297718048096,
+      "eval_pr_auc": 0.8683643115135102,
+      "eval_precision": 0.7561475409836066,
+      "eval_recall": 0.8692579505300353,
+      "eval_roc_auc": 0.853145761220778,
+      "eval_runtime": 3.3665,
+      "eval_samples_per_second": 911.627,
+      "eval_steps_per_second": 1.782,
+      "step": 516
+    },
+    {
+      "epoch": 86.16666666666667,
+      "grad_norm": 43307.9453125,
+      "learning_rate": 0.0007020387649935959,
+      "loss": 0.4768,
+      "step": 517
+    },
+    {
+      "epoch": 86.33333333333333,
+      "grad_norm": 224318.390625,
+      "learning_rate": 0.0007007074988802946,
+      "loss": 0.4983,
+      "step": 518
+    },
+    {
+      "epoch": 86.5,
+      "grad_norm": 232254.71875,
+      "learning_rate": 0.0006993745344626231,
+      "loss": 0.5024,
+      "step": 519
+    },
+    {
+      "epoch": 86.66666666666667,
+      "grad_norm": 95864.921875,
+      "learning_rate": 0.0006980398830195785,
+      "loss": 0.4668,
+      "step": 520
+    },
+    {
+      "epoch": 86.83333333333333,
+      "grad_norm": 156923.25,
+      "learning_rate": 0.0006967035558444326,
+      "loss": 0.4887,
+      "step": 521
+    },
+    {
+      "epoch": 87.0,
+      "grad_norm": 167827.109375,
+      "learning_rate": 0.0006953655642446368,
+      "loss": 0.4985,
+      "step": 522
+    },
+    {
+      "epoch": 87.0,
+      "eval_accuracy": 0.7481264255457803,
+      "eval_best_threshold": 0.4086613953113556,
+      "eval_f1": 0.7982250065257113,
+      "eval_loss": 0.4912816286087036,
+      "eval_pr_auc": 0.8606898747895984,
+      "eval_precision": 0.7168307548054383,
+      "eval_recall": 0.9004711425206124,
+      "eval_roc_auc": 0.841153062039779,
+      "eval_runtime": 3.5763,
+      "eval_samples_per_second": 858.159,
+      "eval_steps_per_second": 1.678,
+      "step": 522
+    },
+    {
+      "epoch": 87.16666666666667,
+      "grad_norm": 99152.1796875,
+      "learning_rate": 0.0006940259195417264,
+      "loss": 0.4867,
+      "step": 523
+    },
+    {
+      "epoch": 87.33333333333333,
+      "grad_norm": 174363.046875,
+      "learning_rate": 0.0006926846330712242,
+      "loss": 0.4909,
+      "step": 524
+    },
+    {
+      "epoch": 87.5,
+      "grad_norm": 181446.703125,
+      "learning_rate": 0.000691341716182545,
+      "loss": 0.4938,
+      "step": 525
+    },
+    {
+      "epoch": 87.66666666666667,
+      "grad_norm": 92542.921875,
+      "learning_rate": 0.0006899971802388996,
+      "loss": 0.4791,
+      "step": 526
+    },
+    {
+      "epoch": 87.83333333333333,
+      "grad_norm": 89334.8046875,
+      "learning_rate": 0.0006886510366171985,
+      "loss": 0.4749,
+      "step": 527
+    },
+    {
+      "epoch": 88.0,
+      "grad_norm": 93162.484375,
+      "learning_rate": 0.0006873032967079561,
+      "loss": 0.488,
+      "step": 528
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.7637666992505702,
+      "eval_best_threshold": 0.4100832939147949,
+      "eval_f1": 0.8053691275167785,
+      "eval_loss": 0.4828181266784668,
+      "eval_pr_auc": 0.8556550349485466,
+      "eval_precision": 0.740009866798224,
+      "eval_recall": 0.8833922261484098,
+      "eval_roc_auc": 0.844597282253374,
+      "eval_runtime": 3.5794,
+      "eval_samples_per_second": 857.413,
+      "eval_steps_per_second": 1.676,
+      "step": 528
+    },
+    {
+      "epoch": 88.16666666666667,
+      "grad_norm": 81783.1015625,
+      "learning_rate": 0.0006859539719151933,
+      "loss": 0.478,
+      "step": 529
+    },
+    {
+      "epoch": 88.33333333333333,
+      "grad_norm": 85834.1875,
+      "learning_rate": 0.0006846030736563422,
+      "loss": 0.4646,
+      "step": 530
+    },
+    {
+      "epoch": 88.5,
+      "grad_norm": 109085.2890625,
+      "learning_rate": 0.0006832506133621487,
+      "loss": 0.4656,
+      "step": 531
+    },
+    {
+      "epoch": 88.66666666666667,
+      "grad_norm": 94811.3515625,
+      "learning_rate": 0.0006818966024765758,
+      "loss": 0.4834,
+      "step": 532
+    },
+    {
+      "epoch": 88.83333333333333,
+      "grad_norm": 119964.9375,
+      "learning_rate": 0.0006805410524567073,
+      "loss": 0.4802,
+      "step": 533
+    },
+    {
+      "epoch": 89.0,
+      "grad_norm": 168818.828125,
+      "learning_rate": 0.0006791839747726501,
+      "loss": 0.4848,
+      "step": 534
+    },
+    {
+      "epoch": 89.0,
+      "eval_accuracy": 0.768654284783317,
+      "eval_best_threshold": 0.41445741057395935,
+      "eval_f1": 0.807796426637791,
+      "eval_loss": 0.4731602966785431,
+      "eval_pr_auc": 0.8652850197288536,
+      "eval_precision": 0.7474949899799599,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.8494418713739681,
+      "eval_runtime": 3.5921,
+      "eval_samples_per_second": 854.381,
+      "eval_steps_per_second": 1.67,
+      "step": 534
+    },
+    {
+      "epoch": 89.16666666666667,
+      "grad_norm": 48302.47265625,
+      "learning_rate": 0.0006778253809074384,
+      "loss": 0.4608,
+      "step": 535
+    },
+    {
+      "epoch": 89.33333333333333,
+      "grad_norm": 186283.75,
+      "learning_rate": 0.0006764652823569344,
+      "loss": 0.4846,
+      "step": 536
+    },
+    {
+      "epoch": 89.5,
+      "grad_norm": 206481.390625,
+      "learning_rate": 0.0006751036906297338,
+      "loss": 0.5059,
+      "step": 537
+    },
+    {
+      "epoch": 89.66666666666667,
+      "grad_norm": 76336.390625,
+      "learning_rate": 0.0006737406172470657,
+      "loss": 0.4759,
+      "step": 538
+    },
+    {
+      "epoch": 89.83333333333333,
+      "grad_norm": 65805.59375,
+      "learning_rate": 0.0006723760737426972,
+      "loss": 0.4615,
+      "step": 539
+    },
+    {
+      "epoch": 90.0,
+      "grad_norm": 52209.296875,
+      "learning_rate": 0.0006710100716628344,
+      "loss": 0.4717,
+      "step": 540
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy": 0.7774519387422614,
+      "eval_best_threshold": 0.4345960021018982,
+      "eval_f1": 0.8100139082058414,
+      "eval_loss": 0.4671099781990051,
+      "eval_pr_auc": 0.8687265581892443,
+      "eval_precision": 0.7675276752767528,
+      "eval_recall": 0.8574793875147232,
+      "eval_roc_auc": 0.8548354394709871,
+      "eval_runtime": 3.4237,
+      "eval_samples_per_second": 896.39,
+      "eval_steps_per_second": 1.752,
+      "step": 540
+    },
+    {
+      "epoch": 90.16666666666667,
+      "grad_norm": 53076.24609375,
+      "learning_rate": 0.0006696426225660256,
+      "loss": 0.4575,
+      "step": 541
+    },
+    {
+      "epoch": 90.33333333333333,
+      "grad_norm": 52598.28515625,
+      "learning_rate": 0.0006682737380230633,
+      "loss": 0.4618,
+      "step": 542
+    },
+    {
+      "epoch": 90.5,
+      "grad_norm": 82825.6015625,
+      "learning_rate": 0.0006669034296168854,
+      "loss": 0.4489,
+      "step": 543
+    },
+    {
+      "epoch": 90.66666666666667,
+      "grad_norm": 62264.2578125,
+      "learning_rate": 0.0006655317089424791,
+      "loss": 0.4657,
+      "step": 544
+    },
+    {
+      "epoch": 90.83333333333333,
+      "grad_norm": 47665.33984375,
+      "learning_rate": 0.0006641585876067806,
+      "loss": 0.4737,
+      "step": 545
+    },
+    {
+      "epoch": 91.0,
+      "grad_norm": 48329.640625,
+      "learning_rate": 0.0006627840772285784,
+      "loss": 0.4799,
+      "step": 546
+    },
+    {
+      "epoch": 91.0,
+      "eval_accuracy": 0.7790811339198436,
+      "eval_best_threshold": 0.4940151870250702,
+      "eval_f1": 0.8072768618533257,
+      "eval_loss": 0.4737241268157959,
+      "eval_pr_auc": 0.8681764998032273,
+      "eval_precision": 0.7802197802197802,
+      "eval_recall": 0.8362779740871613,
+      "eval_roc_auc": 0.8499154623923628,
+      "eval_runtime": 3.3936,
+      "eval_samples_per_second": 904.341,
+      "eval_steps_per_second": 1.768,
+      "step": 546
+    },
+    {
+      "epoch": 91.16666666666667,
+      "grad_norm": 84555.46875,
+      "learning_rate": 0.0006614081894384145,
+      "loss": 0.4674,
+      "step": 547
+    },
+    {
+      "epoch": 91.33333333333333,
+      "grad_norm": 92255.9375,
+      "learning_rate": 0.0006600309358784857,
+      "loss": 0.4891,
+      "step": 548
+    },
+    {
+      "epoch": 91.5,
+      "grad_norm": 66551.3359375,
+      "learning_rate": 0.0006586523282025462,
+      "loss": 0.4754,
+      "step": 549
+    },
+    {
+      "epoch": 91.66666666666667,
+      "grad_norm": 157776.9375,
+      "learning_rate": 0.0006572723780758069,
+      "loss": 0.4624,
+      "step": 550
+    },
+    {
+      "epoch": 91.83333333333333,
+      "grad_norm": 97448.4453125,
+      "learning_rate": 0.000655891097174839,
+      "loss": 0.4661,
+      "step": 551
+    },
+    {
+      "epoch": 92.0,
+      "grad_norm": 73972.796875,
+      "learning_rate": 0.0006545084971874737,
+      "loss": 0.4549,
+      "step": 552
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.7673509286412512,
+      "eval_best_threshold": 0.46922558546066284,
+      "eval_f1": 0.8092948717948718,
+      "eval_loss": 0.475037544965744,
+      "eval_pr_auc": 0.8710701531358553,
+      "eval_precision": 0.7404692082111437,
+      "eval_recall": 0.892226148409894,
+      "eval_roc_auc": 0.8553062383427881,
+      "eval_runtime": 3.6512,
+      "eval_samples_per_second": 840.538,
+      "eval_steps_per_second": 1.643,
+      "step": 552
+    },
+    {
+      "epoch": 92.16666666666667,
+      "grad_norm": 121098.90625,
+      "learning_rate": 0.000653124589812704,
+      "loss": 0.4572,
+      "step": 553
+    },
+    {
+      "epoch": 92.33333333333333,
+      "grad_norm": 56371.390625,
+      "learning_rate": 0.0006517393867605854,
+      "loss": 0.4593,
+      "step": 554
+    },
+    {
+      "epoch": 92.5,
+      "grad_norm": 163690.765625,
+      "learning_rate": 0.0006503528997521365,
+      "loss": 0.4792,
+      "step": 555
+    },
+    {
+      "epoch": 92.66666666666667,
+      "grad_norm": 53005.6796875,
+      "learning_rate": 0.0006489651405192409,
+      "loss": 0.463,
+      "step": 556
+    },
+    {
+      "epoch": 92.83333333333333,
+      "grad_norm": 174509.890625,
+      "learning_rate": 0.000647576120804547,
+      "loss": 0.4603,
+      "step": 557
+    },
+    {
+      "epoch": 93.0,
+      "grad_norm": 189806.515625,
+      "learning_rate": 0.0006461858523613684,
+      "loss": 0.4642,
+      "step": 558
+    },
+    {
+      "epoch": 93.0,
+      "eval_accuracy": 0.7575757575757576,
+      "eval_best_threshold": 0.3204557001590729,
+      "eval_f1": 0.8059467918622848,
+      "eval_loss": 0.4740186333656311,
+      "eval_pr_auc": 0.864116644561191,
+      "eval_precision": 0.723314606741573,
+      "eval_recall": 0.9098939929328622,
+      "eval_roc_auc": 0.8504285730240838,
+      "eval_runtime": 3.6601,
+      "eval_samples_per_second": 838.5,
+      "eval_steps_per_second": 1.639,
+      "step": 558
+    },
+    {
+      "epoch": 93.16666666666667,
+      "grad_norm": 74287.015625,
+      "learning_rate": 0.0006447943469535855,
+      "loss": 0.449,
+      "step": 559
+    },
+    {
+      "epoch": 93.33333333333333,
+      "grad_norm": 51398.22265625,
+      "learning_rate": 0.0006434016163555452,
+      "loss": 0.4631,
+      "step": 560
+    },
+    {
+      "epoch": 93.5,
+      "grad_norm": 69963.7265625,
+      "learning_rate": 0.0006420076723519614,
+      "loss": 0.4422,
+      "step": 561
+    },
+    {
+      "epoch": 93.66666666666667,
+      "grad_norm": 106419.03125,
+      "learning_rate": 0.0006406125267378153,
+      "loss": 0.474,
+      "step": 562
+    },
+    {
+      "epoch": 93.83333333333333,
+      "grad_norm": 123178.09375,
+      "learning_rate": 0.0006392161913182559,
+      "loss": 0.4536,
+      "step": 563
+    },
+    {
+      "epoch": 94.0,
+      "grad_norm": 218872.578125,
+      "learning_rate": 0.0006378186779084996,
+      "loss": 0.489,
+      "step": 564
+    },
+    {
+      "epoch": 94.0,
+      "eval_accuracy": 0.7774519387422614,
+      "eval_best_threshold": 0.4742838740348816,
+      "eval_f1": 0.8142507478923036,
+      "eval_loss": 0.4738493859767914,
+      "eval_pr_auc": 0.8734095284617952,
+      "eval_precision": 0.7564426478019202,
+      "eval_recall": 0.8816254416961131,
+      "eval_roc_auc": 0.8591533867879061,
+      "eval_runtime": 3.6414,
+      "eval_samples_per_second": 842.802,
+      "eval_steps_per_second": 1.648,
+      "step": 564
+    },
+    {
+      "epoch": 94.16666666666667,
+      "grad_norm": 217597.546875,
+      "learning_rate": 0.0006364199983337305,
+      "loss": 0.4792,
+      "step": 565
+    },
+    {
+      "epoch": 94.33333333333333,
+      "grad_norm": 116264.5703125,
+      "learning_rate": 0.0006350201644290005,
+      "loss": 0.4711,
+      "step": 566
+    },
+    {
+      "epoch": 94.5,
+      "grad_norm": 124475.5703125,
+      "learning_rate": 0.0006336191880391284,
+      "loss": 0.4516,
+      "step": 567
+    },
+    {
+      "epoch": 94.66666666666667,
+      "grad_norm": 26623.99609375,
+      "learning_rate": 0.0006322170810186012,
+      "loss": 0.4416,
+      "step": 568
+    },
+    {
+      "epoch": 94.83333333333333,
+      "grad_norm": 85539.7421875,
+      "learning_rate": 0.0006308138552314717,
+      "loss": 0.4391,
+      "step": 569
+    },
+    {
+      "epoch": 95.0,
+      "grad_norm": 116239.9375,
+      "learning_rate": 0.0006294095225512603,
+      "loss": 0.4649,
+      "step": 570
+    },
+    {
+      "epoch": 95.0,
+      "eval_accuracy": 0.7673509286412512,
+      "eval_best_threshold": 0.4234827756881714,
+      "eval_f1": 0.80878414568827,
+      "eval_loss": 0.4616597294807434,
+      "eval_pr_auc": 0.8767909548555617,
+      "eval_precision": 0.7416502946954814,
+      "eval_recall": 0.889281507656066,
+      "eval_roc_auc": 0.8578664219887129,
+      "eval_runtime": 3.6402,
+      "eval_samples_per_second": 843.075,
+      "eval_steps_per_second": 1.648,
+      "step": 570
+    },
+    {
+      "epoch": 95.16666666666667,
+      "grad_norm": 53956.39453125,
+      "learning_rate": 0.0006280040948608529,
+      "loss": 0.4726,
+      "step": 571
+    },
+    {
+      "epoch": 95.33333333333333,
+      "grad_norm": 132776.96875,
+      "learning_rate": 0.000626597584052401,
+      "loss": 0.4812,
+      "step": 572
+    },
+    {
+      "epoch": 95.5,
+      "grad_norm": 110310.21875,
+      "learning_rate": 0.0006251900020272208,
+      "loss": 0.4419,
+      "step": 573
+    },
+    {
+      "epoch": 95.66666666666667,
+      "grad_norm": 213141.046875,
+      "learning_rate": 0.000623781360695693,
+      "loss": 0.479,
+      "step": 574
+    },
+    {
+      "epoch": 95.83333333333333,
+      "grad_norm": 265038.9375,
+      "learning_rate": 0.0006223716719771619,
+      "loss": 0.5025,
+      "step": 575
+    },
+    {
+      "epoch": 96.0,
+      "grad_norm": 100577.78125,
+      "learning_rate": 0.0006209609477998338,
+      "loss": 0.445,
+      "step": 576
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.7653958944281525,
+      "eval_best_threshold": 0.4986065626144409,
+      "eval_f1": 0.8094229751191107,
+      "eval_loss": 0.48587265610694885,
+      "eval_pr_auc": 0.8728447339217204,
+      "eval_precision": 0.7350961538461539,
+      "eval_recall": 0.9004711425206124,
+      "eval_roc_auc": 0.8566103855825578,
+      "eval_runtime": 3.6655,
+      "eval_samples_per_second": 837.276,
+      "eval_steps_per_second": 1.637,
+      "step": 576
+    },
+    {
+      "epoch": 96.16666666666667,
+      "grad_norm": 218718.828125,
+      "learning_rate": 0.0006195492001006772,
+      "loss": 0.4864,
+      "step": 577
+    },
+    {
+      "epoch": 96.33333333333333,
+      "grad_norm": 238896.375,
+      "learning_rate": 0.0006181364408253208,
+      "loss": 0.4732,
+      "step": 578
+    },
+    {
+      "epoch": 96.5,
+      "grad_norm": 144929.984375,
+      "learning_rate": 0.0006167226819279528,
+      "loss": 0.4766,
+      "step": 579
+    },
+    {
+      "epoch": 96.66666666666667,
+      "grad_norm": 345105.125,
+      "learning_rate": 0.0006153079353712201,
+      "loss": 0.5188,
+      "step": 580
+    },
+    {
+      "epoch": 96.83333333333333,
+      "grad_norm": 405794.65625,
+      "learning_rate": 0.0006138922131261268,
+      "loss": 0.5247,
+      "step": 581
+    },
+    {
+      "epoch": 97.0,
+      "grad_norm": 258567.859375,
+      "learning_rate": 0.0006124755271719325,
+      "loss": 0.4781,
+      "step": 582
+    },
+    {
+      "epoch": 97.0,
+      "eval_accuracy": 0.7758227435646791,
+      "eval_best_threshold": 0.4811469614505768,
+      "eval_f1": 0.8101545253863135,
+      "eval_loss": 0.4663538932800293,
+      "eval_pr_auc": 0.8747418466683401,
+      "eval_precision": 0.7622014537902388,
+      "eval_recall": 0.8645465253239105,
+      "eval_roc_auc": 0.8566623624653023,
+      "eval_runtime": 3.6983,
+      "eval_samples_per_second": 829.85,
+      "eval_steps_per_second": 1.622,
+      "step": 582
+    },
+    {
+      "epoch": 97.16666666666667,
+      "grad_norm": 42506.12109375,
+      "learning_rate": 0.0006110578894960517,
+      "loss": 0.453,
+      "step": 583
+    },
+    {
+      "epoch": 97.33333333333333,
+      "grad_norm": 252258.21875,
+      "learning_rate": 0.0006096393120939516,
+      "loss": 0.4965,
+      "step": 584
+    },
+    {
+      "epoch": 97.5,
+      "grad_norm": 216667.84375,
+      "learning_rate": 0.0006082198069690515,
+      "loss": 0.4718,
+      "step": 585
+    },
+    {
+      "epoch": 97.66666666666667,
+      "grad_norm": 63770.66015625,
+      "learning_rate": 0.0006067993861326201,
+      "loss": 0.439,
+      "step": 586
+    },
+    {
+      "epoch": 97.83333333333333,
+      "grad_norm": 117137.3984375,
+      "learning_rate": 0.0006053780616036748,
+      "loss": 0.4545,
+      "step": 587
+    },
+    {
+      "epoch": 98.0,
+      "grad_norm": 81328.671875,
+      "learning_rate": 0.0006039558454088796,
+      "loss": 0.4616,
+      "step": 588
+    },
+    {
+      "epoch": 98.0,
+      "eval_accuracy": 0.7820136852394917,
+      "eval_best_threshold": 0.5103572607040405,
+      "eval_f1": 0.8164609053497942,
+      "eval_loss": 0.46580880880355835,
+      "eval_pr_auc": 0.8780657638908369,
+      "eval_precision": 0.7642526964560863,
+      "eval_recall": 0.8763250883392226,
+      "eval_roc_auc": 0.8624738934293489,
+      "eval_runtime": 3.7401,
+      "eval_samples_per_second": 820.564,
+      "eval_steps_per_second": 1.604,
+      "step": 588
+    },
+    {
+      "epoch": 98.16666666666667,
+      "grad_norm": 96831.2265625,
+      "learning_rate": 0.0006025327495824439,
+      "loss": 0.4453,
+      "step": 589
+    },
+    {
+      "epoch": 98.33333333333333,
+      "grad_norm": 140212.84375,
+      "learning_rate": 0.000601108786166019,
+      "loss": 0.4752,
+      "step": 590
+    },
+    {
+      "epoch": 98.5,
+      "grad_norm": 32939.4609375,
+      "learning_rate": 0.0005996839672085986,
+      "loss": 0.4524,
+      "step": 591
+    },
+    {
+      "epoch": 98.66666666666667,
+      "grad_norm": 217863.0625,
+      "learning_rate": 0.0005982583047664151,
+      "loss": 0.4733,
+      "step": 592
+    },
+    {
+      "epoch": 98.83333333333333,
+      "grad_norm": 245090.515625,
+      "learning_rate": 0.0005968318109028381,
+      "loss": 0.4867,
+      "step": 593
+    },
+    {
+      "epoch": 99.0,
+      "grad_norm": 75009.2265625,
+      "learning_rate": 0.0005954044976882724,
+      "loss": 0.4464,
+      "step": 594
+    },
+    {
+      "epoch": 99.0,
+      "eval_accuracy": 0.7813620071684588,
+      "eval_best_threshold": 0.5514122247695923,
+      "eval_f1": 0.8162147356888524,
+      "eval_loss": 0.4710569381713867,
+      "eval_pr_auc": 0.8805845064525083,
+      "eval_precision": 0.7629288274449565,
+      "eval_recall": 0.8775029446407538,
+      "eval_roc_auc": 0.8641551952397766,
+      "eval_runtime": 3.6598,
+      "eval_samples_per_second": 838.564,
+      "eval_steps_per_second": 1.639,
+      "step": 594
+    },
+    {
+      "epoch": 99.16666666666667,
+      "grad_norm": 174301.109375,
+      "learning_rate": 0.0005939763772000559,
+      "loss": 0.464,
+      "step": 595
+    },
+    {
+      "epoch": 99.33333333333333,
+      "grad_norm": 214547.453125,
+      "learning_rate": 0.0005925474615223572,
+      "loss": 0.4796,
+      "step": 596
+    },
+    {
+      "epoch": 99.5,
+      "grad_norm": 73995.5703125,
+      "learning_rate": 0.0005911177627460738,
+      "loss": 0.4753,
+      "step": 597
+    },
+    {
+      "epoch": 99.66666666666667,
+      "grad_norm": 254424.515625,
+      "learning_rate": 0.0005896872929687287,
+      "loss": 0.4835,
+      "step": 598
+    },
+    {
+      "epoch": 99.83333333333333,
+      "grad_norm": 196407.71875,
+      "learning_rate": 0.0005882560642943696,
+      "loss": 0.4804,
+      "step": 599
+    },
+    {
+      "epoch": 100.0,
+      "grad_norm": 59902.87890625,
+      "learning_rate": 0.0005868240888334653,
+      "loss": 0.4595,
+      "step": 600
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.7758227435646791,
+      "eval_best_threshold": 0.5762914419174194,
+      "eval_f1": 0.8124318429661941,
+      "eval_loss": 0.48100268840789795,
+      "eval_pr_auc": 0.8751349295776516,
+      "eval_precision": 0.7563451776649747,
+      "eval_recall": 0.8775029446407538,
+      "eval_roc_auc": 0.8608797925048475,
+      "eval_runtime": 3.7446,
+      "eval_samples_per_second": 819.576,
+      "eval_steps_per_second": 1.602,
+      "step": 600
+    },
+    {
+      "epoch": 100.16666666666667,
+      "grad_norm": 185637.796875,
+      "learning_rate": 0.0005853913787028035,
+      "loss": 0.4682,
+      "step": 601
+    },
+    {
+      "epoch": 100.33333333333333,
+      "grad_norm": 145423.65625,
+      "learning_rate": 0.0005839579460253886,
+      "loss": 0.4791,
+      "step": 602
+    },
+    {
+      "epoch": 100.5,
+      "grad_norm": 82947.40625,
+      "learning_rate": 0.0005825238029303388,
+      "loss": 0.4639,
+      "step": 603
+    },
+    {
+      "epoch": 100.66666666666667,
+      "grad_norm": 125656.6875,
+      "learning_rate": 0.0005810889615527838,
+      "loss": 0.462,
+      "step": 604
+    },
+    {
+      "epoch": 100.83333333333333,
+      "grad_norm": 42078.859375,
+      "learning_rate": 0.0005796534340337614,
+      "loss": 0.4258,
+      "step": 605
+    },
+    {
+      "epoch": 101.0,
+      "grad_norm": 35350.0625,
+      "learning_rate": 0.0005782172325201155,
+      "loss": 0.4494,
+      "step": 606
+    },
+    {
+      "epoch": 101.0,
+      "eval_accuracy": 0.7741935483870968,
+      "eval_best_threshold": 0.3599238991737366,
+      "eval_f1": 0.8139597315436241,
+      "eval_loss": 0.46240806579589844,
+      "eval_pr_auc": 0.8759762303440317,
+      "eval_precision": 0.747903305377405,
+      "eval_recall": 0.8928150765606596,
+      "eval_roc_auc": 0.8611521341879879,
+      "eval_runtime": 3.7767,
+      "eval_samples_per_second": 812.608,
+      "eval_steps_per_second": 1.589,
+      "step": 606
+    },
+    {
+      "epoch": 101.16666666666667,
+      "grad_norm": 99813.265625,
+      "learning_rate": 0.000576780369164393,
+      "loss": 0.426,
+      "step": 607
+    },
+    {
+      "epoch": 101.33333333333333,
+      "grad_norm": 104900.5703125,
+      "learning_rate": 0.0005753428561247416,
+      "loss": 0.4602,
+      "step": 608
+    },
+    {
+      "epoch": 101.5,
+      "grad_norm": 99518.796875,
+      "learning_rate": 0.0005739047055648054,
+      "loss": 0.4712,
+      "step": 609
+    },
+    {
+      "epoch": 101.66666666666667,
+      "grad_norm": 86323.15625,
+      "learning_rate": 0.0005724659296536234,
+      "loss": 0.4648,
+      "step": 610
+    },
+    {
+      "epoch": 101.83333333333333,
+      "grad_norm": 62814.87890625,
+      "learning_rate": 0.0005710265405655262,
+      "loss": 0.4333,
+      "step": 611
+    },
+    {
+      "epoch": 102.0,
+      "grad_norm": 71986.578125,
+      "learning_rate": 0.0005695865504800327,
+      "loss": 0.4499,
+      "step": 612
+    },
+    {
+      "epoch": 102.0,
+      "eval_accuracy": 0.7803844900619095,
+      "eval_best_threshold": 0.3741658627986908,
+      "eval_f1": 0.8187197417966648,
+      "eval_loss": 0.44404909014701843,
+      "eval_pr_auc": 0.8861759100281312,
+      "eval_precision": 0.7534653465346535,
+      "eval_recall": 0.8963486454652533,
+      "eval_roc_auc": 0.8697824445286384,
+      "eval_runtime": 3.7218,
+      "eval_samples_per_second": 824.608,
+      "eval_steps_per_second": 1.612,
+      "step": 612
+    },
+    {
+      "epoch": 102.16666666666667,
+      "grad_norm": 43484.6484375,
+      "learning_rate": 0.0005681459715817473,
+      "loss": 0.452,
+      "step": 613
+    },
+    {
+      "epoch": 102.33333333333333,
+      "grad_norm": 138327.375,
+      "learning_rate": 0.0005667048160602564,
+      "loss": 0.4387,
+      "step": 614
+    },
+    {
+      "epoch": 102.5,
+      "grad_norm": 70967.2734375,
+      "learning_rate": 0.000565263096110026,
+      "loss": 0.4281,
+      "step": 615
+    },
+    {
+      "epoch": 102.66666666666667,
+      "grad_norm": 100818.53125,
+      "learning_rate": 0.0005638208239302974,
+      "loss": 0.4618,
+      "step": 616
+    },
+    {
+      "epoch": 102.83333333333333,
+      "grad_norm": 50325.85546875,
+      "learning_rate": 0.0005623780117249852,
+      "loss": 0.4533,
+      "step": 617
+    },
+    {
+      "epoch": 103.0,
+      "grad_norm": 149581.5625,
+      "learning_rate": 0.0005609346717025737,
+      "loss": 0.4503,
+      "step": 618
+    },
+    {
+      "epoch": 103.0,
+      "eval_accuracy": 0.7908113391984359,
+      "eval_best_threshold": 0.48392462730407715,
+      "eval_f1": 0.8210702341137124,
+      "eval_loss": 0.44357526302337646,
+      "eval_pr_auc": 0.8885960306219365,
+      "eval_precision": 0.7793650793650794,
+      "eval_recall": 0.8674911660777385,
+      "eval_roc_auc": 0.8710221576162457,
+      "eval_runtime": 3.7582,
+      "eval_samples_per_second": 816.614,
+      "eval_steps_per_second": 1.597,
+      "step": 618
+    },
+    {
+      "epoch": 103.16666666666667,
+      "grad_norm": 82695.859375,
+      "learning_rate": 0.0005594908160760127,
+      "loss": 0.4568,
+      "step": 619
+    },
+    {
+      "epoch": 103.33333333333333,
+      "grad_norm": 107545.03125,
+      "learning_rate": 0.0005580464570626152,
+      "loss": 0.4497,
+      "step": 620
+    },
+    {
+      "epoch": 103.5,
+      "grad_norm": 125753.75,
+      "learning_rate": 0.0005566016068839535,
+      "loss": 0.4361,
+      "step": 621
+    },
+    {
+      "epoch": 103.66666666666667,
+      "grad_norm": 56653.78125,
+      "learning_rate": 0.0005551562777657559,
+      "loss": 0.4209,
+      "step": 622
+    },
+    {
+      "epoch": 103.83333333333333,
+      "grad_norm": 41988.1796875,
+      "learning_rate": 0.0005537104819378037,
+      "loss": 0.4091,
+      "step": 623
+    },
+    {
+      "epoch": 104.0,
+      "grad_norm": 80381.1328125,
+      "learning_rate": 0.0005522642316338268,
+      "loss": 0.4407,
+      "step": 624
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.7914630172694689,
+      "eval_best_threshold": 0.462395578622818,
+      "eval_f1": 0.819718309859155,
+      "eval_loss": 0.44202810525894165,
+      "eval_pr_auc": 0.8876987764595818,
+      "eval_precision": 0.7856371490280778,
+      "eval_recall": 0.8568904593639576,
+      "eval_roc_auc": 0.871470619315297,
+      "eval_runtime": 3.7572,
+      "eval_samples_per_second": 816.841,
+      "eval_steps_per_second": 1.597,
+      "step": 624
+    },
+    {
+      "epoch": 104.16666666666667,
+      "grad_norm": 78504.8828125,
+      "learning_rate": 0.000550817539091401,
+      "loss": 0.4259,
+      "step": 625
+    },
+    {
+      "epoch": 104.33333333333333,
+      "grad_norm": 67419.703125,
+      "learning_rate": 0.000549370416551844,
+      "loss": 0.4346,
+      "step": 626
+    },
+    {
+      "epoch": 104.5,
+      "grad_norm": 166526.515625,
+      "learning_rate": 0.000547922876260112,
+      "loss": 0.4574,
+      "step": 627
+    },
+    {
+      "epoch": 104.66666666666667,
+      "grad_norm": 87268.5703125,
+      "learning_rate": 0.0005464749304646962,
+      "loss": 0.429,
+      "step": 628
+    },
+    {
+      "epoch": 104.83333333333333,
+      "grad_norm": 197419.03125,
+      "learning_rate": 0.0005450265914175187,
+      "loss": 0.438,
+      "step": 629
+    },
+    {
+      "epoch": 105.0,
+      "grad_norm": 163018.390625,
+      "learning_rate": 0.0005435778713738292,
+      "loss": 0.4512,
+      "step": 630
+    },
+    {
+      "epoch": 105.0,
+      "eval_accuracy": 0.7875529488432714,
+      "eval_best_threshold": 0.4379901885986328,
+      "eval_f1": 0.8206820682068207,
+      "eval_loss": 0.4458414316177368,
+      "eval_pr_auc": 0.8821672284197818,
+      "eval_precision": 0.7698658410732714,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.8695427494825937,
+      "eval_runtime": 3.7782,
+      "eval_samples_per_second": 812.298,
+      "eval_steps_per_second": 1.588,
+      "step": 630
+    },
+    {
+      "epoch": 105.16666666666667,
+      "grad_norm": 89169.0546875,
+      "learning_rate": 0.0005421287825921014,
+      "loss": 0.4349,
+      "step": 631
+    },
+    {
+      "epoch": 105.33333333333333,
+      "grad_norm": 61004.5703125,
+      "learning_rate": 0.0005406793373339292,
+      "loss": 0.4277,
+      "step": 632
+    },
+    {
+      "epoch": 105.5,
+      "grad_norm": 52870.4921875,
+      "learning_rate": 0.0005392295478639225,
+      "loss": 0.4479,
+      "step": 633
+    },
+    {
+      "epoch": 105.66666666666667,
+      "grad_norm": 127358.6796875,
+      "learning_rate": 0.0005377794264496041,
+      "loss": 0.4339,
+      "step": 634
+    },
+    {
+      "epoch": 105.83333333333333,
+      "grad_norm": 81547.8203125,
+      "learning_rate": 0.0005363289853613053,
+      "loss": 0.4399,
+      "step": 635
+    },
+    {
+      "epoch": 106.0,
+      "grad_norm": 82232.0390625,
+      "learning_rate": 0.0005348782368720626,
+      "loss": 0.4348,
+      "step": 636
+    },
+    {
+      "epoch": 106.0,
+      "eval_accuracy": 0.7882046269143044,
+      "eval_best_threshold": 0.39316704869270325,
+      "eval_f1": 0.823177366702938,
+      "eval_loss": 0.4380728304386139,
+      "eval_pr_auc": 0.887453308801846,
+      "eval_precision": 0.7649140546006067,
+      "eval_recall": 0.8910482921083628,
+      "eval_roc_auc": 0.8738138746489412,
+      "eval_runtime": 4.2587,
+      "eval_samples_per_second": 720.643,
+      "eval_steps_per_second": 1.409,
+      "step": 636
+    },
+    {
+      "epoch": 106.16666666666667,
+      "grad_norm": 57085.0390625,
+      "learning_rate": 0.0005334271932575137,
+      "loss": 0.4276,
+      "step": 637
+    },
+    {
+      "epoch": 106.33333333333333,
+      "grad_norm": 73160.03125,
+      "learning_rate": 0.0005319758667957928,
+      "loss": 0.4374,
+      "step": 638
+    },
+    {
+      "epoch": 106.5,
+      "grad_norm": 94396.4921875,
+      "learning_rate": 0.0005305242697674285,
+      "loss": 0.4285,
+      "step": 639
+    },
+    {
+      "epoch": 106.66666666666667,
+      "grad_norm": 44217.97265625,
+      "learning_rate": 0.0005290724144552379,
+      "loss": 0.4334,
+      "step": 640
+    },
+    {
+      "epoch": 106.83333333333333,
+      "grad_norm": 47426.1015625,
+      "learning_rate": 0.0005276203131442243,
+      "loss": 0.424,
+      "step": 641
+    },
+    {
+      "epoch": 107.0,
+      "grad_norm": 45280.8984375,
+      "learning_rate": 0.000526167978121472,
+      "loss": 0.4096,
+      "step": 642
+    },
+    {
+      "epoch": 107.0,
+      "eval_accuracy": 0.7963506028022157,
+      "eval_best_threshold": 0.4691428244113922,
+      "eval_f1": 0.826533444351929,
+      "eval_loss": 0.43517568707466125,
+      "eval_pr_auc": 0.8899689908598397,
+      "eval_precision": 0.7816272965879265,
+      "eval_recall": 0.8769140164899882,
+      "eval_roc_auc": 0.8777233094411498,
+      "eval_runtime": 3.8895,
+      "eval_samples_per_second": 789.042,
+      "eval_steps_per_second": 1.543,
+      "step": 642
+    },
+    {
+      "epoch": 107.16666666666667,
+      "grad_norm": 79869.1796875,
+      "learning_rate": 0.0005247154216760431,
+      "loss": 0.4257,
+      "step": 643
+    },
+    {
+      "epoch": 107.33333333333333,
+      "grad_norm": 120783.3203125,
+      "learning_rate": 0.0005232626560988735,
+      "loss": 0.4233,
+      "step": 644
+    },
+    {
+      "epoch": 107.5,
+      "grad_norm": 64844.3515625,
+      "learning_rate": 0.0005218096936826681,
+      "loss": 0.4262,
+      "step": 645
+    },
+    {
+      "epoch": 107.66666666666667,
+      "grad_norm": 144929.3125,
+      "learning_rate": 0.000520356546721798,
+      "loss": 0.4377,
+      "step": 646
+    },
+    {
+      "epoch": 107.83333333333333,
+      "grad_norm": 81265.203125,
+      "learning_rate": 0.0005189032275121955,
+      "loss": 0.4215,
+      "step": 647
+    },
+    {
+      "epoch": 108.0,
+      "grad_norm": 280190.40625,
+      "learning_rate": 0.0005174497483512506,
+      "loss": 0.4443,
+      "step": 648
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.7956989247311828,
+      "eval_best_threshold": 0.5434492230415344,
+      "eval_f1": 0.8280778722237455,
+      "eval_loss": 0.46423980593681335,
+      "eval_pr_auc": 0.891939864139048,
+      "eval_precision": 0.7747562852744997,
+      "eval_recall": 0.889281507656066,
+      "eval_roc_auc": 0.8753121834672275,
+      "eval_runtime": 3.7395,
+      "eval_samples_per_second": 820.687,
+      "eval_steps_per_second": 1.604,
+      "step": 648
+    },
+    {
+      "epoch": 108.16666666666667,
+      "grad_norm": 245879.5,
+      "learning_rate": 0.0005159961215377065,
+      "loss": 0.4513,
+      "step": 649
+    },
+    {
+      "epoch": 108.33333333333333,
+      "grad_norm": 51109.69921875,
+      "learning_rate": 0.0005145423593715557,
+      "loss": 0.4371,
+      "step": 650
+    },
+    {
+      "epoch": 108.5,
+      "grad_norm": 309565.03125,
+      "learning_rate": 0.0005130884741539367,
+      "loss": 0.4463,
+      "step": 651
+    },
+    {
+      "epoch": 108.66666666666667,
+      "grad_norm": 229982.4375,
+      "learning_rate": 0.0005116344781870281,
+      "loss": 0.4492,
+      "step": 652
+    },
+    {
+      "epoch": 108.83333333333333,
+      "grad_norm": 80197.65625,
+      "learning_rate": 0.0005101803837739468,
+      "loss": 0.4464,
+      "step": 653
+    },
+    {
+      "epoch": 109.0,
+      "grad_norm": 138026.71875,
+      "learning_rate": 0.0005087262032186418,
+      "loss": 0.449,
+      "step": 654
+    },
+    {
+      "epoch": 109.0,
+      "eval_accuracy": 0.7927663734115347,
+      "eval_best_threshold": 0.4488154947757721,
+      "eval_f1": 0.8223463687150838,
+      "eval_loss": 0.43212890625,
+      "eval_pr_auc": 0.8937612583840494,
+      "eval_precision": 0.7821466524973433,
+      "eval_recall": 0.866902237926973,
+      "eval_roc_auc": 0.8779436742415456,
+      "eval_runtime": 3.8059,
+      "eval_samples_per_second": 806.371,
+      "eval_steps_per_second": 1.576,
+      "step": 654
+    },
+    {
+      "epoch": 109.16666666666667,
+      "grad_norm": 61663.25390625,
+      "learning_rate": 0.0005072719488257914,
+      "loss": 0.4213,
+      "step": 655
+    },
+    {
+      "epoch": 109.33333333333333,
+      "grad_norm": 72922.8671875,
+      "learning_rate": 0.0005058176329006986,
+      "loss": 0.4282,
+      "step": 656
+    },
+    {
+      "epoch": 109.5,
+      "grad_norm": 150465.5625,
+      "learning_rate": 0.0005043632677491869,
+      "loss": 0.4205,
+      "step": 657
+    },
+    {
+      "epoch": 109.66666666666667,
+      "grad_norm": 131988.890625,
+      "learning_rate": 0.000502908865677497,
+      "loss": 0.435,
+      "step": 658
+    },
+    {
+      "epoch": 109.83333333333333,
+      "grad_norm": 119979.9375,
+      "learning_rate": 0.0005014544389921811,
+      "loss": 0.4199,
+      "step": 659
+    },
+    {
+      "epoch": 110.0,
+      "grad_norm": 69464.1171875,
+      "learning_rate": 0.0005,
+      "loss": 0.4125,
+      "step": 660
+    },
+    {
+      "epoch": 110.0,
+      "eval_accuracy": 0.7859237536656891,
+      "eval_best_threshold": 0.3834899067878723,
+      "eval_f1": 0.8240963855421687,
+      "eval_loss": 0.4334057867527008,
+      "eval_pr_auc": 0.8916199472812893,
+      "eval_precision": 0.7555228276877761,
+      "eval_recall": 0.9063604240282686,
+      "eval_roc_auc": 0.8792104496730611,
+      "eval_runtime": 3.7717,
+      "eval_samples_per_second": 813.687,
+      "eval_steps_per_second": 1.591,
+      "step": 660
+    },
+    {
+      "epoch": 110.16666666666667,
+      "grad_norm": 87378.875,
+      "learning_rate": 0.0004985455610078191,
+      "loss": 0.4207,
+      "step": 661
+    },
+    {
+      "epoch": 110.33333333333333,
+      "grad_norm": 71451.0546875,
+      "learning_rate": 0.0004970911343225031,
+      "loss": 0.4227,
+      "step": 662
+    },
+    {
+      "epoch": 110.5,
+      "grad_norm": 115093.7578125,
+      "learning_rate": 0.0004956367322508131,
+      "loss": 0.4317,
+      "step": 663
+    },
+    {
+      "epoch": 110.66666666666667,
+      "grad_norm": 77691.953125,
+      "learning_rate": 0.0004941823670993016,
+      "loss": 0.4072,
+      "step": 664
+    },
+    {
+      "epoch": 110.83333333333333,
+      "grad_norm": 123772.8359375,
+      "learning_rate": 0.0004927280511742086,
+      "loss": 0.4193,
+      "step": 665
+    },
+    {
+      "epoch": 111.0,
+      "grad_norm": 71867.2109375,
+      "learning_rate": 0.0004912737967813582,
+      "loss": 0.4164,
+      "step": 666
+    },
+    {
+      "epoch": 111.0,
+      "eval_accuracy": 0.7924405343760182,
+      "eval_best_threshold": 0.3640596866607666,
+      "eval_f1": 0.8279773156899811,
+      "eval_loss": 0.42901697754859924,
+      "eval_pr_auc": 0.8968072773162334,
+      "eval_precision": 0.7645885286783043,
+      "eval_recall": 0.9028268551236749,
+      "eval_roc_auc": 0.8803616731916986,
+      "eval_runtime": 3.7708,
+      "eval_samples_per_second": 813.877,
+      "eval_steps_per_second": 1.591,
+      "step": 666
+    },
+    {
+      "epoch": 111.16666666666667,
+      "grad_norm": 63211.99609375,
+      "learning_rate": 0.0004898196162260531,
+      "loss": 0.422,
+      "step": 667
+    },
+    {
+      "epoch": 111.33333333333333,
+      "grad_norm": 112762.171875,
+      "learning_rate": 0.0004883655218129719,
+      "loss": 0.4188,
+      "step": 668
+    },
+    {
+      "epoch": 111.5,
+      "grad_norm": 165300.15625,
+      "learning_rate": 0.0004869115258460635,
+      "loss": 0.4226,
+      "step": 669
+    },
+    {
+      "epoch": 111.66666666666667,
+      "grad_norm": 96502.1953125,
+      "learning_rate": 0.0004854576406284443,
+      "loss": 0.4226,
+      "step": 670
+    },
+    {
+      "epoch": 111.83333333333333,
+      "grad_norm": 103810.5546875,
+      "learning_rate": 0.0004840038784622937,
+      "loss": 0.4239,
+      "step": 671
+    },
+    {
+      "epoch": 112.0,
+      "grad_norm": 110581.7421875,
+      "learning_rate": 0.0004825502516487497,
+      "loss": 0.4106,
+      "step": 672
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.7917888563049853,
+      "eval_best_threshold": 0.3643791079521179,
+      "eval_f1": 0.8244023083264633,
+      "eval_loss": 0.43213963508605957,
+      "eval_pr_auc": 0.8929182333546763,
+      "eval_precision": 0.7727975270479135,
+      "eval_recall": 0.8833922261484098,
+      "eval_roc_auc": 0.8802027356163642,
+      "eval_runtime": 3.8148,
+      "eval_samples_per_second": 804.496,
+      "eval_steps_per_second": 1.573,
+      "step": 672
+    },
+    {
+      "epoch": 112.16666666666667,
+      "grad_norm": 70169.6171875,
+      "learning_rate": 0.0004810967724878046,
+      "loss": 0.4186,
+      "step": 673
+    },
+    {
+      "epoch": 112.33333333333333,
+      "grad_norm": 100069.8359375,
+      "learning_rate": 0.0004796434532782021,
+      "loss": 0.4334,
+      "step": 674
+    },
+    {
+      "epoch": 112.5,
+      "grad_norm": 54957.6328125,
+      "learning_rate": 0.0004781903063173321,
+      "loss": 0.3984,
+      "step": 675
+    },
+    {
+      "epoch": 112.66666666666667,
+      "grad_norm": 144818.28125,
+      "learning_rate": 0.0004767373439011267,
+      "loss": 0.4173,
+      "step": 676
+    },
+    {
+      "epoch": 112.83333333333333,
+      "grad_norm": 128438.8046875,
+      "learning_rate": 0.00047528457832395705,
+      "loss": 0.4267,
+      "step": 677
+    },
+    {
+      "epoch": 113.0,
+      "grad_norm": 191549.34375,
+      "learning_rate": 0.0004738320218785281,
+      "loss": 0.434,
+      "step": 678
+    },
+    {
+      "epoch": 113.0,
+      "eval_accuracy": 0.7898338220918866,
+      "eval_best_threshold": 0.5195454955101013,
+      "eval_f1": 0.8216754216201272,
+      "eval_loss": 0.44215479493141174,
+      "eval_pr_auc": 0.895288529089388,
+      "eval_precision": 0.7743616466909848,
+      "eval_recall": 0.8751472320376914,
+      "eval_roc_auc": 0.8801997286892632,
+      "eval_runtime": 3.8745,
+      "eval_samples_per_second": 792.093,
+      "eval_steps_per_second": 1.549,
+      "step": 678
+    },
+    {
+      "epoch": 113.16666666666667,
+      "grad_norm": 180644.59375,
+      "learning_rate": 0.0004723796868557758,
+      "loss": 0.4344,
+      "step": 679
+    },
+    {
+      "epoch": 113.33333333333333,
+      "grad_norm": 89088.8515625,
+      "learning_rate": 0.0004709275855447621,
+      "loss": 0.4423,
+      "step": 680
+    },
+    {
+      "epoch": 113.5,
+      "grad_norm": 154950.390625,
+      "learning_rate": 0.00046947573023257153,
+      "loss": 0.4279,
+      "step": 681
+    },
+    {
+      "epoch": 113.66666666666667,
+      "grad_norm": 75882.7578125,
+      "learning_rate": 0.0004680241332042072,
+      "loss": 0.4217,
+      "step": 682
+    },
+    {
+      "epoch": 113.83333333333333,
+      "grad_norm": 178440.0,
+      "learning_rate": 0.00046657280674248646,
+      "loss": 0.4251,
+      "step": 683
+    },
+    {
+      "epoch": 114.0,
+      "grad_norm": 185205.1875,
+      "learning_rate": 0.00046512176312793734,
+      "loss": 0.4287,
+      "step": 684
+    },
+    {
+      "epoch": 114.0,
+      "eval_accuracy": 0.7973281199087651,
+      "eval_best_threshold": 0.43903276324272156,
+      "eval_f1": 0.827318156579678,
+      "eval_loss": 0.429295152425766,
+      "eval_pr_auc": 0.8894820330082692,
+      "eval_precision": 0.782563025210084,
+      "eval_recall": 0.8775029446407538,
+      "eval_roc_auc": 0.8793955904702748,
+      "eval_runtime": 3.9972,
+      "eval_samples_per_second": 767.793,
+      "eval_steps_per_second": 1.501,
+      "step": 684
+    },
+    {
+      "epoch": 114.16666666666667,
+      "grad_norm": 57033.11328125,
+      "learning_rate": 0.0004636710146386947,
+      "loss": 0.4343,
+      "step": 685
+    },
+    {
+      "epoch": 114.33333333333333,
+      "grad_norm": 190587.890625,
+      "learning_rate": 0.0004622205735503961,
+      "loss": 0.4285,
+      "step": 686
+    },
+    {
+      "epoch": 114.5,
+      "grad_norm": 188324.6875,
+      "learning_rate": 0.0004607704521360776,
+      "loss": 0.427,
+      "step": 687
+    },
+    {
+      "epoch": 114.66666666666667,
+      "grad_norm": 77043.28125,
+      "learning_rate": 0.00045932066266607093,
+      "loss": 0.4032,
+      "step": 688
+    },
+    {
+      "epoch": 114.83333333333333,
+      "grad_norm": 104234.28125,
+      "learning_rate": 0.0004578712174078986,
+      "loss": 0.4,
+      "step": 689
+    },
+    {
+      "epoch": 115.0,
+      "grad_norm": 109458.25,
+      "learning_rate": 0.00045642212862617086,
+      "loss": 0.4238,
+      "step": 690
+    },
+    {
+      "epoch": 115.0,
+      "eval_accuracy": 0.7983056370153144,
+      "eval_best_threshold": 0.35520094633102417,
+      "eval_f1": 0.8281987232861504,
+      "eval_loss": 0.43845149874687195,
+      "eval_pr_auc": 0.8946518775878574,
+      "eval_precision": 0.7832020997375329,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.8802770496718584,
+      "eval_runtime": 3.8083,
+      "eval_samples_per_second": 805.875,
+      "eval_steps_per_second": 1.576,
+      "step": 690
+    },
+    {
+      "epoch": 115.16666666666667,
+      "grad_norm": 164871.96875,
+      "learning_rate": 0.0004549734085824816,
+      "loss": 0.4166,
+      "step": 691
+    },
+    {
+      "epoch": 115.33333333333333,
+      "grad_norm": 112055.2265625,
+      "learning_rate": 0.0004535250695353039,
+      "loss": 0.4082,
+      "step": 692
+    },
+    {
+      "epoch": 115.5,
+      "grad_norm": 155252.046875,
+      "learning_rate": 0.000452077123739888,
+      "loss": 0.4193,
+      "step": 693
+    },
+    {
+      "epoch": 115.66666666666667,
+      "grad_norm": 116719.890625,
+      "learning_rate": 0.0004506295834481561,
+      "loss": 0.4435,
+      "step": 694
+    },
+    {
+      "epoch": 115.83333333333333,
+      "grad_norm": 128234.1640625,
+      "learning_rate": 0.00044918246090859906,
+      "loss": 0.4177,
+      "step": 695
+    },
+    {
+      "epoch": 116.0,
+      "grad_norm": 184487.40625,
+      "learning_rate": 0.00044773576836617336,
+      "loss": 0.4256,
+      "step": 696
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy": 0.797979797979798,
+      "eval_best_threshold": 0.4464789927005768,
+      "eval_f1": 0.8266219239373602,
+      "eval_loss": 0.42530736327171326,
+      "eval_pr_auc": 0.8979053279883333,
+      "eval_precision": 0.7870074547390842,
+      "eval_recall": 0.8704358068315665,
+      "eval_roc_auc": 0.882127168960952,
+      "eval_runtime": 3.8355,
+      "eval_samples_per_second": 800.146,
+      "eval_steps_per_second": 1.564,
+      "step": 696
+    },
+    {
+      "epoch": 116.16666666666667,
+      "grad_norm": 107760.5078125,
+      "learning_rate": 0.0004462895180621965,
+      "loss": 0.405,
+      "step": 697
+    },
+    {
+      "epoch": 116.33333333333333,
+      "grad_norm": 200850.09375,
+      "learning_rate": 0.00044484372223424414,
+      "loss": 0.4161,
+      "step": 698
+    },
+    {
+      "epoch": 116.5,
+      "grad_norm": 245987.421875,
+      "learning_rate": 0.00044339839311604667,
+      "loss": 0.4531,
+      "step": 699
+    },
+    {
+      "epoch": 116.66666666666667,
+      "grad_norm": 103315.1171875,
+      "learning_rate": 0.0004419535429373848,
+      "loss": 0.4008,
+      "step": 700
+    },
+    {
+      "epoch": 116.83333333333333,
+      "grad_norm": 121910.796875,
+      "learning_rate": 0.0004405091839239874,
+      "loss": 0.4229,
+      "step": 701
+    },
+    {
+      "epoch": 117.0,
+      "grad_norm": 148540.953125,
+      "learning_rate": 0.00043906532829742634,
+      "loss": 0.4003,
+      "step": 702
+    },
+    {
+      "epoch": 117.0,
+      "eval_accuracy": 0.7940697295536006,
+      "eval_best_threshold": 0.35628262162208557,
+      "eval_f1": 0.8281674823273518,
+      "eval_loss": 0.41951417922973633,
+      "eval_pr_auc": 0.9025554076344122,
+      "eval_precision": 0.7691919191919192,
+      "eval_recall": 0.8969375736160189,
+      "eval_roc_auc": 0.8862505251383402,
+      "eval_runtime": 3.8482,
+      "eval_samples_per_second": 797.513,
+      "eval_steps_per_second": 1.559,
+      "step": 702
+    },
+    {
+      "epoch": 117.16666666666667,
+      "grad_norm": 61846.95703125,
+      "learning_rate": 0.00043762198827501463,
+      "loss": 0.418,
+      "step": 703
+    },
+    {
+      "epoch": 117.33333333333333,
+      "grad_norm": 62604.359375,
+      "learning_rate": 0.0004361791760697027,
+      "loss": 0.4135,
+      "step": 704
+    },
+    {
+      "epoch": 117.5,
+      "grad_norm": 64781.78125,
+      "learning_rate": 0.00043473690388997434,
+      "loss": 0.4109,
+      "step": 705
+    },
+    {
+      "epoch": 117.66666666666667,
+      "grad_norm": 78247.5859375,
+      "learning_rate": 0.00043329518393974364,
+      "loss": 0.4041,
+      "step": 706
+    },
+    {
+      "epoch": 117.83333333333333,
+      "grad_norm": 120075.2890625,
+      "learning_rate": 0.00043185402841825285,
+      "loss": 0.3946,
+      "step": 707
+    },
+    {
+      "epoch": 118.0,
+      "grad_norm": 133093.421875,
+      "learning_rate": 0.0004304134495199674,
+      "loss": 0.3965,
+      "step": 708
+    },
+    {
+      "epoch": 118.0,
+      "eval_accuracy": 0.8041707396546106,
+      "eval_best_threshold": 0.3818773925304413,
+      "eval_f1": 0.830847171404447,
+      "eval_loss": 0.4185047745704651,
+      "eval_pr_auc": 0.9035732506643357,
+      "eval_precision": 0.7956873315363882,
+      "eval_recall": 0.8692579505300353,
+      "eval_roc_auc": 0.888738542533843,
+      "eval_runtime": 3.9051,
+      "eval_samples_per_second": 785.905,
+      "eval_steps_per_second": 1.536,
+      "step": 708
+    },
+    {
+      "epoch": 118.16666666666667,
+      "grad_norm": 106172.3125,
+      "learning_rate": 0.00042897345943447373,
+      "loss": 0.4074,
+      "step": 709
+    },
+    {
+      "epoch": 118.33333333333333,
+      "grad_norm": 66229.796875,
+      "learning_rate": 0.0004275340703463767,
+      "loss": 0.3781,
+      "step": 710
+    },
+    {
+      "epoch": 118.5,
+      "grad_norm": 49387.58984375,
+      "learning_rate": 0.0004260952944351947,
+      "loss": 0.3981,
+      "step": 711
+    },
+    {
+      "epoch": 118.66666666666667,
+      "grad_norm": 282733.625,
+      "learning_rate": 0.00042465714387525844,
+      "loss": 0.4425,
+      "step": 712
+    },
+    {
+      "epoch": 118.83333333333333,
+      "grad_norm": 188722.796875,
+      "learning_rate": 0.000423219630835607,
+      "loss": 0.4469,
+      "step": 713
+    },
+    {
+      "epoch": 119.0,
+      "grad_norm": 162376.21875,
+      "learning_rate": 0.0004217827674798845,
+      "loss": 0.4318,
+      "step": 714
+    },
+    {
+      "epoch": 119.0,
+      "eval_accuracy": 0.8106875203649397,
+      "eval_best_threshold": 0.5838644504547119,
+      "eval_f1": 0.8339525578736782,
+      "eval_loss": 0.4320233464241028,
+      "eval_pr_auc": 0.9027477013201601,
+      "eval_precision": 0.810105496946141,
+      "eval_recall": 0.85924617196702,
+      "eval_roc_auc": 0.8869300906631478,
+      "eval_runtime": 3.9503,
+      "eval_samples_per_second": 776.896,
+      "eval_steps_per_second": 1.519,
+      "step": 714
+    },
+    {
+      "epoch": 119.16666666666667,
+      "grad_norm": 147409.125,
+      "learning_rate": 0.00042034656596623876,
+      "loss": 0.4207,
+      "step": 715
+    },
+    {
+      "epoch": 119.33333333333333,
+      "grad_norm": 55680.04296875,
+      "learning_rate": 0.00041891103844721633,
+      "loss": 0.3979,
+      "step": 716
+    },
+    {
+      "epoch": 119.5,
+      "grad_norm": 169613.6875,
+      "learning_rate": 0.00041747619706966117,
+      "loss": 0.4194,
+      "step": 717
+    },
+    {
+      "epoch": 119.66666666666667,
+      "grad_norm": 159846.125,
+      "learning_rate": 0.0004160420539746115,
+      "loss": 0.4207,
+      "step": 718
+    },
+    {
+      "epoch": 119.83333333333333,
+      "grad_norm": 62629.86328125,
+      "learning_rate": 0.00041460862129719675,
+      "loss": 0.4122,
+      "step": 719
+    },
+    {
+      "epoch": 120.0,
+      "grad_norm": 70529.6953125,
+      "learning_rate": 0.00041317591116653486,
+      "loss": 0.3997,
+      "step": 720
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy": 0.8071032909742587,
+      "eval_best_threshold": 0.43030133843421936,
+      "eval_f1": 0.8332394366197183,
+      "eval_loss": 0.41646894812583923,
+      "eval_pr_auc": 0.9015496978574711,
+      "eval_precision": 0.7985961123110151,
+      "eval_recall": 0.8710247349823321,
+      "eval_roc_auc": 0.8878643858695044,
+      "eval_runtime": 3.9469,
+      "eval_samples_per_second": 777.573,
+      "eval_steps_per_second": 1.52,
+      "step": 720
+    },
+    {
+      "epoch": 120.16666666666667,
+      "grad_norm": 65910.9921875,
+      "learning_rate": 0.0004117439357056305,
+      "loss": 0.3941,
+      "step": 721
+    },
+    {
+      "epoch": 120.33333333333333,
+      "grad_norm": 76004.28125,
+      "learning_rate": 0.0004103127070312713,
+      "loss": 0.4048,
+      "step": 722
+    },
+    {
+      "epoch": 120.5,
+      "grad_norm": 169147.484375,
+      "learning_rate": 0.00040888223725392626,
+      "loss": 0.4036,
+      "step": 723
+    },
+    {
+      "epoch": 120.66666666666667,
+      "grad_norm": 118840.078125,
+      "learning_rate": 0.0004074525384776428,
+      "loss": 0.4056,
+      "step": 724
+    },
+    {
+      "epoch": 120.83333333333333,
+      "grad_norm": 267710.28125,
+      "learning_rate": 0.0004060236227999441,
+      "loss": 0.4361,
+      "step": 725
+    },
+    {
+      "epoch": 121.0,
+      "grad_norm": 303997.96875,
+      "learning_rate": 0.0004045955023117276,
+      "loss": 0.432,
+      "step": 726
+    },
+    {
+      "epoch": 121.0,
+      "eval_accuracy": 0.7908113391984359,
+      "eval_best_threshold": 0.33801013231277466,
+      "eval_f1": 0.8248772504091653,
+      "eval_loss": 0.4277116358280182,
+      "eval_pr_auc": 0.9012340785625325,
+      "eval_precision": 0.7682926829268293,
+      "eval_recall": 0.8904593639575972,
+      "eval_roc_auc": 0.8837801197444284,
+      "eval_runtime": 3.8194,
+      "eval_samples_per_second": 803.521,
+      "eval_steps_per_second": 1.571,
+      "step": 726
+    },
+    {
+      "epoch": 121.16666666666667,
+      "grad_norm": 138695.34375,
+      "learning_rate": 0.00040316818909716206,
+      "loss": 0.4303,
+      "step": 727
+    },
+    {
+      "epoch": 121.33333333333333,
+      "grad_norm": 116743.2265625,
+      "learning_rate": 0.0004017416952335849,
+      "loss": 0.3981,
+      "step": 728
+    },
+    {
+      "epoch": 121.5,
+      "grad_norm": 90517.8828125,
+      "learning_rate": 0.0004003160327914015,
+      "loss": 0.392,
+      "step": 729
+    },
+    {
+      "epoch": 121.66666666666667,
+      "grad_norm": 116381.640625,
+      "learning_rate": 0.00039889121383398113,
+      "loss": 0.4085,
+      "step": 730
+    },
+    {
+      "epoch": 121.83333333333333,
+      "grad_norm": 98319.7421875,
+      "learning_rate": 0.00039746725041755625,
+      "loss": 0.4026,
+      "step": 731
+    },
+    {
+      "epoch": 122.0,
+      "grad_norm": 110849.65625,
+      "learning_rate": 0.0003960441545911204,
+      "loss": 0.4035,
+      "step": 732
+    },
+    {
+      "epoch": 122.0,
+      "eval_accuracy": 0.8002606712284132,
+      "eval_best_threshold": 0.4917587339878082,
+      "eval_f1": 0.8283394007280873,
+      "eval_loss": 0.4172866940498352,
+      "eval_pr_auc": 0.9087252692436502,
+      "eval_precision": 0.789642285104111,
+      "eval_recall": 0.8710247349823321,
+      "eval_roc_auc": 0.8915839546933406,
+      "eval_runtime": 3.8054,
+      "eval_samples_per_second": 806.475,
+      "eval_steps_per_second": 1.577,
+      "step": 732
+    },
+    {
+      "epoch": 122.16666666666667,
+      "grad_norm": 174083.859375,
+      "learning_rate": 0.0003946219383963254,
+      "loss": 0.4315,
+      "step": 733
+    },
+    {
+      "epoch": 122.33333333333333,
+      "grad_norm": 55083.375,
+      "learning_rate": 0.00039320061386738007,
+      "loss": 0.3932,
+      "step": 734
+    },
+    {
+      "epoch": 122.5,
+      "grad_norm": 136627.65625,
+      "learning_rate": 0.0003917801930309486,
+      "loss": 0.3989,
+      "step": 735
+    },
+    {
+      "epoch": 122.66666666666667,
+      "grad_norm": 64387.43359375,
+      "learning_rate": 0.0003903606879060483,
+      "loss": 0.4018,
+      "step": 736
+    },
+    {
+      "epoch": 122.83333333333333,
+      "grad_norm": 127286.59375,
+      "learning_rate": 0.00038894211050394847,
+      "loss": 0.4137,
+      "step": 737
+    },
+    {
+      "epoch": 123.0,
+      "grad_norm": 135965.734375,
+      "learning_rate": 0.0003875244728280676,
+      "loss": 0.3932,
+      "step": 738
+    },
+    {
+      "epoch": 123.0,
+      "eval_accuracy": 0.8084066471163245,
+      "eval_best_threshold": 0.43941736221313477,
+      "eval_f1": 0.8357541899441341,
+      "eval_loss": 0.41049131751060486,
+      "eval_pr_auc": 0.9031017937895613,
+      "eval_precision": 0.7948990435706695,
+      "eval_recall": 0.8810365135453475,
+      "eval_roc_auc": 0.8904202739052852,
+      "eval_runtime": 3.958,
+      "eval_samples_per_second": 775.396,
+      "eval_steps_per_second": 1.516,
+      "step": 738
+    },
+    {
+      "epoch": 123.16666666666667,
+      "grad_norm": 83984.4375,
+      "learning_rate": 0.00038610778687387325,
+      "loss": 0.4004,
+      "step": 739
+    },
+    {
+      "epoch": 123.33333333333333,
+      "grad_norm": 118075.1640625,
+      "learning_rate": 0.00038469206462877993,
+      "loss": 0.4059,
+      "step": 740
+    },
+    {
+      "epoch": 123.5,
+      "grad_norm": 69527.3828125,
+      "learning_rate": 0.00038327731807204744,
+      "loss": 0.3804,
+      "step": 741
+    },
+    {
+      "epoch": 123.66666666666667,
+      "grad_norm": 115279.078125,
+      "learning_rate": 0.0003818635591746793,
+      "loss": 0.3858,
+      "step": 742
+    },
+    {
+      "epoch": 123.83333333333333,
+      "grad_norm": 90124.2890625,
+      "learning_rate": 0.0003804507998993229,
+      "loss": 0.418,
+      "step": 743
+    },
+    {
+      "epoch": 124.0,
+      "grad_norm": 197003.6875,
+      "learning_rate": 0.0003790390522001662,
+      "loss": 0.4173,
+      "step": 744
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy": 0.8012381883349625,
+      "eval_best_threshold": 0.2784575819969177,
+      "eval_f1": 0.8348673524634542,
+      "eval_loss": 0.43013083934783936,
+      "eval_pr_auc": 0.9060761496011254,
+      "eval_precision": 0.7725450901803608,
+      "eval_recall": 0.9081272084805654,
+      "eval_roc_auc": 0.8903390868735606,
+      "eval_runtime": 3.8243,
+      "eval_samples_per_second": 802.5,
+      "eval_steps_per_second": 1.569,
+      "step": 744
+    },
+    {
+      "epoch": 124.16666666666667,
+      "grad_norm": 315667.375,
+      "learning_rate": 0.0003776283280228381,
+      "loss": 0.4234,
+      "step": 745
+    },
+    {
+      "epoch": 124.33333333333333,
+      "grad_norm": 58933.265625,
+      "learning_rate": 0.00037621863930430713,
+      "loss": 0.3804,
+      "step": 746
+    },
+    {
+      "epoch": 124.5,
+      "grad_norm": 193578.453125,
+      "learning_rate": 0.0003748099979727792,
+      "loss": 0.3968,
+      "step": 747
+    },
+    {
+      "epoch": 124.66666666666667,
+      "grad_norm": 213663.515625,
+      "learning_rate": 0.00037340241594759917,
+      "loss": 0.4235,
+      "step": 748
+    },
+    {
+      "epoch": 124.83333333333333,
+      "grad_norm": 92057.046875,
+      "learning_rate": 0.00037199590513914714,
+      "loss": 0.3882,
+      "step": 749
+    },
+    {
+      "epoch": 125.0,
+      "grad_norm": 58977.01171875,
+      "learning_rate": 0.0003705904774487396,
+      "loss": 0.3975,
+      "step": 750
+    },
+    {
+      "epoch": 125.0,
+      "eval_accuracy": 0.8113391984359726,
+      "eval_best_threshold": 0.42779427766799927,
+      "eval_f1": 0.8374052232518955,
+      "eval_loss": 0.39996033906936646,
+      "eval_pr_auc": 0.9114790385715704,
+      "eval_precision": 0.8003220611916264,
+      "eval_recall": 0.8780918727915195,
+      "eval_roc_auc": 0.8961544838867368,
+      "eval_runtime": 3.8438,
+      "eval_samples_per_second": 798.422,
+      "eval_steps_per_second": 1.561,
+      "step": 750
+    },
+    {
+      "epoch": 125.16666666666667,
+      "grad_norm": 56553.06640625,
+      "learning_rate": 0.0003691861447685283,
+      "loss": 0.4024,
+      "step": 751
+    },
+    {
+      "epoch": 125.33333333333333,
+      "grad_norm": 85294.8671875,
+      "learning_rate": 0.00036778291898139905,
+      "loss": 0.4007,
+      "step": 752
+    },
+    {
+      "epoch": 125.5,
+      "grad_norm": 55662.74609375,
+      "learning_rate": 0.0003663808119608716,
+      "loss": 0.3915,
+      "step": 753
+    },
+    {
+      "epoch": 125.66666666666667,
+      "grad_norm": 59557.08984375,
+      "learning_rate": 0.0003649798355709997,
+      "loss": 0.3726,
+      "step": 754
+    },
+    {
+      "epoch": 125.83333333333333,
+      "grad_norm": 185125.28125,
+      "learning_rate": 0.0003635800016662696,
+      "loss": 0.4103,
+      "step": 755
+    },
+    {
+      "epoch": 126.0,
+      "grad_norm": 95126.6640625,
+      "learning_rate": 0.00036218132209150044,
+      "loss": 0.3871,
+      "step": 756
+    },
+    {
+      "epoch": 126.0,
+      "eval_accuracy": 0.8074291300097751,
+      "eval_best_threshold": 0.39693698287010193,
+      "eval_f1": 0.8332863187588152,
+      "eval_loss": 0.4095918834209442,
+      "eval_pr_auc": 0.9093813330693225,
+      "eval_precision": 0.7996751488900921,
+      "eval_recall": 0.8698468786808009,
+      "eval_roc_auc": 0.8925401575114328,
+      "eval_runtime": 3.6274,
+      "eval_samples_per_second": 846.051,
+      "eval_steps_per_second": 1.654,
+      "step": 756
+    },
+    {
+      "epoch": 126.16666666666667,
+      "grad_norm": 111067.421875,
+      "learning_rate": 0.00036078380868174423,
+      "loss": 0.4079,
+      "step": 757
+    },
+    {
+      "epoch": 126.33333333333333,
+      "grad_norm": 161514.3125,
+      "learning_rate": 0.00035938747326218467,
+      "loss": 0.4236,
+      "step": 758
+    },
+    {
+      "epoch": 126.5,
+      "grad_norm": 105326.0703125,
+      "learning_rate": 0.0003579923276480387,
+      "loss": 0.3951,
+      "step": 759
+    },
+    {
+      "epoch": 126.66666666666667,
+      "grad_norm": 153800.453125,
+      "learning_rate": 0.00035659838364445503,
+      "loss": 0.3889,
+      "step": 760
+    },
+    {
+      "epoch": 126.83333333333333,
+      "grad_norm": 164338.90625,
+      "learning_rate": 0.0003552056530464146,
+      "loss": 0.4136,
+      "step": 761
+    },
+    {
+      "epoch": 127.0,
+      "grad_norm": 68888.59375,
+      "learning_rate": 0.00035381414763863166,
+      "loss": 0.4016,
+      "step": 762
+    },
+    {
+      "epoch": 127.0,
+      "eval_accuracy": 0.8080808080808081,
+      "eval_best_threshold": 0.3385215401649475,
+      "eval_f1": 0.8349677780891006,
+      "eval_loss": 0.4206610918045044,
+      "eval_pr_auc": 0.9069561308305243,
+      "eval_precision": 0.7963655799037948,
+      "eval_recall": 0.8775029446407538,
+      "eval_roc_auc": 0.8931857877161014,
+      "eval_runtime": 3.6927,
+      "eval_samples_per_second": 831.105,
+      "eval_steps_per_second": 1.625,
+      "step": 762
+    },
+    {
+      "epoch": 127.16666666666667,
+      "grad_norm": 190998.125,
+      "learning_rate": 0.00035242387919545313,
+      "loss": 0.3974,
+      "step": 763
+    },
+    {
+      "epoch": 127.33333333333333,
+      "grad_norm": 243973.671875,
+      "learning_rate": 0.000351034859480759,
+      "loss": 0.3905,
+      "step": 764
+    },
+    {
+      "epoch": 127.5,
+      "grad_norm": 100218.6484375,
+      "learning_rate": 0.0003496471002478635,
+      "loss": 0.4081,
+      "step": 765
+    },
+    {
+      "epoch": 127.66666666666667,
+      "grad_norm": 171532.484375,
+      "learning_rate": 0.00034826061323941484,
+      "loss": 0.3994,
+      "step": 766
+    },
+    {
+      "epoch": 127.83333333333333,
+      "grad_norm": 133002.640625,
+      "learning_rate": 0.00034687541018729604,
+      "loss": 0.4055,
+      "step": 767
+    },
+    {
+      "epoch": 128.0,
+      "grad_norm": 74934.3359375,
+      "learning_rate": 0.00034549150281252633,
+      "loss": 0.4062,
+      "step": 768
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy": 0.8028673835125448,
+      "eval_best_threshold": 0.3128413259983063,
+      "eval_f1": 0.8354636932281752,
+      "eval_loss": 0.4116682708263397,
+      "eval_pr_auc": 0.9072100448142456,
+      "eval_precision": 0.7761495704901465,
+      "eval_recall": 0.9045936395759717,
+      "eval_roc_auc": 0.8940595148194255,
+      "eval_runtime": 3.6699,
+      "eval_samples_per_second": 836.257,
+      "eval_steps_per_second": 1.635,
+      "step": 768
+    },
+    {
+      "epoch": 128.16666666666666,
+      "grad_norm": 187347.5625,
+      "learning_rate": 0.000344108902825161,
+      "loss": 0.416,
+      "step": 769
+    },
+    {
+      "epoch": 128.33333333333334,
+      "grad_norm": 81035.7890625,
+      "learning_rate": 0.00034272762192419327,
+      "loss": 0.3729,
+      "step": 770
+    },
+    {
+      "epoch": 128.5,
+      "grad_norm": 162363.390625,
+      "learning_rate": 0.000341347671797454,
+      "loss": 0.4092,
+      "step": 771
+    },
+    {
+      "epoch": 128.66666666666666,
+      "grad_norm": 112333.1328125,
+      "learning_rate": 0.00033996906412151417,
+      "loss": 0.398,
+      "step": 772
+    },
+    {
+      "epoch": 128.83333333333334,
+      "grad_norm": 106592.90625,
+      "learning_rate": 0.00033859181056158564,
+      "loss": 0.4205,
+      "step": 773
+    },
+    {
+      "epoch": 129.0,
+      "grad_norm": 107416.6640625,
+      "learning_rate": 0.00033721592277142175,
+      "loss": 0.4021,
+      "step": 774
+    },
+    {
+      "epoch": 129.0,
+      "eval_accuracy": 0.8123167155425219,
+      "eval_best_threshold": 0.40762990713119507,
+      "eval_f1": 0.8406198118428334,
+      "eval_loss": 0.39988085627555847,
+      "eval_pr_auc": 0.9105294646161908,
+      "eval_precision": 0.7927974947807933,
+      "eval_recall": 0.8945818610129564,
+      "eval_roc_auc": 0.8966293635881746,
+      "eval_runtime": 3.5481,
+      "eval_samples_per_second": 864.965,
+      "eval_steps_per_second": 1.691,
+      "step": 774
+    },
+    {
+      "epoch": 129.16666666666666,
+      "grad_norm": 41239.49609375,
+      "learning_rate": 0.00033584141239321953,
+      "loss": 0.3802,
+      "step": 775
+    },
+    {
+      "epoch": 129.33333333333334,
+      "grad_norm": 85460.1328125,
+      "learning_rate": 0.000334468291057521,
+      "loss": 0.401,
+      "step": 776
+    },
+    {
+      "epoch": 129.5,
+      "grad_norm": 89357.3984375,
+      "learning_rate": 0.00033309657038311456,
+      "loss": 0.4056,
+      "step": 777
+    },
+    {
+      "epoch": 129.66666666666666,
+      "grad_norm": 111458.8046875,
+      "learning_rate": 0.0003317262619769368,
+      "loss": 0.3956,
+      "step": 778
+    },
+    {
+      "epoch": 129.83333333333334,
+      "grad_norm": 109909.875,
+      "learning_rate": 0.00033035737743397446,
+      "loss": 0.4093,
+      "step": 779
+    },
+    {
+      "epoch": 130.0,
+      "grad_norm": 50348.46484375,
+      "learning_rate": 0.0003289899283371657,
+      "loss": 0.3804,
+      "step": 780
+    },
+    {
+      "epoch": 130.0,
+      "eval_accuracy": 0.8113391984359726,
+      "eval_best_threshold": 0.48834142088890076,
+      "eval_f1": 0.8335728657660247,
+      "eval_loss": 0.4086468517780304,
+      "eval_pr_auc": 0.9062419900377866,
+      "eval_precision": 0.8141493542953397,
+      "eval_recall": 0.8539458186101295,
+      "eval_roc_auc": 0.8933133673373831,
+      "eval_runtime": 3.6984,
+      "eval_samples_per_second": 829.81,
+      "eval_steps_per_second": 1.622,
+      "step": 780
+    },
+    {
+      "epoch": 130.16666666666666,
+      "grad_norm": 61747.39453125,
+      "learning_rate": 0.00032762392625730293,
+      "loss": 0.3776,
+      "step": 781
+    },
+    {
+      "epoch": 130.33333333333334,
+      "grad_norm": 81545.625,
+      "learning_rate": 0.0003262593827529343,
+      "loss": 0.3911,
+      "step": 782
+    },
+    {
+      "epoch": 130.5,
+      "grad_norm": 50101.1015625,
+      "learning_rate": 0.00032489630937026625,
+      "loss": 0.3805,
+      "step": 783
+    },
+    {
+      "epoch": 130.66666666666666,
+      "grad_norm": 47948.2265625,
+      "learning_rate": 0.00032353471764306563,
+      "loss": 0.3778,
+      "step": 784
+    },
+    {
+      "epoch": 130.83333333333334,
+      "grad_norm": 70858.6015625,
+      "learning_rate": 0.0003221746190925618,
+      "loss": 0.4055,
+      "step": 785
+    },
+    {
+      "epoch": 131.0,
+      "grad_norm": 65349.390625,
+      "learning_rate": 0.00032081602522734986,
+      "loss": 0.3874,
+      "step": 786
+    },
+    {
+      "epoch": 131.0,
+      "eval_accuracy": 0.8165526230042359,
+      "eval_best_threshold": 0.3926054537296295,
+      "eval_f1": 0.8408255583828103,
+      "eval_loss": 0.40172040462493896,
+      "eval_pr_auc": 0.9120142027521659,
+      "eval_precision": 0.8085916258836324,
+      "eval_recall": 0.875736160188457,
+      "eval_roc_auc": 0.8970475412357095,
+      "eval_runtime": 3.7845,
+      "eval_samples_per_second": 810.935,
+      "eval_steps_per_second": 1.585,
+      "step": 786
+    },
+    {
+      "epoch": 131.16666666666666,
+      "grad_norm": 118622.21875,
+      "learning_rate": 0.00031945894754329286,
+      "loss": 0.3902,
+      "step": 787
+    },
+    {
+      "epoch": 131.33333333333334,
+      "grad_norm": 109528.3515625,
+      "learning_rate": 0.0003181033975234244,
+      "loss": 0.4079,
+      "step": 788
+    },
+    {
+      "epoch": 131.5,
+      "grad_norm": 78493.7265625,
+      "learning_rate": 0.0003167493866378514,
+      "loss": 0.3847,
+      "step": 789
+    },
+    {
+      "epoch": 131.66666666666666,
+      "grad_norm": 198052.96875,
+      "learning_rate": 0.00031539692634365783,
+      "loss": 0.3853,
+      "step": 790
+    },
+    {
+      "epoch": 131.83333333333334,
+      "grad_norm": 139994.390625,
+      "learning_rate": 0.0003140460280848066,
+      "loss": 0.379,
+      "step": 791
+    },
+    {
+      "epoch": 132.0,
+      "grad_norm": 99828.71875,
+      "learning_rate": 0.00031269670329204396,
+      "loss": 0.3816,
+      "step": 792
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy": 0.8211143695014663,
+      "eval_best_threshold": 0.5147361159324646,
+      "eval_f1": 0.8431876606683805,
+      "eval_loss": 0.39976704120635986,
+      "eval_pr_auc": 0.9126112501936106,
+      "eval_precision": 0.8186356073211315,
+      "eval_recall": 0.8692579505300353,
+      "eval_roc_auc": 0.8986343396229659,
+      "eval_runtime": 3.6082,
+      "eval_samples_per_second": 850.569,
+      "eval_steps_per_second": 1.663,
+      "step": 792
+    },
+    {
+      "epoch": 132.16666666666666,
+      "grad_norm": 137741.453125,
+      "learning_rate": 0.0003113489633828016,
+      "loss": 0.3943,
+      "step": 793
+    },
+    {
+      "epoch": 132.33333333333334,
+      "grad_norm": 64978.36328125,
+      "learning_rate": 0.00031000281976110056,
+      "loss": 0.3808,
+      "step": 794
+    },
+    {
+      "epoch": 132.5,
+      "grad_norm": 253367.953125,
+      "learning_rate": 0.0003086582838174551,
+      "loss": 0.3871,
+      "step": 795
+    },
+    {
+      "epoch": 132.66666666666666,
+      "grad_norm": 175149.796875,
+      "learning_rate": 0.00030731536692877595,
+      "loss": 0.3883,
+      "step": 796
+    },
+    {
+      "epoch": 132.83333333333334,
+      "grad_norm": 62742.25,
+      "learning_rate": 0.0003059740804582736,
+      "loss": 0.3913,
+      "step": 797
+    },
+    {
+      "epoch": 133.0,
+      "grad_norm": 69216.578125,
+      "learning_rate": 0.0003046344357553632,
+      "loss": 0.3498,
+      "step": 798
+    },
+    {
+      "epoch": 133.0,
+      "eval_accuracy": 0.8145975887911372,
+      "eval_best_threshold": 0.3728158175945282,
+      "eval_f1": 0.8408391608391609,
+      "eval_loss": 0.3914491534233093,
+      "eval_pr_auc": 0.915878808081855,
+      "eval_precision": 0.8007458710708577,
+      "eval_recall": 0.8851590106007067,
+      "eval_roc_auc": 0.9024999591917036,
+      "eval_runtime": 3.6398,
+      "eval_samples_per_second": 843.171,
+      "eval_steps_per_second": 1.648,
+      "step": 798
+    },
+    {
+      "epoch": 133.16666666666666,
+      "grad_norm": 78250.9921875,
+      "learning_rate": 0.00030329644415556757,
+      "loss": 0.3738,
+      "step": 799
+    },
+    {
+      "epoch": 133.33333333333334,
+      "grad_norm": 51423.38671875,
+      "learning_rate": 0.0003019601169804216,
+      "loss": 0.3597,
+      "step": 800
+    },
+    {
+      "epoch": 133.5,
+      "grad_norm": 128579.515625,
+      "learning_rate": 0.00030062546553737693,
+      "loss": 0.3987,
+      "step": 801
+    },
+    {
+      "epoch": 133.66666666666666,
+      "grad_norm": 49651.3046875,
+      "learning_rate": 0.0002992925011197053,
+      "loss": 0.377,
+      "step": 802
+    },
+    {
+      "epoch": 133.83333333333334,
+      "grad_norm": 95432.3671875,
+      "learning_rate": 0.0002979612350064042,
+      "loss": 0.3948,
+      "step": 803
+    },
+    {
+      "epoch": 134.0,
+      "grad_norm": 179643.421875,
+      "learning_rate": 0.0002966316784621,
+      "loss": 0.3748,
+      "step": 804
+    },
+    {
+      "epoch": 134.0,
+      "eval_accuracy": 0.820136852394917,
+      "eval_best_threshold": 0.6017637848854065,
+      "eval_f1": 0.8398142774230992,
+      "eval_loss": 0.40153124928474426,
+      "eval_pr_auc": 0.9159341444123951,
+      "eval_precision": 0.8278032036613272,
+      "eval_recall": 0.8521790341578327,
+      "eval_roc_auc": 0.9012782876667019,
+      "eval_runtime": 3.6233,
+      "eval_samples_per_second": 847.024,
+      "eval_steps_per_second": 1.656,
+      "step": 804
+    },
+    {
+      "epoch": 134.16666666666666,
+      "grad_norm": 152972.578125,
+      "learning_rate": 0.0002953038427369537,
+      "loss": 0.3877,
+      "step": 805
+    },
+    {
+      "epoch": 134.33333333333334,
+      "grad_norm": 75853.0078125,
+      "learning_rate": 0.0002939777390665658,
+      "loss": 0.3917,
+      "step": 806
+    },
+    {
+      "epoch": 134.5,
+      "grad_norm": 87845.7734375,
+      "learning_rate": 0.00029265337867188056,
+      "loss": 0.3859,
+      "step": 807
+    },
+    {
+      "epoch": 134.66666666666666,
+      "grad_norm": 91740.2421875,
+      "learning_rate": 0.0002913307727590911,
+      "loss": 0.366,
+      "step": 808
+    },
+    {
+      "epoch": 134.83333333333334,
+      "grad_norm": 178002.421875,
+      "learning_rate": 0.0002900099325195453,
+      "loss": 0.3757,
+      "step": 809
+    },
+    {
+      "epoch": 135.0,
+      "grad_norm": 134209.953125,
+      "learning_rate": 0.0002886908691296504,
+      "loss": 0.3992,
+      "step": 810
+    },
+    {
+      "epoch": 135.0,
+      "eval_accuracy": 0.8243727598566308,
+      "eval_best_threshold": 0.3863644599914551,
+      "eval_f1": 0.8455144740613356,
+      "eval_loss": 0.39555782079696655,
+      "eval_pr_auc": 0.9174050733417477,
+      "eval_precision": 0.8235622557230597,
+      "eval_recall": 0.8686690223792697,
+      "eval_roc_auc": 0.902185090968136,
+      "eval_runtime": 3.6528,
+      "eval_samples_per_second": 840.174,
+      "eval_steps_per_second": 1.643,
+      "step": 810
+    },
+    {
+      "epoch": 135.16666666666666,
+      "grad_norm": 109344.46875,
+      "learning_rate": 0.0002873735937507786,
+      "loss": 0.3929,
+      "step": 811
+    },
+    {
+      "epoch": 135.33333333333334,
+      "grad_norm": 193133.703125,
+      "learning_rate": 0.000286058117529173,
+      "loss": 0.3721,
+      "step": 812
+    },
+    {
+      "epoch": 135.5,
+      "grad_norm": 97225.890625,
+      "learning_rate": 0.0002847444515958523,
+      "loss": 0.3845,
+      "step": 813
+    },
+    {
+      "epoch": 135.66666666666666,
+      "grad_norm": 65620.78125,
+      "learning_rate": 0.00028343260706651866,
+      "loss": 0.3739,
+      "step": 814
+    },
+    {
+      "epoch": 135.83333333333334,
+      "grad_norm": 121686.9609375,
+      "learning_rate": 0.00028212259504146043,
+      "loss": 0.3798,
+      "step": 815
+    },
+    {
+      "epoch": 136.0,
+      "grad_norm": 145372.125,
+      "learning_rate": 0.00028081442660546124,
+      "loss": 0.4034,
+      "step": 816
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy": 0.8139459107201042,
+      "eval_best_threshold": 0.434948205947876,
+      "eval_f1": 0.8400112076211824,
+      "eval_loss": 0.3963012099266052,
+      "eval_pr_auc": 0.9114395336778187,
+      "eval_precision": 0.8011758417958311,
+      "eval_recall": 0.8828032979976443,
+      "eval_roc_auc": 0.8988340854946696,
+      "eval_runtime": 3.8337,
+      "eval_samples_per_second": 800.528,
+      "eval_steps_per_second": 1.565,
+      "step": 816
+    },
+    {
+      "epoch": 136.16666666666666,
+      "grad_norm": 69139.9921875,
+      "learning_rate": 0.00027950811282770506,
+      "loss": 0.363,
+      "step": 817
+    },
+    {
+      "epoch": 136.33333333333334,
+      "grad_norm": 87535.4453125,
+      "learning_rate": 0.00027820366476168226,
+      "loss": 0.3627,
+      "step": 818
+    },
+    {
+      "epoch": 136.5,
+      "grad_norm": 64158.21484375,
+      "learning_rate": 0.0002769010934450956,
+      "loss": 0.3796,
+      "step": 819
+    },
+    {
+      "epoch": 136.66666666666666,
+      "grad_norm": 134014.46875,
+      "learning_rate": 0.00027560040989976894,
+      "loss": 0.4012,
+      "step": 820
+    },
+    {
+      "epoch": 136.83333333333334,
+      "grad_norm": 122497.1953125,
+      "learning_rate": 0.0002743016251315517,
+      "loss": 0.3701,
+      "step": 821
+    },
+    {
+      "epoch": 137.0,
+      "grad_norm": 41029.91796875,
+      "learning_rate": 0.00027300475013022663,
+      "loss": 0.3771,
+      "step": 822
+    },
+    {
+      "epoch": 137.0,
+      "eval_accuracy": 0.8224177256435321,
+      "eval_best_threshold": 0.4527401924133301,
+      "eval_f1": 0.8470390120684816,
+      "eval_loss": 0.3848971724510193,
+      "eval_pr_auc": 0.918773658242021,
+      "eval_precision": 0.8091152815013405,
+      "eval_recall": 0.8886925795053003,
+      "eval_roc_auc": 0.9052160734858619,
+      "eval_runtime": 3.7139,
+      "eval_samples_per_second": 826.362,
+      "eval_steps_per_second": 1.616,
+      "step": 822
+    },
+    {
+      "epoch": 137.16666666666666,
+      "grad_norm": 78552.4296875,
+      "learning_rate": 0.00027170979586941727,
+      "loss": 0.3637,
+      "step": 823
+    },
+    {
+      "epoch": 137.33333333333334,
+      "grad_norm": 83702.828125,
+      "learning_rate": 0.0002704167733064941,
+      "loss": 0.3808,
+      "step": 824
+    },
+    {
+      "epoch": 137.5,
+      "grad_norm": 101689.9453125,
+      "learning_rate": 0.00026912569338248315,
+      "loss": 0.3703,
+      "step": 825
+    },
+    {
+      "epoch": 137.66666666666666,
+      "grad_norm": 108093.1015625,
+      "learning_rate": 0.00026783656702197156,
+      "loss": 0.3886,
+      "step": 826
+    },
+    {
+      "epoch": 137.83333333333334,
+      "grad_norm": 85036.0546875,
+      "learning_rate": 0.00026654940513301696,
+      "loss": 0.3702,
+      "step": 827
+    },
+    {
+      "epoch": 138.0,
+      "grad_norm": 92898.703125,
+      "learning_rate": 0.00026526421860705474,
+      "loss": 0.3831,
+      "step": 828
+    },
+    {
+      "epoch": 138.0,
+      "eval_accuracy": 0.8233952427500815,
+      "eval_best_threshold": 0.36562278866767883,
+      "eval_f1": 0.848434004474273,
+      "eval_loss": 0.38462093472480774,
+      "eval_pr_auc": 0.9204799802032648,
+      "eval_precision": 0.8077742279020235,
+      "eval_recall": 0.8934040047114252,
+      "eval_roc_auc": 0.9060363202428912,
+      "eval_runtime": 3.6435,
+      "eval_samples_per_second": 842.317,
+      "eval_steps_per_second": 1.647,
+      "step": 828
+    },
+    {
+      "epoch": 138.16666666666666,
+      "grad_norm": 105777.234375,
+      "learning_rate": 0.0002639810183188045,
+      "loss": 0.3695,
+      "step": 829
+    },
+    {
+      "epoch": 138.33333333333334,
+      "grad_norm": 72163.3515625,
+      "learning_rate": 0.0002626998151261798,
+      "loss": 0.3792,
+      "step": 830
+    },
+    {
+      "epoch": 138.5,
+      "grad_norm": 78065.9921875,
+      "learning_rate": 0.00026142061987019576,
+      "loss": 0.3676,
+      "step": 831
+    },
+    {
+      "epoch": 138.66666666666666,
+      "grad_norm": 149816.9375,
+      "learning_rate": 0.0002601434433748771,
+      "loss": 0.3829,
+      "step": 832
+    },
+    {
+      "epoch": 138.83333333333334,
+      "grad_norm": 99635.90625,
+      "learning_rate": 0.0002588682964471657,
+      "loss": 0.361,
+      "step": 833
+    },
+    {
+      "epoch": 139.0,
+      "grad_norm": 153759.828125,
+      "learning_rate": 0.0002575951898768315,
+      "loss": 0.3732,
+      "step": 834
+    },
+    {
+      "epoch": 139.0,
+      "eval_accuracy": 0.826653633105246,
+      "eval_best_threshold": 0.5249126553535461,
+      "eval_f1": 0.8463316002310803,
+      "eval_loss": 0.3846147060394287,
+      "eval_pr_auc": 0.920402157746111,
+      "eval_precision": 0.8304988662131519,
+      "eval_recall": 0.8627797408716137,
+      "eval_roc_auc": 0.9073454074343266,
+      "eval_runtime": 3.6743,
+      "eval_samples_per_second": 835.272,
+      "eval_steps_per_second": 1.633,
+      "step": 834
+    },
+    {
+      "epoch": 139.16666666666666,
+      "grad_norm": 115450.0859375,
+      "learning_rate": 0.00025632413443637885,
+      "loss": 0.3688,
+      "step": 835
+    },
+    {
+      "epoch": 139.33333333333334,
+      "grad_norm": 108277.234375,
+      "learning_rate": 0.00025505514088095655,
+      "loss": 0.3803,
+      "step": 836
+    },
+    {
+      "epoch": 139.5,
+      "grad_norm": 167415.609375,
+      "learning_rate": 0.0002537882199482665,
+      "loss": 0.3723,
+      "step": 837
+    },
+    {
+      "epoch": 139.66666666666666,
+      "grad_norm": 69005.0859375,
+      "learning_rate": 0.000252523382358473,
+      "loss": 0.3635,
+      "step": 838
+    },
+    {
+      "epoch": 139.83333333333334,
+      "grad_norm": 161154.0,
+      "learning_rate": 0.00025126063881411185,
+      "loss": 0.4026,
+      "step": 839
+    },
+    {
+      "epoch": 140.0,
+      "grad_norm": 224129.625,
+      "learning_rate": 0.0002500000000000001,
+      "loss": 0.3813,
+      "step": 840
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy": 0.8207885304659498,
+      "eval_best_threshold": 0.43965643644332886,
+      "eval_f1": 0.8441043083900227,
+      "eval_loss": 0.3839028477668762,
+      "eval_pr_auc": 0.9200467798517515,
+      "eval_precision": 0.8136612021857923,
+      "eval_recall": 0.8769140164899882,
+      "eval_roc_auc": 0.9065846548777942,
+      "eval_runtime": 3.8677,
+      "eval_samples_per_second": 793.494,
+      "eval_steps_per_second": 1.551,
+      "step": 840
+    },
+    {
+      "epoch": 140.16666666666666,
+      "grad_norm": 138304.515625,
+      "learning_rate": 0.000248741476583145,
+      "loss": 0.3825,
+      "step": 841
+    },
+    {
+      "epoch": 140.33333333333334,
+      "grad_norm": 193517.5625,
+      "learning_rate": 0.00024748507921265513,
+      "loss": 0.3782,
+      "step": 842
+    },
+    {
+      "epoch": 140.5,
+      "grad_norm": 210903.09375,
+      "learning_rate": 0.00024623081851964806,
+      "loss": 0.4004,
+      "step": 843
+    },
+    {
+      "epoch": 140.66666666666666,
+      "grad_norm": 137417.265625,
+      "learning_rate": 0.0002449787051171624,
+      "loss": 0.3664,
+      "step": 844
+    },
+    {
+      "epoch": 140.83333333333334,
+      "grad_norm": 112819.9765625,
+      "learning_rate": 0.00024372874960006742,
+      "loss": 0.3757,
+      "step": 845
+    },
+    {
+      "epoch": 141.0,
+      "grad_norm": 204521.53125,
+      "learning_rate": 0.00024248096254497287,
+      "loss": 0.3845,
+      "step": 846
+    },
+    {
+      "epoch": 141.0,
+      "eval_accuracy": 0.8204626914304334,
+      "eval_best_threshold": 0.39028987288475037,
+      "eval_f1": 0.8484181568088033,
+      "eval_loss": 0.37648096680641174,
+      "eval_pr_auc": 0.9232867403907821,
+      "eval_precision": 0.7960764068146619,
+      "eval_recall": 0.9081272084805654,
+      "eval_roc_auc": 0.909695106183187,
+      "eval_runtime": 3.9144,
+      "eval_samples_per_second": 784.022,
+      "eval_steps_per_second": 1.533,
+      "step": 846
+    },
+    {
+      "epoch": 141.16666666666666,
+      "grad_norm": 53121.89453125,
+      "learning_rate": 0.00024123535451014012,
+      "loss": 0.3609,
+      "step": 847
+    },
+    {
+      "epoch": 141.33333333333334,
+      "grad_norm": 133370.984375,
+      "learning_rate": 0.00023999193603539232,
+      "loss": 0.3849,
+      "step": 848
+    },
+    {
+      "epoch": 141.5,
+      "grad_norm": 164482.40625,
+      "learning_rate": 0.00023875071764202561,
+      "loss": 0.3461,
+      "step": 849
+    },
+    {
+      "epoch": 141.66666666666666,
+      "grad_norm": 76757.4609375,
+      "learning_rate": 0.00023751170983271996,
+      "loss": 0.3632,
+      "step": 850
+    },
+    {
+      "epoch": 141.83333333333334,
+      "grad_norm": 128235.109375,
+      "learning_rate": 0.00023627492309145028,
+      "loss": 0.3693,
+      "step": 851
+    },
+    {
+      "epoch": 142.0,
+      "grad_norm": 63393.671875,
+      "learning_rate": 0.0002350403678833976,
+      "loss": 0.3608,
+      "step": 852
+    },
+    {
+      "epoch": 142.0,
+      "eval_accuracy": 0.8269794721407625,
+      "eval_best_threshold": 0.37692126631736755,
+      "eval_f1": 0.8489331436699857,
+      "eval_loss": 0.38506025075912476,
+      "eval_pr_auc": 0.9212251720339615,
+      "eval_precision": 0.8211337369290038,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.9076310655089139,
+      "eval_runtime": 3.9427,
+      "eval_samples_per_second": 778.401,
+      "eval_steps_per_second": 1.522,
+      "step": 852
+    },
+    {
+      "epoch": 142.16666666666666,
+      "grad_norm": 165122.5,
+      "learning_rate": 0.00023380805465486083,
+      "loss": 0.3744,
+      "step": 853
+    },
+    {
+      "epoch": 142.33333333333334,
+      "grad_norm": 152976.96875,
+      "learning_rate": 0.00023257799383316798,
+      "loss": 0.3712,
+      "step": 854
+    },
+    {
+      "epoch": 142.5,
+      "grad_norm": 85063.9921875,
+      "learning_rate": 0.000231350195826588,
+      "loss": 0.355,
+      "step": 855
+    },
+    {
+      "epoch": 142.66666666666666,
+      "grad_norm": 120685.4609375,
+      "learning_rate": 0.00023012467102424372,
+      "loss": 0.3713,
+      "step": 856
+    },
+    {
+      "epoch": 142.83333333333334,
+      "grad_norm": 102459.109375,
+      "learning_rate": 0.00022890142979602147,
+      "loss": 0.3722,
+      "step": 857
+    },
+    {
+      "epoch": 143.0,
+      "grad_norm": 72757.78125,
+      "learning_rate": 0.00022768048249248646,
+      "loss": 0.3411,
+      "step": 858
+    },
+    {
+      "epoch": 143.0,
+      "eval_accuracy": 0.83088954056696,
+      "eval_best_threshold": 0.5101819634437561,
+      "eval_f1": 0.8498698293317906,
+      "eval_loss": 0.3803495764732361,
+      "eval_pr_auc": 0.9233289689301184,
+      "eval_precision": 0.8351335986355884,
+      "eval_recall": 0.8651354534746761,
+      "eval_roc_auc": 0.9097028382814465,
+      "eval_runtime": 3.9443,
+      "eval_samples_per_second": 778.087,
+      "eval_steps_per_second": 1.521,
+      "step": 858
+    },
+    {
+      "epoch": 143.16666666666666,
+      "grad_norm": 126750.984375,
+      "learning_rate": 0.00022646183944479272,
+      "loss": 0.3623,
+      "step": 859
+    },
+    {
+      "epoch": 143.33333333333334,
+      "grad_norm": 146971.515625,
+      "learning_rate": 0.000225245510964597,
+      "loss": 0.3923,
+      "step": 860
+    },
+    {
+      "epoch": 143.5,
+      "grad_norm": 92141.0234375,
+      "learning_rate": 0.00022403150734397092,
+      "loss": 0.3584,
+      "step": 861
+    },
+    {
+      "epoch": 143.66666666666666,
+      "grad_norm": 158932.25,
+      "learning_rate": 0.00022281983885531425,
+      "loss": 0.3741,
+      "step": 862
+    },
+    {
+      "epoch": 143.83333333333334,
+      "grad_norm": 72797.109375,
+      "learning_rate": 0.00022161051575126783,
+      "loss": 0.3628,
+      "step": 863
+    },
+    {
+      "epoch": 144.0,
+      "grad_norm": 189629.625,
+      "learning_rate": 0.00022040354826462666,
+      "loss": 0.3652,
+      "step": 864
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy": 0.8305637015314434,
+      "eval_best_threshold": 0.5188665390014648,
+      "eval_f1": 0.8505747126436781,
+      "eval_loss": 0.3749193251132965,
+      "eval_pr_auc": 0.9253198082355598,
+      "eval_precision": 0.8305274971941639,
+      "eval_recall": 0.8716136631330977,
+      "eval_roc_auc": 0.912020319954226,
+      "eval_runtime": 3.8321,
+      "eval_samples_per_second": 800.873,
+      "eval_steps_per_second": 1.566,
+      "step": 864
+    },
+    {
+      "epoch": 144.16666666666666,
+      "grad_norm": 105332.4609375,
+      "learning_rate": 0.00021919894660825363,
+      "loss": 0.3674,
+      "step": 865
+    },
+    {
+      "epoch": 144.33333333333334,
+      "grad_norm": 91438.1640625,
+      "learning_rate": 0.0002179967209749929,
+      "loss": 0.3586,
+      "step": 866
+    },
+    {
+      "epoch": 144.5,
+      "grad_norm": 93499.1796875,
+      "learning_rate": 0.0002167968815375837,
+      "loss": 0.3675,
+      "step": 867
+    },
+    {
+      "epoch": 144.66666666666666,
+      "grad_norm": 66955.734375,
+      "learning_rate": 0.0002155994384485742,
+      "loss": 0.3729,
+      "step": 868
+    },
+    {
+      "epoch": 144.83333333333334,
+      "grad_norm": 67269.2578125,
+      "learning_rate": 0.00021440440184023564,
+      "loss": 0.3571,
+      "step": 869
+    },
+    {
+      "epoch": 145.0,
+      "grad_norm": 103688.8515625,
+      "learning_rate": 0.00021321178182447708,
+      "loss": 0.3786,
+      "step": 870
+    },
+    {
+      "epoch": 145.0,
+      "eval_accuracy": 0.8240469208211144,
+      "eval_best_threshold": 0.3671317994594574,
+      "eval_f1": 0.8512396694214877,
+      "eval_loss": 0.37529799342155457,
+      "eval_pr_auc": 0.9232630000576302,
+      "eval_precision": 0.7996894409937888,
+      "eval_recall": 0.9098939929328622,
+      "eval_roc_auc": 0.9101865239836802,
+      "eval_runtime": 3.8101,
+      "eval_samples_per_second": 805.489,
+      "eval_steps_per_second": 1.575,
+      "step": 870
+    },
+    {
+      "epoch": 145.16666666666666,
+      "grad_norm": 91547.3515625,
+      "learning_rate": 0.00021202158849275833,
+      "loss": 0.3695,
+      "step": 871
+    },
+    {
+      "epoch": 145.33333333333334,
+      "grad_norm": 89502.46875,
+      "learning_rate": 0.00021083383191600674,
+      "loss": 0.3704,
+      "step": 872
+    },
+    {
+      "epoch": 145.5,
+      "grad_norm": 96104.1875,
+      "learning_rate": 0.0002096485221445301,
+      "loss": 0.358,
+      "step": 873
+    },
+    {
+      "epoch": 145.66666666666666,
+      "grad_norm": 108982.1796875,
+      "learning_rate": 0.00020846566920793263,
+      "loss": 0.3532,
+      "step": 874
+    },
+    {
+      "epoch": 145.83333333333334,
+      "grad_norm": 114974.890625,
+      "learning_rate": 0.00020728528311502974,
+      "loss": 0.3778,
+      "step": 875
+    },
+    {
+      "epoch": 146.0,
+      "grad_norm": 75530.546875,
+      "learning_rate": 0.00020610737385376348,
+      "loss": 0.3627,
+      "step": 876
+    },
+    {
+      "epoch": 146.0,
+      "eval_accuracy": 0.8312153796024764,
+      "eval_best_threshold": 0.4211173951625824,
+      "eval_f1": 0.853257790368272,
+      "eval_loss": 0.37683817744255066,
+      "eval_pr_auc": 0.9232412956565166,
+      "eval_precision": 0.8220524017467249,
+      "eval_recall": 0.8869257950530035,
+      "eval_roc_auc": 0.9097780114589695,
+      "eval_runtime": 3.9666,
+      "eval_samples_per_second": 773.701,
+      "eval_steps_per_second": 1.513,
+      "step": 876
+    },
+    {
+      "epoch": 146.16666666666666,
+      "grad_norm": 65020.58984375,
+      "learning_rate": 0.00020493195139111787,
+      "loss": 0.3548,
+      "step": 877
+    },
+    {
+      "epoch": 146.33333333333334,
+      "grad_norm": 67554.9609375,
+      "learning_rate": 0.00020375902567303471,
+      "loss": 0.3666,
+      "step": 878
+    },
+    {
+      "epoch": 146.5,
+      "grad_norm": 80817.4921875,
+      "learning_rate": 0.00020258860662432943,
+      "loss": 0.3667,
+      "step": 879
+    },
+    {
+      "epoch": 146.66666666666666,
+      "grad_norm": 129111.09375,
+      "learning_rate": 0.00020142070414860702,
+      "loss": 0.364,
+      "step": 880
+    },
+    {
+      "epoch": 146.83333333333334,
+      "grad_norm": 68261.125,
+      "learning_rate": 0.0002002553281281782,
+      "loss": 0.3613,
+      "step": 881
+    },
+    {
+      "epoch": 147.0,
+      "grad_norm": 85333.3671875,
+      "learning_rate": 0.00019909248842397582,
+      "loss": 0.3684,
+      "step": 882
+    },
+    {
+      "epoch": 147.0,
+      "eval_accuracy": 0.8305637015314434,
+      "eval_best_threshold": 0.45273062586784363,
+      "eval_f1": 0.8525241066364152,
+      "eval_loss": 0.3769237995147705,
+      "eval_pr_auc": 0.9225581489341609,
+      "eval_precision": 0.8222100656455142,
+      "eval_recall": 0.8851590106007067,
+      "eval_roc_auc": 0.9094597067472867,
+      "eval_runtime": 4.0322,
+      "eval_samples_per_second": 761.121,
+      "eval_steps_per_second": 1.488,
+      "step": 882
+    },
+    {
+      "epoch": 147.16666666666666,
+      "grad_norm": 111886.7578125,
+      "learning_rate": 0.00019793219487547183,
+      "loss": 0.3575,
+      "step": 883
+    },
+    {
+      "epoch": 147.33333333333334,
+      "grad_norm": 163052.109375,
+      "learning_rate": 0.00019677445730059345,
+      "loss": 0.3625,
+      "step": 884
+    },
+    {
+      "epoch": 147.5,
+      "grad_norm": 66131.125,
+      "learning_rate": 0.00019561928549563967,
+      "loss": 0.3651,
+      "step": 885
+    },
+    {
+      "epoch": 147.66666666666666,
+      "grad_norm": 199836.4375,
+      "learning_rate": 0.0001944666892352001,
+      "loss": 0.3858,
+      "step": 886
+    },
+    {
+      "epoch": 147.83333333333334,
+      "grad_norm": 204624.03125,
+      "learning_rate": 0.00019331667827207044,
+      "loss": 0.3718,
+      "step": 887
+    },
+    {
+      "epoch": 148.0,
+      "grad_norm": 118003.953125,
+      "learning_rate": 0.00019216926233717085,
+      "loss": 0.3635,
+      "step": 888
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy": 0.83088954056696,
+      "eval_best_threshold": 0.5705100893974304,
+      "eval_f1": 0.8493468795355588,
+      "eval_loss": 0.3833242356777191,
+      "eval_pr_auc": 0.9233202281747017,
+      "eval_precision": 0.8374356038923869,
+      "eval_recall": 0.8616018845700825,
+      "eval_roc_auc": 0.9097142216483287,
+      "eval_runtime": 3.9143,
+      "eval_samples_per_second": 784.048,
+      "eval_steps_per_second": 1.533,
+      "step": 888
+    },
+    {
+      "epoch": 148.16666666666666,
+      "grad_norm": 222922.71875,
+      "learning_rate": 0.00019102445113946344,
+      "loss": 0.3759,
+      "step": 889
+    },
+    {
+      "epoch": 148.33333333333334,
+      "grad_norm": 234529.6875,
+      "learning_rate": 0.00018988225436587002,
+      "loss": 0.3723,
+      "step": 890
+    },
+    {
+      "epoch": 148.5,
+      "grad_norm": 210516.421875,
+      "learning_rate": 0.0001887426816811903,
+      "loss": 0.3804,
+      "step": 891
+    },
+    {
+      "epoch": 148.66666666666666,
+      "grad_norm": 59804.4765625,
+      "learning_rate": 0.00018760574272802001,
+      "loss": 0.3669,
+      "step": 892
+    },
+    {
+      "epoch": 148.83333333333334,
+      "grad_norm": 221783.171875,
+      "learning_rate": 0.0001864714471266691,
+      "loss": 0.3675,
+      "step": 893
+    },
+    {
+      "epoch": 149.0,
+      "grad_norm": 220364.734375,
+      "learning_rate": 0.00018533980447508135,
+      "loss": 0.3643,
+      "step": 894
+    },
+    {
+      "epoch": 149.0,
+      "eval_accuracy": 0.8227435646790485,
+      "eval_best_threshold": 0.4187193512916565,
+      "eval_f1": 0.8465011286681715,
+      "eval_loss": 0.37492498755455017,
+      "eval_pr_auc": 0.924585944838121,
+      "eval_precision": 0.8125677139761647,
+      "eval_recall": 0.8833922261484098,
+      "eval_roc_auc": 0.9101375540280366,
+      "eval_runtime": 3.9895,
+      "eval_samples_per_second": 769.264,
+      "eval_steps_per_second": 1.504,
+      "step": 894
+    },
+    {
+      "epoch": 149.16666666666666,
+      "grad_norm": 81329.15625,
+      "learning_rate": 0.00018421082434875132,
+      "loss": 0.3556,
+      "step": 895
+    },
+    {
+      "epoch": 149.33333333333334,
+      "grad_norm": 156673.375,
+      "learning_rate": 0.0001830845163006448,
+      "loss": 0.3886,
+      "step": 896
+    },
+    {
+      "epoch": 149.5,
+      "grad_norm": 65551.578125,
+      "learning_rate": 0.00018196088986111798,
+      "loss": 0.3523,
+      "step": 897
+    },
+    {
+      "epoch": 149.66666666666666,
+      "grad_norm": 121681.234375,
+      "learning_rate": 0.00018083995453783603,
+      "loss": 0.3617,
+      "step": 898
+    },
+    {
+      "epoch": 149.83333333333334,
+      "grad_norm": 119770.8984375,
+      "learning_rate": 0.0001797217198156924,
+      "loss": 0.3687,
+      "step": 899
+    },
+    {
+      "epoch": 150.0,
+      "grad_norm": 62330.62890625,
+      "learning_rate": 0.0001786061951567303,
+      "loss": 0.3559,
+      "step": 900
+    },
+    {
+      "epoch": 150.0,
+      "eval_accuracy": 0.826653633105246,
+      "eval_best_threshold": 0.47909417748451233,
+      "eval_f1": 0.8487777146105742,
+      "eval_loss": 0.3842035233974457,
+      "eval_pr_auc": 0.9212527914064633,
+      "eval_precision": 0.8203296703296703,
+      "eval_recall": 0.8792697290930507,
+      "eval_roc_auc": 0.9077642294233831,
+      "eval_runtime": 4.006,
+      "eval_samples_per_second": 766.103,
+      "eval_steps_per_second": 1.498,
+      "step": 900
+    },
+    {
+      "epoch": 150.16666666666666,
+      "grad_norm": 181080.671875,
+      "learning_rate": 0.00017749339000006098,
+      "loss": 0.3687,
+      "step": 901
+    },
+    {
+      "epoch": 150.33333333333334,
+      "grad_norm": 173832.859375,
+      "learning_rate": 0.00017638331376178384,
+      "loss": 0.3865,
+      "step": 902
+    },
+    {
+      "epoch": 150.5,
+      "grad_norm": 120683.6015625,
+      "learning_rate": 0.00017527597583490823,
+      "loss": 0.3556,
+      "step": 903
+    },
+    {
+      "epoch": 150.66666666666666,
+      "grad_norm": 215743.546875,
+      "learning_rate": 0.00017417138558927244,
+      "loss": 0.3401,
+      "step": 904
+    },
+    {
+      "epoch": 150.83333333333334,
+      "grad_norm": 303968.78125,
+      "learning_rate": 0.00017306955237146522,
+      "loss": 0.3778,
+      "step": 905
+    },
+    {
+      "epoch": 151.0,
+      "grad_norm": 278908.25,
+      "learning_rate": 0.00017197048550474643,
+      "loss": 0.3826,
+      "step": 906
+    },
+    {
+      "epoch": 151.0,
+      "eval_accuracy": 0.83088954056696,
+      "eval_best_threshold": 0.4328542947769165,
+      "eval_f1": 0.8517566409597258,
+      "eval_loss": 0.37284591794013977,
+      "eval_pr_auc": 0.9236686774663367,
+      "eval_precision": 0.826955074875208,
+      "eval_recall": 0.8780918727915195,
+      "eval_roc_auc": 0.9110409208413552,
+      "eval_runtime": 4.0186,
+      "eval_samples_per_second": 763.692,
+      "eval_steps_per_second": 1.493,
+      "step": 906
+    },
+    {
+      "epoch": 151.16666666666666,
+      "grad_norm": 151452.84375,
+      "learning_rate": 0.00017087419428896806,
+      "loss": 0.3363,
+      "step": 907
+    },
+    {
+      "epoch": 151.33333333333334,
+      "grad_norm": 205682.671875,
+      "learning_rate": 0.00016978068800049622,
+      "loss": 0.3828,
+      "step": 908
+    },
+    {
+      "epoch": 151.5,
+      "grad_norm": 306064.1875,
+      "learning_rate": 0.00016868997589213137,
+      "loss": 0.3857,
+      "step": 909
+    },
+    {
+      "epoch": 151.66666666666666,
+      "grad_norm": 246629.0,
+      "learning_rate": 0.00016760206719303107,
+      "loss": 0.3897,
+      "step": 910
+    },
+    {
+      "epoch": 151.83333333333334,
+      "grad_norm": 102579.609375,
+      "learning_rate": 0.0001665169711086319,
+      "loss": 0.3532,
+      "step": 911
+    },
+    {
+      "epoch": 152.0,
+      "grad_norm": 155984.015625,
+      "learning_rate": 0.00016543469682057105,
+      "loss": 0.3756,
+      "step": 912
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy": 0.8204626914304334,
+      "eval_best_threshold": 0.2976920008659363,
+      "eval_f1": 0.8457878533445284,
+      "eval_loss": 0.40224388241767883,
+      "eval_pr_auc": 0.9200705150322268,
+      "eval_precision": 0.8058666666666666,
+      "eval_recall": 0.8898704358068316,
+      "eval_roc_auc": 0.906433019839705,
+      "eval_runtime": 3.9788,
+      "eval_samples_per_second": 771.348,
+      "eval_steps_per_second": 1.508,
+      "step": 912
+    },
+    {
+      "epoch": 152.16666666666666,
+      "grad_norm": 299503.15625,
+      "learning_rate": 0.00016435525348660823,
+      "loss": 0.3917,
+      "step": 913
+    },
+    {
+      "epoch": 152.33333333333334,
+      "grad_norm": 268243.5625,
+      "learning_rate": 0.00016327865024054982,
+      "loss": 0.3717,
+      "step": 914
+    },
+    {
+      "epoch": 152.5,
+      "grad_norm": 88446.296875,
+      "learning_rate": 0.0001622048961921699,
+      "loss": 0.3505,
+      "step": 915
+    },
+    {
+      "epoch": 152.66666666666666,
+      "grad_norm": 175196.296875,
+      "learning_rate": 0.0001611340004271339,
+      "loss": 0.3534,
+      "step": 916
+    },
+    {
+      "epoch": 152.83333333333334,
+      "grad_norm": 200618.140625,
+      "learning_rate": 0.00016006597200692168,
+      "loss": 0.3837,
+      "step": 917
+    },
+    {
+      "epoch": 153.0,
+      "grad_norm": 138927.59375,
+      "learning_rate": 0.00015900081996875082,
+      "loss": 0.3716,
+      "step": 918
+    },
+    {
+      "epoch": 153.0,
+      "eval_accuracy": 0.8334962528510915,
+      "eval_best_threshold": 0.49080690741539,
+      "eval_f1": 0.8498383779018513,
+      "eval_loss": 0.37779301404953003,
+      "eval_pr_auc": 0.9236738139526288,
+      "eval_precision": 0.8480938416422288,
+      "eval_recall": 0.8515901060070671,
+      "eval_roc_auc": 0.909127226522128,
+      "eval_runtime": 4.0135,
+      "eval_samples_per_second": 764.662,
+      "eval_steps_per_second": 1.495,
+      "step": 918
+    },
+    {
+      "epoch": 153.16666666666666,
+      "grad_norm": 81768.3828125,
+      "learning_rate": 0.0001579385533255001,
+      "loss": 0.3556,
+      "step": 919
+    },
+    {
+      "epoch": 153.33333333333334,
+      "grad_norm": 165935.765625,
+      "learning_rate": 0.00015687918106563326,
+      "loss": 0.352,
+      "step": 920
+    },
+    {
+      "epoch": 153.5,
+      "grad_norm": 228723.890625,
+      "learning_rate": 0.00015582271215312294,
+      "loss": 0.3631,
+      "step": 921
+    },
+    {
+      "epoch": 153.66666666666666,
+      "grad_norm": 213650.1875,
+      "learning_rate": 0.00015476915552737532,
+      "loss": 0.3663,
+      "step": 922
+    },
+    {
+      "epoch": 153.83333333333334,
+      "grad_norm": 126728.4453125,
+      "learning_rate": 0.00015371852010315302,
+      "loss": 0.3653,
+      "step": 923
+    },
+    {
+      "epoch": 154.0,
+      "grad_norm": 166727.078125,
+      "learning_rate": 0.00015267081477050133,
+      "loss": 0.3684,
+      "step": 924
+    },
+    {
+      "epoch": 154.0,
+      "eval_accuracy": 0.8305637015314434,
+      "eval_best_threshold": 0.570549488067627,
+      "eval_f1": 0.8510028653295129,
+      "eval_loss": 0.3899708092212677,
+      "eval_pr_auc": 0.9218689165850973,
+      "eval_precision": 0.8286830357142857,
+      "eval_recall": 0.8745583038869258,
+      "eval_roc_auc": 0.9100284455303747,
+      "eval_runtime": 3.9666,
+      "eval_samples_per_second": 773.719,
+      "eval_steps_per_second": 1.513,
+      "step": 924
+    },
+    {
+      "epoch": 154.16666666666666,
+      "grad_norm": 238350.765625,
+      "learning_rate": 0.00015162604839467266,
+      "loss": 0.3786,
+      "step": 925
+    },
+    {
+      "epoch": 154.33333333333334,
+      "grad_norm": 208803.765625,
+      "learning_rate": 0.00015058422981604997,
+      "loss": 0.3754,
+      "step": 926
+    },
+    {
+      "epoch": 154.5,
+      "grad_norm": 146063.75,
+      "learning_rate": 0.00014954536785007454,
+      "loss": 0.3698,
+      "step": 927
+    },
+    {
+      "epoch": 154.66666666666666,
+      "grad_norm": 141785.609375,
+      "learning_rate": 0.00014850947128716913,
+      "loss": 0.3579,
+      "step": 928
+    },
+    {
+      "epoch": 154.83333333333334,
+      "grad_norm": 191827.84375,
+      "learning_rate": 0.0001474765488926648,
+      "loss": 0.3634,
+      "step": 929
+    },
+    {
+      "epoch": 155.0,
+      "grad_norm": 185356.5,
+      "learning_rate": 0.00014644660940672628,
+      "loss": 0.3723,
+      "step": 930
+    },
+    {
+      "epoch": 155.0,
+      "eval_accuracy": 0.8367546432062561,
+      "eval_best_threshold": 0.40736904740333557,
+      "eval_f1": 0.8571428571428571,
+      "eval_loss": 0.37203943729400635,
+      "eval_pr_auc": 0.9237786897972513,
+      "eval_precision": 0.8308457711442786,
+      "eval_recall": 0.8851590106007067,
+      "eval_roc_auc": 0.9116079413803857,
+      "eval_runtime": 3.9799,
+      "eval_samples_per_second": 771.132,
+      "eval_steps_per_second": 1.508,
+      "step": 930
+    },
+    {
+      "epoch": 155.16666666666666,
+      "grad_norm": 92037.6640625,
+      "learning_rate": 0.00014541966154427822,
+      "loss": 0.3771,
+      "step": 931
+    },
+    {
+      "epoch": 155.33333333333334,
+      "grad_norm": 62586.59375,
+      "learning_rate": 0.00014439571399493145,
+      "loss": 0.3654,
+      "step": 932
+    },
+    {
+      "epoch": 155.5,
+      "grad_norm": 77046.0390625,
+      "learning_rate": 0.00014337477542290927,
+      "loss": 0.3472,
+      "step": 933
+    },
+    {
+      "epoch": 155.66666666666666,
+      "grad_norm": 54610.234375,
+      "learning_rate": 0.00014235685446697434,
+      "loss": 0.3664,
+      "step": 934
+    },
+    {
+      "epoch": 155.83333333333334,
+      "grad_norm": 109650.8125,
+      "learning_rate": 0.00014134195974035524,
+      "loss": 0.3338,
+      "step": 935
+    },
+    {
+      "epoch": 156.0,
+      "grad_norm": 68304.0625,
+      "learning_rate": 0.00014033009983067452,
+      "loss": 0.3468,
+      "step": 936
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy": 0.8328445747800587,
+      "eval_best_threshold": 0.46987348794937134,
+      "eval_f1": 0.8525438344351826,
+      "eval_loss": 0.37313708662986755,
+      "eval_pr_auc": 0.9235315191777744,
+      "eval_precision": 0.832678270634475,
+      "eval_recall": 0.8733804475853946,
+      "eval_roc_auc": 0.9111019185054027,
+      "eval_runtime": 4.1093,
+      "eval_samples_per_second": 746.848,
+      "eval_steps_per_second": 1.46,
+      "step": 936
+    },
+    {
+      "epoch": 156.16666666666666,
+      "grad_norm": 97583.125,
+      "learning_rate": 0.0001393212832998743,
+      "loss": 0.3536,
+      "step": 937
+    },
+    {
+      "epoch": 156.33333333333334,
+      "grad_norm": 62628.94921875,
+      "learning_rate": 0.000138315518684146,
+      "loss": 0.3457,
+      "step": 938
+    },
+    {
+      "epoch": 156.5,
+      "grad_norm": 79858.0078125,
+      "learning_rate": 0.0001373128144938563,
+      "loss": 0.3513,
+      "step": 939
+    },
+    {
+      "epoch": 156.66666666666666,
+      "grad_norm": 87592.1015625,
+      "learning_rate": 0.00013631317921347562,
+      "loss": 0.3656,
+      "step": 940
+    },
+    {
+      "epoch": 156.83333333333334,
+      "grad_norm": 72757.2578125,
+      "learning_rate": 0.00013531662130150736,
+      "loss": 0.3523,
+      "step": 941
+    },
+    {
+      "epoch": 157.0,
+      "grad_norm": 77319.5859375,
+      "learning_rate": 0.00013432314919041476,
+      "loss": 0.3549,
+      "step": 942
+    },
+    {
+      "epoch": 157.0,
+      "eval_accuracy": 0.83088954056696,
+      "eval_best_threshold": 0.424466073513031,
+      "eval_f1": 0.85099052540913,
+      "eval_loss": 0.3759530484676361,
+      "eval_pr_auc": 0.9233009083235354,
+      "eval_precision": 0.8302521008403362,
+      "eval_recall": 0.872791519434629,
+      "eval_roc_auc": 0.9103029350185871,
+      "eval_runtime": 4.1782,
+      "eval_samples_per_second": 734.535,
+      "eval_steps_per_second": 1.436,
+      "step": 942
+    },
+    {
+      "epoch": 157.16666666666666,
+      "grad_norm": 95107.390625,
+      "learning_rate": 0.00013333277128655064,
+      "loss": 0.3339,
+      "step": 943
+    },
+    {
+      "epoch": 157.33333333333334,
+      "grad_norm": 86475.1171875,
+      "learning_rate": 0.00013234549597008571,
+      "loss": 0.3561,
+      "step": 944
+    },
+    {
+      "epoch": 157.5,
+      "grad_norm": 80413.9765625,
+      "learning_rate": 0.000131361331594938,
+      "loss": 0.3521,
+      "step": 945
+    },
+    {
+      "epoch": 157.66666666666666,
+      "grad_norm": 131372.859375,
+      "learning_rate": 0.00013038028648870205,
+      "loss": 0.3606,
+      "step": 946
+    },
+    {
+      "epoch": 157.83333333333334,
+      "grad_norm": 89998.09375,
+      "learning_rate": 0.00012940236895257824,
+      "loss": 0.3661,
+      "step": 947
+    },
+    {
+      "epoch": 158.0,
+      "grad_norm": 110346.8359375,
+      "learning_rate": 0.00012842758726130281,
+      "loss": 0.3606,
+      "step": 948
+    },
+    {
+      "epoch": 158.0,
+      "eval_accuracy": 0.8253502769631802,
+      "eval_best_threshold": 0.3530597686767578,
+      "eval_f1": 0.8529088913282108,
+      "eval_loss": 0.37229081988334656,
+      "eval_pr_auc": 0.9241729081786076,
+      "eval_precision": 0.7985611510791367,
+      "eval_recall": 0.9151943462897526,
+      "eval_roc_auc": 0.9118356087180268,
+      "eval_runtime": 4.152,
+      "eval_samples_per_second": 739.162,
+      "eval_steps_per_second": 1.445,
+      "step": 948
+    },
+    {
+      "epoch": 158.16666666666666,
+      "grad_norm": 80206.3359375,
+      "learning_rate": 0.00012745594966307822,
+      "loss": 0.3499,
+      "step": 949
+    },
+    {
+      "epoch": 158.33333333333334,
+      "grad_norm": 65314.18359375,
+      "learning_rate": 0.00012648746437950208,
+      "loss": 0.3384,
+      "step": 950
+    },
+    {
+      "epoch": 158.5,
+      "grad_norm": 89174.2578125,
+      "learning_rate": 0.0001255221396054989,
+      "loss": 0.3593,
+      "step": 951
+    },
+    {
+      "epoch": 158.66666666666666,
+      "grad_norm": 67226.4375,
+      "learning_rate": 0.0001245599835092504,
+      "loss": 0.364,
+      "step": 952
+    },
+    {
+      "epoch": 158.83333333333334,
+      "grad_norm": 83786.7734375,
+      "learning_rate": 0.00012360100423212605,
+      "loss": 0.3524,
+      "step": 953
+    },
+    {
+      "epoch": 159.0,
+      "grad_norm": 67511.8359375,
+      "learning_rate": 0.000122645209888614,
+      "loss": 0.3537,
+      "step": 954
+    },
+    {
+      "epoch": 159.0,
+      "eval_accuracy": 0.8413163897034864,
+      "eval_best_threshold": 0.5131648182868958,
+      "eval_f1": 0.8578102189781022,
+      "eval_loss": 0.3645898997783661,
+      "eval_pr_auc": 0.9277610171354891,
+      "eval_precision": 0.8506079907353793,
+      "eval_recall": 0.8651354534746761,
+      "eval_roc_auc": 0.9155506671512115,
+      "eval_runtime": 4.1736,
+      "eval_samples_per_second": 735.334,
+      "eval_steps_per_second": 1.438,
+      "step": 954
+    },
+    {
+      "epoch": 159.16666666666666,
+      "grad_norm": 53345.765625,
+      "learning_rate": 0.00012169260856625358,
+      "loss": 0.351,
+      "step": 955
+    },
+    {
+      "epoch": 159.33333333333334,
+      "grad_norm": 168120.953125,
+      "learning_rate": 0.00012074320832556557,
+      "loss": 0.3663,
+      "step": 956
+    },
+    {
+      "epoch": 159.5,
+      "grad_norm": 154755.359375,
+      "learning_rate": 0.00011979701719998454,
+      "loss": 0.3637,
+      "step": 957
+    },
+    {
+      "epoch": 159.66666666666666,
+      "grad_norm": 81563.53125,
+      "learning_rate": 0.00011885404319579107,
+      "loss": 0.3413,
+      "step": 958
+    },
+    {
+      "epoch": 159.83333333333334,
+      "grad_norm": 119309.7578125,
+      "learning_rate": 0.00011791429429204342,
+      "loss": 0.3577,
+      "step": 959
+    },
+    {
+      "epoch": 160.0,
+      "grad_norm": 72382.3359375,
+      "learning_rate": 0.00011697777844051105,
+      "loss": 0.3399,
+      "step": 960
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy": 0.8292603453893777,
+      "eval_best_threshold": 0.44661715626716614,
+      "eval_f1": 0.8506271379703535,
+      "eval_loss": 0.3717084228992462,
+      "eval_pr_auc": 0.9246020036155943,
+      "eval_precision": 0.8243093922651934,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.9123667609123532,
+      "eval_runtime": 4.1104,
+      "eval_samples_per_second": 746.647,
+      "eval_steps_per_second": 1.46,
+      "step": 960
+    },
+    {
+      "epoch": 160.16666666666666,
+      "grad_norm": 67661.6640625,
+      "learning_rate": 0.00011604450356560575,
+      "loss": 0.3437,
+      "step": 961
+    },
+    {
+      "epoch": 160.33333333333334,
+      "grad_norm": 119658.078125,
+      "learning_rate": 0.00011511447756431604,
+      "loss": 0.3525,
+      "step": 962
+    },
+    {
+      "epoch": 160.5,
+      "grad_norm": 106822.890625,
+      "learning_rate": 0.00011418770830614011,
+      "loss": 0.3454,
+      "step": 963
+    },
+    {
+      "epoch": 160.66666666666666,
+      "grad_norm": 57645.9296875,
+      "learning_rate": 0.00011326420363301809,
+      "loss": 0.3424,
+      "step": 964
+    },
+    {
+      "epoch": 160.83333333333334,
+      "grad_norm": 77775.6796875,
+      "learning_rate": 0.00011234397135926705,
+      "loss": 0.3647,
+      "step": 965
+    },
+    {
+      "epoch": 161.0,
+      "grad_norm": 90241.984375,
+      "learning_rate": 0.00011142701927151455,
+      "loss": 0.3537,
+      "step": 966
+    },
+    {
+      "epoch": 161.0,
+      "eval_accuracy": 0.8400130335614207,
+      "eval_best_threshold": 0.4722670316696167,
+      "eval_f1": 0.858948577994829,
+      "eval_loss": 0.36644211411476135,
+      "eval_pr_auc": 0.9255285397987627,
+      "eval_precision": 0.8384744812114414,
+      "eval_recall": 0.8804475853945819,
+      "eval_roc_auc": 0.9146513811675296,
+      "eval_runtime": 4.2747,
+      "eval_samples_per_second": 717.945,
+      "eval_steps_per_second": 1.404,
+      "step": 966
+    },
+    {
+      "epoch": 161.16666666666666,
+      "grad_norm": 81642.9296875,
+      "learning_rate": 0.00011051335512863232,
+      "loss": 0.3504,
+      "step": 967
+    },
+    {
+      "epoch": 161.33333333333334,
+      "grad_norm": 160026.0,
+      "learning_rate": 0.0001096029866616704,
+      "loss": 0.3406,
+      "step": 968
+    },
+    {
+      "epoch": 161.5,
+      "grad_norm": 120099.6171875,
+      "learning_rate": 0.00010869592157379304,
+      "loss": 0.3677,
+      "step": 969
+    },
+    {
+      "epoch": 161.66666666666666,
+      "grad_norm": 75987.7421875,
+      "learning_rate": 0.00010779216754021215,
+      "loss": 0.3498,
+      "step": 970
+    },
+    {
+      "epoch": 161.83333333333334,
+      "grad_norm": 116359.1484375,
+      "learning_rate": 0.00010689173220812293,
+      "loss": 0.3335,
+      "step": 971
+    },
+    {
+      "epoch": 162.0,
+      "grad_norm": 79084.625,
+      "learning_rate": 0.00010599462319663906,
+      "loss": 0.3546,
+      "step": 972
+    },
+    {
+      "epoch": 162.0,
+      "eval_accuracy": 0.8289345063538612,
+      "eval_best_threshold": 0.4081013798713684,
+      "eval_f1": 0.8509792790235595,
+      "eval_loss": 0.3715569078922272,
+      "eval_pr_auc": 0.9236155693773005,
+      "eval_precision": 0.8213698630136986,
+      "eval_recall": 0.8828032979976443,
+      "eval_roc_auc": 0.9119734978036544,
+      "eval_runtime": 4.2416,
+      "eval_samples_per_second": 723.546,
+      "eval_steps_per_second": 1.415,
+      "step": 972
+    },
+    {
+      "epoch": 162.16666666666666,
+      "grad_norm": 94386.2265625,
+      "learning_rate": 0.00010510084809672838,
+      "loss": 0.3692,
+      "step": 973
+    },
+    {
+      "epoch": 162.33333333333334,
+      "grad_norm": 74503.6953125,
+      "learning_rate": 0.00010421041447114837,
+      "loss": 0.3599,
+      "step": 974
+    },
+    {
+      "epoch": 162.5,
+      "grad_norm": 100313.7578125,
+      "learning_rate": 0.00010332332985438247,
+      "loss": 0.3539,
+      "step": 975
+    },
+    {
+      "epoch": 162.66666666666666,
+      "grad_norm": 101121.765625,
+      "learning_rate": 0.00010243960175257604,
+      "loss": 0.3283,
+      "step": 976
+    },
+    {
+      "epoch": 162.83333333333334,
+      "grad_norm": 100584.296875,
+      "learning_rate": 0.00010155923764347352,
+      "loss": 0.3514,
+      "step": 977
+    },
+    {
+      "epoch": 163.0,
+      "grad_norm": 88137.6328125,
+      "learning_rate": 0.00010068224497635369,
+      "loss": 0.3506,
+      "step": 978
+    },
+    {
+      "epoch": 163.0,
+      "eval_accuracy": 0.830237862495927,
+      "eval_best_threshold": 0.4215361773967743,
+      "eval_f1": 0.8511853756069694,
+      "eval_loss": 0.3692292869091034,
+      "eval_pr_auc": 0.9259921792261762,
+      "eval_precision": 0.8264004437049363,
+      "eval_recall": 0.8775029446407538,
+      "eval_roc_auc": 0.9133734371496394,
+      "eval_runtime": 4.2429,
+      "eval_samples_per_second": 723.323,
+      "eval_steps_per_second": 1.414,
+      "step": 978
+    },
+    {
+      "epoch": 163.16666666666666,
+      "grad_norm": 71053.8828125,
+      "learning_rate": 9.980863117196814e-05,
+      "loss": 0.3631,
+      "step": 979
+    },
+    {
+      "epoch": 163.33333333333334,
+      "grad_norm": 158759.703125,
+      "learning_rate": 9.893840362247809e-05,
+      "loss": 0.3573,
+      "step": 980
+    },
+    {
+      "epoch": 163.5,
+      "grad_norm": 124987.6015625,
+      "learning_rate": 9.807156969139135e-05,
+      "loss": 0.3565,
+      "step": 981
+    },
+    {
+      "epoch": 163.66666666666666,
+      "grad_norm": 74836.234375,
+      "learning_rate": 9.720813671350031e-05,
+      "loss": 0.3333,
+      "step": 982
+    },
+    {
+      "epoch": 163.83333333333334,
+      "grad_norm": 70798.34375,
+      "learning_rate": 9.634811199482008e-05,
+      "loss": 0.3462,
+      "step": 983
+    },
+    {
+      "epoch": 164.0,
+      "grad_norm": 82296.0234375,
+      "learning_rate": 9.549150281252633e-05,
+      "loss": 0.3453,
+      "step": 984
+    },
+    {
+      "epoch": 164.0,
+      "eval_accuracy": 0.8328445747800587,
+      "eval_best_threshold": 0.4308677613735199,
+      "eval_f1": 0.8538877812589006,
+      "eval_loss": 0.36614447832107544,
+      "eval_pr_auc": 0.9269770160926805,
+      "eval_precision": 0.8268063982349697,
+      "eval_recall": 0.8828032979976443,
+      "eval_roc_auc": 0.9147209700518653,
+      "eval_runtime": 4.2745,
+      "eval_samples_per_second": 717.975,
+      "eval_steps_per_second": 1.404,
+      "step": 984
+    },
+    {
+      "epoch": 164.16666666666666,
+      "grad_norm": 41647.91015625,
+      "learning_rate": 9.46383164148939e-05,
+      "loss": 0.3571,
+      "step": 985
+    },
+    {
+      "epoch": 164.33333333333334,
+      "grad_norm": 49898.0625,
+      "learning_rate": 9.378856002123548e-05,
+      "loss": 0.3608,
+      "step": 986
+    },
+    {
+      "epoch": 164.5,
+      "grad_norm": 142842.78125,
+      "learning_rate": 9.294224082184043e-05,
+      "loss": 0.3591,
+      "step": 987
+    },
+    {
+      "epoch": 164.66666666666666,
+      "grad_norm": 97403.40625,
+      "learning_rate": 9.209936597791407e-05,
+      "loss": 0.3492,
+      "step": 988
+    },
+    {
+      "epoch": 164.83333333333334,
+      "grad_norm": 72412.1328125,
+      "learning_rate": 9.125994262151683e-05,
+      "loss": 0.3452,
+      "step": 989
+    },
+    {
+      "epoch": 165.0,
+      "grad_norm": 80854.3046875,
+      "learning_rate": 9.042397785550405e-05,
+      "loss": 0.3233,
+      "step": 990
+    },
+    {
+      "epoch": 165.0,
+      "eval_accuracy": 0.8351254480286738,
+      "eval_best_threshold": 0.4565604031085968,
+      "eval_f1": 0.8544303797468354,
+      "eval_loss": 0.36564016342163086,
+      "eval_pr_auc": 0.9278390368820122,
+      "eval_precision": 0.8352080989876266,
+      "eval_recall": 0.8745583038869258,
+      "eval_roc_auc": 0.9147793903498259,
+      "eval_runtime": 4.2545,
+      "eval_samples_per_second": 721.352,
+      "eval_steps_per_second": 1.41,
+      "step": 990
+    },
+    {
+      "epoch": 165.16666666666666,
+      "grad_norm": 49396.58984375,
+      "learning_rate": 8.959147875346652e-05,
+      "loss": 0.3176,
+      "step": 991
+    },
+    {
+      "epoch": 165.33333333333334,
+      "grad_norm": 112044.234375,
+      "learning_rate": 8.876245235966885e-05,
+      "loss": 0.3341,
+      "step": 992
+    },
+    {
+      "epoch": 165.5,
+      "grad_norm": 80428.34375,
+      "learning_rate": 8.793690568899215e-05,
+      "loss": 0.3543,
+      "step": 993
+    },
+    {
+      "epoch": 165.66666666666666,
+      "grad_norm": 66167.703125,
+      "learning_rate": 8.711484572687295e-05,
+      "loss": 0.3357,
+      "step": 994
+    },
+    {
+      "epoch": 165.83333333333334,
+      "grad_norm": 87920.015625,
+      "learning_rate": 8.629627942924473e-05,
+      "loss": 0.3829,
+      "step": 995
+    },
+    {
+      "epoch": 166.0,
+      "grad_norm": 150330.1875,
+      "learning_rate": 8.548121372247918e-05,
+      "loss": 0.3576,
+      "step": 996
+    },
+    {
+      "epoch": 166.0,
+      "eval_accuracy": 0.8341479309221245,
+      "eval_best_threshold": 0.41728565096855164,
+      "eval_f1": 0.8544466685730626,
+      "eval_loss": 0.36713624000549316,
+      "eval_pr_auc": 0.92653588836402,
+      "eval_precision": 0.830461367426348,
+      "eval_recall": 0.8798586572438163,
+      "eval_roc_auc": 0.9145203650581325,
+      "eval_runtime": 4.2872,
+      "eval_samples_per_second": 715.845,
+      "eval_steps_per_second": 1.4,
+      "step": 996
+    },
+    {
+      "epoch": 166.16666666666666,
+      "grad_norm": 73645.6640625,
+      "learning_rate": 8.466965550332728e-05,
+      "loss": 0.3528,
+      "step": 997
+    },
+    {
+      "epoch": 166.33333333333334,
+      "grad_norm": 83714.1015625,
+      "learning_rate": 8.386161163886118e-05,
+      "loss": 0.3432,
+      "step": 998
+    },
+    {
+      "epoch": 166.5,
+      "grad_norm": 146188.8125,
+      "learning_rate": 8.305708896641595e-05,
+      "loss": 0.341,
+      "step": 999
+    },
+    {
+      "epoch": 166.66666666666666,
+      "grad_norm": 69561.3671875,
+      "learning_rate": 8.225609429353187e-05,
+      "loss": 0.35,
+      "step": 1000
+    },
+    {
+      "epoch": 166.83333333333334,
+      "grad_norm": 170722.265625,
+      "learning_rate": 8.145863439789658e-05,
+      "loss": 0.3656,
+      "step": 1001
+    },
+    {
+      "epoch": 167.0,
+      "grad_norm": 155031.265625,
+      "learning_rate": 8.066471602728804e-05,
+      "loss": 0.3356,
+      "step": 1002
+    },
+    {
+      "epoch": 167.0,
+      "eval_accuracy": 0.8344737699576409,
+      "eval_best_threshold": 0.4809587299823761,
+      "eval_f1": 0.8542742398164085,
+      "eval_loss": 0.3662647604942322,
+      "eval_pr_auc": 0.926724531939227,
+      "eval_precision": 0.8327740492170023,
+      "eval_recall": 0.8769140164899882,
+      "eval_roc_auc": 0.915205085315113,
+      "eval_runtime": 4.3238,
+      "eval_samples_per_second": 709.793,
+      "eval_steps_per_second": 1.388,
+      "step": 1002
+    },
+    {
+      "epoch": 167.16666666666666,
+      "grad_norm": 101407.078125,
+      "learning_rate": 7.987434589951726e-05,
+      "loss": 0.3447,
+      "step": 1003
+    },
+    {
+      "epoch": 167.33333333333334,
+      "grad_norm": 103118.65625,
+      "learning_rate": 7.908753070237123e-05,
+      "loss": 0.3635,
+      "step": 1004
+    },
+    {
+      "epoch": 167.5,
+      "grad_norm": 123300.7265625,
+      "learning_rate": 7.830427709355725e-05,
+      "loss": 0.3502,
+      "step": 1005
+    },
+    {
+      "epoch": 167.66666666666666,
+      "grad_norm": 127009.6328125,
+      "learning_rate": 7.752459170064491e-05,
+      "loss": 0.3332,
+      "step": 1006
+    },
+    {
+      "epoch": 167.83333333333334,
+      "grad_norm": 69865.4765625,
+      "learning_rate": 7.674848112101196e-05,
+      "loss": 0.3415,
+      "step": 1007
+    },
+    {
+      "epoch": 168.0,
+      "grad_norm": 90881.4140625,
+      "learning_rate": 7.597595192178702e-05,
+      "loss": 0.3486,
+      "step": 1008
+    },
+    {
+      "epoch": 168.0,
+      "eval_accuracy": 0.8393613554903877,
+      "eval_best_threshold": 0.5172176361083984,
+      "eval_f1": 0.8541851523217983,
+      "eval_loss": 0.3668026328086853,
+      "eval_pr_auc": 0.927415132113945,
+      "eval_precision": 0.8579916815210933,
+      "eval_recall": 0.850412249705536,
+      "eval_roc_auc": 0.9148575704544499,
+      "eval_runtime": 3.9261,
+      "eval_samples_per_second": 781.695,
+      "eval_steps_per_second": 1.528,
+      "step": 1008
+    },
+    {
+      "epoch": 168.16666666666666,
+      "grad_norm": 106595.0625,
+      "learning_rate": 7.520701063979474e-05,
+      "loss": 0.3614,
+      "step": 1009
+    },
+    {
+      "epoch": 168.33333333333334,
+      "grad_norm": 93386.671875,
+      "learning_rate": 7.444166378150013e-05,
+      "loss": 0.3379,
+      "step": 1010
+    },
+    {
+      "epoch": 168.5,
+      "grad_norm": 45564.28515625,
+      "learning_rate": 7.367991782295391e-05,
+      "loss": 0.3381,
+      "step": 1011
+    },
+    {
+      "epoch": 168.66666666666666,
+      "grad_norm": 68190.890625,
+      "learning_rate": 7.292177920973725e-05,
+      "loss": 0.3379,
+      "step": 1012
+    },
+    {
+      "epoch": 168.83333333333334,
+      "grad_norm": 114296.609375,
+      "learning_rate": 7.216725435690751e-05,
+      "loss": 0.3523,
+      "step": 1013
+    },
+    {
+      "epoch": 169.0,
+      "grad_norm": 118999.3515625,
+      "learning_rate": 7.14163496489439e-05,
+      "loss": 0.3421,
+      "step": 1014
+    },
+    {
+      "epoch": 169.0,
+      "eval_accuracy": 0.8396871945259042,
+      "eval_best_threshold": 0.4463706612586975,
+      "eval_f1": 0.858701895462378,
+      "eval_loss": 0.3626120686531067,
+      "eval_pr_auc": 0.9278995895335498,
+      "eval_precision": 0.8380044843049327,
+      "eval_recall": 0.8804475853945819,
+      "eval_roc_auc": 0.916600084709432,
+      "eval_runtime": 3.9409,
+      "eval_samples_per_second": 778.747,
+      "eval_steps_per_second": 1.522,
+      "step": 1014
+    },
+    {
+      "epoch": 169.16666666666666,
+      "grad_norm": 55655.57421875,
+      "learning_rate": 7.066907143969353e-05,
+      "loss": 0.3158,
+      "step": 1015
+    },
+    {
+      "epoch": 169.33333333333334,
+      "grad_norm": 67258.390625,
+      "learning_rate": 6.992542605231739e-05,
+      "loss": 0.3561,
+      "step": 1016
+    },
+    {
+      "epoch": 169.5,
+      "grad_norm": 63820.6328125,
+      "learning_rate": 6.91854197792371e-05,
+      "loss": 0.3378,
+      "step": 1017
+    },
+    {
+      "epoch": 169.66666666666666,
+      "grad_norm": 63111.68359375,
+      "learning_rate": 6.844905888208181e-05,
+      "loss": 0.3529,
+      "step": 1018
+    },
+    {
+      "epoch": 169.83333333333334,
+      "grad_norm": 98256.8203125,
+      "learning_rate": 6.77163495916348e-05,
+      "loss": 0.3369,
+      "step": 1019
+    },
+    {
+      "epoch": 170.0,
+      "grad_norm": 59946.46484375,
+      "learning_rate": 6.698729810778065e-05,
+      "loss": 0.3437,
+      "step": 1020
+    },
+    {
+      "epoch": 170.0,
+      "eval_accuracy": 0.8380579993483219,
+      "eval_best_threshold": 0.45613446831703186,
+      "eval_f1": 0.8571428571428571,
+      "eval_loss": 0.3628476858139038,
+      "eval_pr_auc": 0.928300039207456,
+      "eval_precision": 0.837170129140932,
+      "eval_recall": 0.8780918727915195,
+      "eval_roc_auc": 0.9168992739559735,
+      "eval_runtime": 3.9447,
+      "eval_samples_per_second": 778.01,
+      "eval_steps_per_second": 1.521,
+      "step": 1020
+    },
+    {
+      "epoch": 170.16666666666666,
+      "grad_norm": 93339.4765625,
+      "learning_rate": 6.626191059945375e-05,
+      "loss": 0.3552,
+      "step": 1021
+    },
+    {
+      "epoch": 170.33333333333334,
+      "grad_norm": 83849.5625,
+      "learning_rate": 6.554019320458493e-05,
+      "loss": 0.3205,
+      "step": 1022
+    },
+    {
+      "epoch": 170.5,
+      "grad_norm": 48642.6015625,
+      "learning_rate": 6.482215203005015e-05,
+      "loss": 0.3427,
+      "step": 1023
+    },
+    {
+      "epoch": 170.66666666666666,
+      "grad_norm": 155210.078125,
+      "learning_rate": 6.410779315161885e-05,
+      "loss": 0.3271,
+      "step": 1024
+    },
+    {
+      "epoch": 170.83333333333334,
+      "grad_norm": 86855.953125,
+      "learning_rate": 6.339712261390212e-05,
+      "loss": 0.3423,
+      "step": 1025
+    },
+    {
+      "epoch": 171.0,
+      "grad_norm": 107046.40625,
+      "learning_rate": 6.269014643030213e-05,
+      "loss": 0.3363,
+      "step": 1026
+    },
+    {
+      "epoch": 171.0,
+      "eval_accuracy": 0.8400130335614207,
+      "eval_best_threshold": 0.5443341135978699,
+      "eval_f1": 0.8557155451072583,
+      "eval_loss": 0.3667296767234802,
+      "eval_pr_auc": 0.9274746367227966,
+      "eval_precision": 0.853958944281525,
+      "eval_recall": 0.8574793875147232,
+      "eval_roc_auc": 0.9158794961077475,
+      "eval_runtime": 3.9867,
+      "eval_samples_per_second": 769.81,
+      "eval_steps_per_second": 1.505,
+      "step": 1026
+    },
+    {
+      "epoch": 171.16666666666666,
+      "grad_norm": 100376.0625,
+      "learning_rate": 6.198687058296071e-05,
+      "loss": 0.3336,
+      "step": 1027
+    },
+    {
+      "epoch": 171.33333333333334,
+      "grad_norm": 150857.78125,
+      "learning_rate": 6.128730102270897e-05,
+      "loss": 0.3689,
+      "step": 1028
+    },
+    {
+      "epoch": 171.5,
+      "grad_norm": 62914.38671875,
+      "learning_rate": 6.059144366901737e-05,
+      "loss": 0.3193,
+      "step": 1029
+    },
+    {
+      "epoch": 171.66666666666666,
+      "grad_norm": 72126.6484375,
+      "learning_rate": 5.989930440994451e-05,
+      "loss": 0.3407,
+      "step": 1030
+    },
+    {
+      "epoch": 171.83333333333334,
+      "grad_norm": 59858.13671875,
+      "learning_rate": 5.921088910208833e-05,
+      "loss": 0.3384,
+      "step": 1031
+    },
+    {
+      "epoch": 172.0,
+      "grad_norm": 107665.28125,
+      "learning_rate": 5.852620357053651e-05,
+      "loss": 0.3605,
+      "step": 1032
+    },
+    {
+      "epoch": 172.0,
+      "eval_accuracy": 0.837406321277289,
+      "eval_best_threshold": 0.42939823865890503,
+      "eval_f1": 0.8546460821438975,
+      "eval_loss": 0.36934569478034973,
+      "eval_pr_auc": 0.9267793551349948,
+      "eval_precision": 0.8455331412103746,
+      "eval_recall": 0.8639575971731449,
+      "eval_roc_auc": 0.9149121247032808,
+      "eval_runtime": 4.0058,
+      "eval_samples_per_second": 766.141,
+      "eval_steps_per_second": 1.498,
+      "step": 1032
+    },
+    {
+      "epoch": 172.16666666666666,
+      "grad_norm": 190367.765625,
+      "learning_rate": 5.78452536088166e-05,
+      "loss": 0.3248,
+      "step": 1033
+    },
+    {
+      "epoch": 172.33333333333334,
+      "grad_norm": 134129.859375,
+      "learning_rate": 5.716804497884698e-05,
+      "loss": 0.3496,
+      "step": 1034
+    },
+    {
+      "epoch": 172.5,
+      "grad_norm": 129405.9609375,
+      "learning_rate": 5.649458341088914e-05,
+      "loss": 0.3654,
+      "step": 1035
+    },
+    {
+      "epoch": 172.66666666666666,
+      "grad_norm": 85490.953125,
+      "learning_rate": 5.5824874603498056e-05,
+      "loss": 0.3546,
+      "step": 1036
+    },
+    {
+      "epoch": 172.83333333333334,
+      "grad_norm": 101800.09375,
+      "learning_rate": 5.515892422347413e-05,
+      "loss": 0.3349,
+      "step": 1037
+    },
+    {
+      "epoch": 173.0,
+      "grad_norm": 102238.9296875,
+      "learning_rate": 5.449673790581611e-05,
+      "loss": 0.3378,
+      "step": 1038
+    },
+    {
+      "epoch": 173.0,
+      "eval_accuracy": 0.8387096774193549,
+      "eval_best_threshold": 0.4776381254196167,
+      "eval_f1": 0.8567293777134588,
+      "eval_loss": 0.36330658197402954,
+      "eval_pr_auc": 0.9276768573910711,
+      "eval_precision": 0.842344906089926,
+      "eval_recall": 0.8716136631330977,
+      "eval_roc_auc": 0.9164031309843219,
+      "eval_runtime": 4.006,
+      "eval_samples_per_second": 766.097,
+      "eval_steps_per_second": 1.498,
+      "step": 1038
+    },
+    {
+      "epoch": 173.16666666666666,
+      "grad_norm": 73306.3046875,
+      "learning_rate": 5.383832125367233e-05,
+      "loss": 0.3338,
+      "step": 1039
+    },
+    {
+      "epoch": 173.33333333333334,
+      "grad_norm": 55874.16015625,
+      "learning_rate": 5.318367983829392e-05,
+      "loss": 0.3408,
+      "step": 1040
+    },
+    {
+      "epoch": 173.5,
+      "grad_norm": 164934.734375,
+      "learning_rate": 5.2532819198987506e-05,
+      "loss": 0.334,
+      "step": 1041
+    },
+    {
+      "epoch": 173.66666666666666,
+      "grad_norm": 159728.6875,
+      "learning_rate": 5.1885744843068295e-05,
+      "loss": 0.3337,
+      "step": 1042
+    },
+    {
+      "epoch": 173.83333333333334,
+      "grad_norm": 125720.4140625,
+      "learning_rate": 5.124246224581375e-05,
+      "loss": 0.3456,
+      "step": 1043
+    },
+    {
+      "epoch": 174.0,
+      "grad_norm": 61294.265625,
+      "learning_rate": 5.060297685041659e-05,
+      "loss": 0.3518,
+      "step": 1044
+    },
+    {
+      "epoch": 174.0,
+      "eval_accuracy": 0.8426197458455523,
+      "eval_best_threshold": 0.4912884533405304,
+      "eval_f1": 0.8591426071741033,
+      "eval_loss": 0.3630110025405884,
+      "eval_pr_auc": 0.9279353248854936,
+      "eval_precision": 0.8509532062391681,
+      "eval_recall": 0.8674911660777385,
+      "eval_roc_auc": 0.9164450131832275,
+      "eval_runtime": 4.055,
+      "eval_samples_per_second": 756.837,
+      "eval_steps_per_second": 1.48,
+      "step": 1044
+    },
+    {
+      "epoch": 174.16666666666666,
+      "grad_norm": 46890.20703125,
+      "learning_rate": 4.996729406793943e-05,
+      "loss": 0.3363,
+      "step": 1045
+    },
+    {
+      "epoch": 174.33333333333334,
+      "grad_norm": 112337.2421875,
+      "learning_rate": 4.933541927726887e-05,
+      "loss": 0.3398,
+      "step": 1046
+    },
+    {
+      "epoch": 174.5,
+      "grad_norm": 91068.9609375,
+      "learning_rate": 4.87073578250698e-05,
+      "loss": 0.3423,
+      "step": 1047
+    },
+    {
+      "epoch": 174.66666666666666,
+      "grad_norm": 97913.828125,
+      "learning_rate": 4.8083115025739754e-05,
+      "loss": 0.3471,
+      "step": 1048
+    },
+    {
+      "epoch": 174.83333333333334,
+      "grad_norm": 63174.71875,
+      "learning_rate": 4.7462696161365006e-05,
+      "loss": 0.329,
+      "step": 1049
+    },
+    {
+      "epoch": 175.0,
+      "grad_norm": 103652.0859375,
+      "learning_rate": 4.684610648167503e-05,
+      "loss": 0.3401,
+      "step": 1050
+    },
+    {
+      "epoch": 175.0,
+      "eval_accuracy": 0.8387096774193549,
+      "eval_best_threshold": 0.43684035539627075,
+      "eval_f1": 0.8576358930112166,
+      "eval_loss": 0.362859845161438,
+      "eval_pr_auc": 0.9280278073336945,
+      "eval_precision": 0.8381112984822934,
+      "eval_recall": 0.8780918727915195,
+      "eval_roc_auc": 0.9166754726674623,
+      "eval_runtime": 4.0499,
+      "eval_samples_per_second": 757.803,
+      "eval_steps_per_second": 1.482,
+      "step": 1050
+    },
+    {
+      "epoch": 175.16666666666666,
+      "grad_norm": 74191.2109375,
+      "learning_rate": 4.6233351203998185e-05,
+      "loss": 0.3401,
+      "step": 1051
+    },
+    {
+      "epoch": 175.33333333333334,
+      "grad_norm": 56000.7421875,
+      "learning_rate": 4.562443551321788e-05,
+      "loss": 0.3453,
+      "step": 1052
+    },
+    {
+      "epoch": 175.5,
+      "grad_norm": 90864.1640625,
+      "learning_rate": 4.501936456172845e-05,
+      "loss": 0.3524,
+      "step": 1053
+    },
+    {
+      "epoch": 175.66666666666666,
+      "grad_norm": 77782.4453125,
+      "learning_rate": 4.441814346939149e-05,
+      "loss": 0.3323,
+      "step": 1054
+    },
+    {
+      "epoch": 175.83333333333334,
+      "grad_norm": 67637.40625,
+      "learning_rate": 4.382077732349299e-05,
+      "loss": 0.3194,
+      "step": 1055
+    },
+    {
+      "epoch": 176.0,
+      "grad_norm": 122488.0390625,
+      "learning_rate": 4.322727117869951e-05,
+      "loss": 0.3407,
+      "step": 1056
+    },
+    {
+      "epoch": 176.0,
+      "eval_accuracy": 0.8422939068100358,
+      "eval_best_threshold": 0.4837534725666046,
+      "eval_f1": 0.8576470588235294,
+      "eval_loss": 0.3607856035232544,
+      "eval_pr_auc": 0.9296632050709258,
+      "eval_precision": 0.8566392479435958,
+      "eval_recall": 0.8586572438162544,
+      "eval_roc_auc": 0.9178060772574076,
+      "eval_runtime": 4.0408,
+      "eval_samples_per_second": 759.5,
+      "eval_steps_per_second": 1.485,
+      "step": 1056
+    },
+    {
+      "epoch": 176.16666666666666,
+      "grad_norm": 102023.53125,
+      "learning_rate": 4.2637630057016487e-05,
+      "loss": 0.3196,
+      "step": 1057
+    },
+    {
+      "epoch": 176.33333333333334,
+      "grad_norm": 90810.1171875,
+      "learning_rate": 4.205185894774455e-05,
+      "loss": 0.3339,
+      "step": 1058
+    },
+    {
+      "epoch": 176.5,
+      "grad_norm": 137295.046875,
+      "learning_rate": 4.1469962807437976e-05,
+      "loss": 0.335,
+      "step": 1059
+    },
+    {
+      "epoch": 176.66666666666666,
+      "grad_norm": 57208.4453125,
+      "learning_rate": 4.0891946559863055e-05,
+      "loss": 0.3573,
+      "step": 1060
+    },
+    {
+      "epoch": 176.83333333333334,
+      "grad_norm": 89965.6640625,
+      "learning_rate": 4.0317815095955446e-05,
+      "loss": 0.3539,
+      "step": 1061
+    },
+    {
+      "epoch": 177.0,
+      "grad_norm": 91107.03125,
+      "learning_rate": 3.974757327377981e-05,
+      "loss": 0.3411,
+      "step": 1062
+    },
+    {
+      "epoch": 177.0,
+      "eval_accuracy": 0.8419680677745194,
+      "eval_best_threshold": 0.46736517548561096,
+      "eval_f1": 0.8590526009880849,
+      "eval_loss": 0.359243780374527,
+      "eval_pr_auc": 0.9296493537780411,
+      "eval_precision": 0.8479632816982214,
+      "eval_recall": 0.8704358068315665,
+      "eval_roc_auc": 0.918315536620506,
+      "eval_runtime": 4.0551,
+      "eval_samples_per_second": 756.821,
+      "eval_steps_per_second": 1.48,
+      "step": 1062
+    },
+    {
+      "epoch": 177.16666666666666,
+      "grad_norm": 60148.11328125,
+      "learning_rate": 3.91812259184881e-05,
+      "loss": 0.3518,
+      "step": 1063
+    },
+    {
+      "epoch": 177.33333333333334,
+      "grad_norm": 69379.609375,
+      "learning_rate": 3.861877782227885e-05,
+      "loss": 0.3265,
+      "step": 1064
+    },
+    {
+      "epoch": 177.5,
+      "grad_norm": 101317.9375,
+      "learning_rate": 3.806023374435663e-05,
+      "loss": 0.3383,
+      "step": 1065
+    },
+    {
+      "epoch": 177.66666666666666,
+      "grad_norm": 55583.95703125,
+      "learning_rate": 3.750559841089196e-05,
+      "loss": 0.3113,
+      "step": 1066
+    },
+    {
+      "epoch": 177.83333333333334,
+      "grad_norm": 91411.53125,
+      "learning_rate": 3.6954876514981085e-05,
+      "loss": 0.3464,
+      "step": 1067
+    },
+    {
+      "epoch": 178.0,
+      "grad_norm": 61728.08984375,
+      "learning_rate": 3.6408072716606344e-05,
+      "loss": 0.3551,
+      "step": 1068
+    },
+    {
+      "epoch": 178.0,
+      "eval_accuracy": 0.8442489410231345,
+      "eval_best_threshold": 0.4724968671798706,
+      "eval_f1": 0.8594117647058823,
+      "eval_loss": 0.3610427677631378,
+      "eval_pr_auc": 0.9291781268367707,
+      "eval_precision": 0.8584018801410106,
+      "eval_recall": 0.8604240282685512,
+      "eval_roc_auc": 0.9179811233707824,
+      "eval_runtime": 4.0883,
+      "eval_samples_per_second": 750.681,
+      "eval_steps_per_second": 1.468,
+      "step": 1068
+    },
+    {
+      "epoch": 178.16666666666666,
+      "grad_norm": 54876.40234375,
+      "learning_rate": 3.586519164259672e-05,
+      "loss": 0.3271,
+      "step": 1069
+    },
+    {
+      "epoch": 178.33333333333334,
+      "grad_norm": 151829.515625,
+      "learning_rate": 3.532623788658873e-05,
+      "loss": 0.3351,
+      "step": 1070
+    },
+    {
+      "epoch": 178.5,
+      "grad_norm": 183532.984375,
+      "learning_rate": 3.479121600898777e-05,
+      "loss": 0.3301,
+      "step": 1071
+    },
+    {
+      "epoch": 178.66666666666666,
+      "grad_norm": 71134.21875,
+      "learning_rate": 3.426013053692878e-05,
+      "loss": 0.3312,
+      "step": 1072
+    },
+    {
+      "epoch": 178.83333333333334,
+      "grad_norm": 65962.1484375,
+      "learning_rate": 3.373298596423901e-05,
+      "loss": 0.3306,
+      "step": 1073
+    },
+    {
+      "epoch": 179.0,
+      "grad_norm": 73190.765625,
+      "learning_rate": 3.3209786751399184e-05,
+      "loss": 0.3645,
+      "step": 1074
+    },
+    {
+      "epoch": 179.0,
+      "eval_accuracy": 0.844574780058651,
+      "eval_best_threshold": 0.5492097735404968,
+      "eval_f1": 0.8584989617324236,
+      "eval_loss": 0.35855212807655334,
+      "eval_pr_auc": 0.9293997822677108,
+      "eval_precision": 0.8649133293484758,
+      "eval_recall": 0.8521790341578327,
+      "eval_roc_auc": 0.9187532592941969,
+      "eval_runtime": 4.0681,
+      "eval_samples_per_second": 754.413,
+      "eval_steps_per_second": 1.475,
+      "step": 1074
+    },
+    {
+      "epoch": 179.16666666666666,
+      "grad_norm": 118103.890625,
+      "learning_rate": 3.2690537325505806e-05,
+      "loss": 0.3578,
+      "step": 1075
+    },
+    {
+      "epoch": 179.33333333333334,
+      "grad_norm": 89901.9375,
+      "learning_rate": 3.217524208023431e-05,
+      "loss": 0.3531,
+      "step": 1076
+    },
+    {
+      "epoch": 179.5,
+      "grad_norm": 116530.1328125,
+      "learning_rate": 3.166390537580122e-05,
+      "loss": 0.3251,
+      "step": 1077
+    },
+    {
+      "epoch": 179.66666666666666,
+      "grad_norm": 64210.16015625,
+      "learning_rate": 3.115653153892761e-05,
+      "loss": 0.3309,
+      "step": 1078
+    },
+    {
+      "epoch": 179.83333333333334,
+      "grad_norm": 115194.0625,
+      "learning_rate": 3.065312486280231e-05,
+      "loss": 0.325,
+      "step": 1079
+    },
+    {
+      "epoch": 180.0,
+      "grad_norm": 74724.2890625,
+      "learning_rate": 3.0153689607045842e-05,
+      "loss": 0.3515,
+      "step": 1080
+    },
+    {
+      "epoch": 180.0,
+      "eval_accuracy": 0.8406647116324536,
+      "eval_best_threshold": 0.44527578353881836,
+      "eval_f1": 0.8587929540860526,
+      "eval_loss": 0.35788804292678833,
+      "eval_pr_auc": 0.929943212647981,
+      "eval_precision": 0.8424929178470255,
+      "eval_recall": 0.875736160188457,
+      "eval_roc_auc": 0.9188731068172191,
+      "eval_runtime": 4.1403,
+      "eval_samples_per_second": 741.258,
+      "eval_steps_per_second": 1.449,
+      "step": 1080
+    },
+    {
+      "epoch": 180.16666666666666,
+      "grad_norm": 91110.6875,
+      "learning_rate": 2.965822999767398e-05,
+      "loss": 0.3468,
+      "step": 1081
+    },
+    {
+      "epoch": 180.33333333333334,
+      "grad_norm": 86685.6640625,
+      "learning_rate": 2.9166750227062387e-05,
+      "loss": 0.3374,
+      "step": 1082
+    },
+    {
+      "epoch": 180.5,
+      "grad_norm": 121932.2734375,
+      "learning_rate": 2.8679254453910786e-05,
+      "loss": 0.346,
+      "step": 1083
+    },
+    {
+      "epoch": 180.66666666666666,
+      "grad_norm": 108420.78125,
+      "learning_rate": 2.8195746803208243e-05,
+      "loss": 0.3369,
+      "step": 1084
+    },
+    {
+      "epoch": 180.83333333333334,
+      "grad_norm": 85976.6953125,
+      "learning_rate": 2.7716231366197665e-05,
+      "loss": 0.3223,
+      "step": 1085
+    },
+    {
+      "epoch": 181.0,
+      "grad_norm": 77798.3046875,
+      "learning_rate": 2.724071220034158e-05,
+      "loss": 0.3413,
+      "step": 1086
+    },
+    {
+      "epoch": 181.0,
+      "eval_accuracy": 0.8426197458455523,
+      "eval_best_threshold": 0.46771353483200073,
+      "eval_f1": 0.8588132125109617,
+      "eval_loss": 0.36123743653297424,
+      "eval_pr_auc": 0.9284586856058963,
+      "eval_precision": 0.8525827045850262,
+      "eval_recall": 0.8651354534746761,
+      "eval_roc_auc": 0.9173434400448806,
+      "eval_runtime": 4.1205,
+      "eval_samples_per_second": 744.82,
+      "eval_steps_per_second": 1.456,
+      "step": 1086
+    },
+    {
+      "epoch": 181.16666666666666,
+      "grad_norm": 69239.671875,
+      "learning_rate": 2.6769193329287845e-05,
+      "loss": 0.34,
+      "step": 1087
+    },
+    {
+      "epoch": 181.33333333333334,
+      "grad_norm": 67383.75,
+      "learning_rate": 2.63016787428354e-05,
+      "loss": 0.3458,
+      "step": 1088
+    },
+    {
+      "epoch": 181.5,
+      "grad_norm": 49835.70703125,
+      "learning_rate": 2.583817239690034e-05,
+      "loss": 0.3425,
+      "step": 1089
+    },
+    {
+      "epoch": 181.66666666666666,
+      "grad_norm": 92561.109375,
+      "learning_rate": 2.537867821348305e-05,
+      "loss": 0.342,
+      "step": 1090
+    },
+    {
+      "epoch": 181.83333333333334,
+      "grad_norm": 65791.2734375,
+      "learning_rate": 2.4923200080634422e-05,
+      "loss": 0.3283,
+      "step": 1091
+    },
+    {
+      "epoch": 182.0,
+      "grad_norm": 90121.1484375,
+      "learning_rate": 2.4471741852423235e-05,
+      "loss": 0.3357,
+      "step": 1092
+    },
+    {
+      "epoch": 182.0,
+      "eval_accuracy": 0.8419680677745194,
+      "eval_best_threshold": 0.451703816652298,
+      "eval_f1": 0.8601902565580859,
+      "eval_loss": 0.3588697612285614,
+      "eval_pr_auc": 0.9288007303907959,
+      "eval_precision": 0.8424618859401468,
+      "eval_recall": 0.8786808009422851,
+      "eval_roc_auc": 0.9184053148725191,
+      "eval_runtime": 4.1365,
+      "eval_samples_per_second": 741.932,
+      "eval_steps_per_second": 1.451,
+      "step": 1092
+    },
+    {
+      "epoch": 182.16666666666666,
+      "grad_norm": 58866.40234375,
+      "learning_rate": 2.402430734890343e-05,
+      "loss": 0.3147,
+      "step": 1093
+    },
+    {
+      "epoch": 182.33333333333334,
+      "grad_norm": 93942.328125,
+      "learning_rate": 2.3580900356081904e-05,
+      "loss": 0.3327,
+      "step": 1094
+    },
+    {
+      "epoch": 182.5,
+      "grad_norm": 83109.65625,
+      "learning_rate": 2.314152462588659e-05,
+      "loss": 0.3356,
+      "step": 1095
+    },
+    {
+      "epoch": 182.66666666666666,
+      "grad_norm": 105172.3359375,
+      "learning_rate": 2.2706183876134045e-05,
+      "loss": 0.3514,
+      "step": 1096
+    },
+    {
+      "epoch": 182.83333333333334,
+      "grad_norm": 65252.1484375,
+      "learning_rate": 2.2274881790498913e-05,
+      "loss": 0.3422,
+      "step": 1097
+    },
+    {
+      "epoch": 183.0,
+      "grad_norm": 58000.1953125,
+      "learning_rate": 2.1847622018482283e-05,
+      "loss": 0.3336,
+      "step": 1098
+    },
+    {
+      "epoch": 183.0,
+      "eval_accuracy": 0.8422939068100358,
+      "eval_best_threshold": 0.4615512788295746,
+      "eval_f1": 0.8593023255813953,
+      "eval_loss": 0.35933783650398254,
+      "eval_pr_auc": 0.928577987507547,
+      "eval_precision": 0.8484500574052812,
+      "eval_recall": 0.8704358068315665,
+      "eval_roc_auc": 0.9181961186584982,
+      "eval_runtime": 4.1192,
+      "eval_samples_per_second": 745.048,
+      "eval_steps_per_second": 1.457,
+      "step": 1098
+    },
+    {
+      "epoch": 183.16666666666666,
+      "grad_norm": 75021.8828125,
+      "learning_rate": 2.142440817538066e-05,
+      "loss": 0.3462,
+      "step": 1099
+    },
+    {
+      "epoch": 183.33333333333334,
+      "grad_norm": 108155.3671875,
+      "learning_rate": 2.100524384225555e-05,
+      "loss": 0.3532,
+      "step": 1100
+    },
+    {
+      "epoch": 183.5,
+      "grad_norm": 60136.75,
+      "learning_rate": 2.0590132565903473e-05,
+      "loss": 0.3307,
+      "step": 1101
+    },
+    {
+      "epoch": 183.66666666666666,
+      "grad_norm": 64039.2109375,
+      "learning_rate": 2.0179077858825447e-05,
+      "loss": 0.326,
+      "step": 1102
+    },
+    {
+      "epoch": 183.83333333333334,
+      "grad_norm": 85661.9375,
+      "learning_rate": 1.977208319919721e-05,
+      "loss": 0.3576,
+      "step": 1103
+    },
+    {
+      "epoch": 184.0,
+      "grad_norm": 97985.703125,
+      "learning_rate": 1.9369152030840554e-05,
+      "loss": 0.3204,
+      "step": 1104
+    },
+    {
+      "epoch": 184.0,
+      "eval_accuracy": 0.8419680677745194,
+      "eval_best_threshold": 0.46132898330688477,
+      "eval_f1": 0.8590526009880849,
+      "eval_loss": 0.35893210768699646,
+      "eval_pr_auc": 0.9290151696778128,
+      "eval_precision": 0.8479632816982214,
+      "eval_recall": 0.8704358068315665,
+      "eval_roc_auc": 0.9183816890167262,
+      "eval_runtime": 4.2619,
+      "eval_samples_per_second": 720.096,
+      "eval_steps_per_second": 1.408,
+      "step": 1104
+    },
+    {
+      "epoch": 184.16666666666666,
+      "grad_norm": 63854.22265625,
+      "learning_rate": 1.8970287763193428e-05,
+      "loss": 0.3424,
+      "step": 1105
+    },
+    {
+      "epoch": 184.33333333333334,
+      "grad_norm": 61403.140625,
+      "learning_rate": 1.8575493771281205e-05,
+      "loss": 0.3312,
+      "step": 1106
+    },
+    {
+      "epoch": 184.5,
+      "grad_norm": 86111.2578125,
+      "learning_rate": 1.8184773395688526e-05,
+      "loss": 0.343,
+      "step": 1107
+    },
+    {
+      "epoch": 184.66666666666666,
+      "grad_norm": 122857.234375,
+      "learning_rate": 1.779812994253055e-05,
+      "loss": 0.346,
+      "step": 1108
+    },
+    {
+      "epoch": 184.83333333333334,
+      "grad_norm": 77238.8671875,
+      "learning_rate": 1.741556668342531e-05,
+      "loss": 0.3356,
+      "step": 1109
+    },
+    {
+      "epoch": 185.0,
+      "grad_norm": 85184.375,
+      "learning_rate": 1.70370868554659e-05,
+      "loss": 0.3307,
+      "step": 1110
+    },
+    {
+      "epoch": 185.0,
+      "eval_accuracy": 0.8439231019876181,
+      "eval_best_threshold": 0.5097879767417908,
+      "eval_f1": 0.858242083456644,
+      "eval_loss": 0.35859715938568115,
+      "eval_pr_auc": 0.9290756714382906,
+      "eval_precision": 0.8625817965496728,
+      "eval_recall": 0.8539458186101295,
+      "eval_roc_auc": 0.918589596547704,
+      "eval_runtime": 4.1985,
+      "eval_samples_per_second": 730.979,
+      "eval_steps_per_second": 1.429,
+      "step": 1110
+    },
+    {
+      "epoch": 185.16666666666666,
+      "grad_norm": 75691.6640625,
+      "learning_rate": 1.6662693661192997e-05,
+      "loss": 0.3324,
+      "step": 1111
+    },
+    {
+      "epoch": 185.33333333333334,
+      "grad_norm": 50374.91796875,
+      "learning_rate": 1.6292390268568102e-05,
+      "loss": 0.3305,
+      "step": 1112
+    },
+    {
+      "epoch": 185.5,
+      "grad_norm": 61129.52734375,
+      "learning_rate": 1.5926179810946185e-05,
+      "loss": 0.3259,
+      "step": 1113
+    },
+    {
+      "epoch": 185.66666666666666,
+      "grad_norm": 71613.9453125,
+      "learning_rate": 1.556406538704963e-05,
+      "loss": 0.3314,
+      "step": 1114
+    },
+    {
+      "epoch": 185.83333333333334,
+      "grad_norm": 62662.45703125,
+      "learning_rate": 1.5206050060942001e-05,
+      "loss": 0.3318,
+      "step": 1115
+    },
+    {
+      "epoch": 186.0,
+      "grad_norm": 68081.0078125,
+      "learning_rate": 1.4852136862001764e-05,
+      "loss": 0.347,
+      "step": 1116
+    },
+    {
+      "epoch": 186.0,
+      "eval_accuracy": 0.844574780058651,
+      "eval_best_threshold": 0.5024138689041138,
+      "eval_f1": 0.858833974548683,
+      "eval_loss": 0.35903245210647583,
+      "eval_pr_auc": 0.9289612048718296,
+      "eval_precision": 0.8631766805472932,
+      "eval_recall": 0.8545347467608951,
+      "eval_roc_auc": 0.9184444049248311,
+      "eval_runtime": 4.2396,
+      "eval_samples_per_second": 723.888,
+      "eval_steps_per_second": 1.415,
+      "step": 1116
+    },
+    {
+      "epoch": 186.16666666666666,
+      "grad_norm": 70388.890625,
+      "learning_rate": 1.450232878489699e-05,
+      "loss": 0.3416,
+      "step": 1117
+    },
+    {
+      "epoch": 186.33333333333334,
+      "grad_norm": 61398.140625,
+      "learning_rate": 1.4156628789559923e-05,
+      "loss": 0.3336,
+      "step": 1118
+    },
+    {
+      "epoch": 186.5,
+      "grad_norm": 92423.7109375,
+      "learning_rate": 1.3815039801161721e-05,
+      "loss": 0.34,
+      "step": 1119
+    },
+    {
+      "epoch": 186.66666666666666,
+      "grad_norm": 96490.03125,
+      "learning_rate": 1.3477564710088097e-05,
+      "loss": 0.3446,
+      "step": 1120
+    },
+    {
+      "epoch": 186.83333333333334,
+      "grad_norm": 83160.28125,
+      "learning_rate": 1.3144206371914601e-05,
+      "loss": 0.3277,
+      "step": 1121
+    },
+    {
+      "epoch": 187.0,
+      "grad_norm": 100917.78125,
+      "learning_rate": 1.2814967607382432e-05,
+      "loss": 0.3291,
+      "step": 1122
+    },
+    {
+      "epoch": 187.0,
+      "eval_accuracy": 0.8429455848810687,
+      "eval_best_threshold": 0.45496270060539246,
+      "eval_f1": 0.8600464576074333,
+      "eval_loss": 0.35840901732444763,
+      "eval_pr_auc": 0.9294939960565185,
+      "eval_precision": 0.8482245131729668,
+      "eval_recall": 0.8722025912838633,
+      "eval_roc_auc": 0.9187990075422323,
+      "eval_runtime": 4.8699,
+      "eval_samples_per_second": 630.198,
+      "eval_steps_per_second": 1.232,
+      "step": 1122
+    },
+    {
+      "epoch": 187.16666666666666,
+      "grad_norm": 71137.640625,
+      "learning_rate": 1.2489851202374724e-05,
+      "loss": 0.3169,
+      "step": 1123
+    },
+    {
+      "epoch": 187.33333333333334,
+      "grad_norm": 117276.4453125,
+      "learning_rate": 1.2168859907892904e-05,
+      "loss": 0.3514,
+      "step": 1124
+    },
+    {
+      "epoch": 187.5,
+      "grad_norm": 114218.09375,
+      "learning_rate": 1.185199644003332e-05,
+      "loss": 0.3436,
+      "step": 1125
+    },
+    {
+      "epoch": 187.66666666666666,
+      "grad_norm": 97303.96875,
+      "learning_rate": 1.1539263479964535e-05,
+      "loss": 0.3552,
+      "step": 1126
+    },
+    {
+      "epoch": 187.83333333333334,
+      "grad_norm": 80702.0546875,
+      "learning_rate": 1.123066367390424e-05,
+      "loss": 0.323,
+      "step": 1127
+    },
+    {
+      "epoch": 188.0,
+      "grad_norm": 59036.5390625,
+      "learning_rate": 1.0926199633097156e-05,
+      "loss": 0.3381,
+      "step": 1128
+    },
+    {
+      "epoch": 188.0,
+      "eval_accuracy": 0.8442489410231345,
+      "eval_best_threshold": 0.49203333258628845,
+      "eval_f1": 0.8590801886792453,
+      "eval_loss": 0.3574254512786865,
+      "eval_pr_auc": 0.9295933422197652,
+      "eval_precision": 0.8600944510035419,
+      "eval_recall": 0.8580683156654888,
+      "eval_roc_auc": 0.9192786124148288,
+      "eval_runtime": 4.3706,
+      "eval_samples_per_second": 702.193,
+      "eval_steps_per_second": 1.373,
+      "step": 1128
+    },
+    {
+      "epoch": 188.16666666666666,
+      "grad_norm": 58683.0,
+      "learning_rate": 1.0625873933792996e-05,
+      "loss": 0.3391,
+      "step": 1129
+    },
+    {
+      "epoch": 188.33333333333334,
+      "grad_norm": 51025.08203125,
+      "learning_rate": 1.0329689117224261e-05,
+      "loss": 0.3307,
+      "step": 1130
+    },
+    {
+      "epoch": 188.5,
+      "grad_norm": 95173.8984375,
+      "learning_rate": 1.0037647689585205e-05,
+      "loss": 0.3309,
+      "step": 1131
+    },
+    {
+      "epoch": 188.66666666666666,
+      "grad_norm": 123481.3359375,
+      "learning_rate": 9.749752122010347e-06,
+      "loss": 0.337,
+      "step": 1132
+    },
+    {
+      "epoch": 188.83333333333334,
+      "grad_norm": 106579.6953125,
+      "learning_rate": 9.46600485055371e-06,
+      "loss": 0.3498,
+      "step": 1133
+    },
+    {
+      "epoch": 189.0,
+      "grad_norm": 87096.609375,
+      "learning_rate": 9.186408276168012e-06,
+      "loss": 0.3346,
+      "step": 1134
+    },
+    {
+      "epoch": 189.0,
+      "eval_accuracy": 0.8419680677745194,
+      "eval_best_threshold": 0.4620799422264099,
+      "eval_f1": 0.859297940237888,
+      "eval_loss": 0.35710543394088745,
+      "eval_pr_auc": 0.9296763020932254,
+      "eval_precision": 0.8467695826186392,
+      "eval_recall": 0.8722025912838633,
+      "eval_roc_auc": 0.9193872913514762,
+      "eval_runtime": 4.154,
+      "eval_samples_per_second": 738.803,
+      "eval_steps_per_second": 1.444,
+      "step": 1134
+    },
+    {
+      "epoch": 189.16666666666666,
+      "grad_norm": 67883.25,
+      "learning_rate": 8.91096476468467e-06,
+      "loss": 0.3172,
+      "step": 1135
+    },
+    {
+      "epoch": 189.33333333333334,
+      "grad_norm": 60549.61328125,
+      "learning_rate": 8.639676646793382e-06,
+      "loss": 0.3289,
+      "step": 1136
+    },
+    {
+      "epoch": 189.5,
+      "grad_norm": 60352.7890625,
+      "learning_rate": 8.372546218022748e-06,
+      "loss": 0.3512,
+      "step": 1137
+    },
+    {
+      "epoch": 189.66666666666666,
+      "grad_norm": 86886.8359375,
+      "learning_rate": 8.109575738720621e-06,
+      "loss": 0.3293,
+      "step": 1138
+    },
+    {
+      "epoch": 189.83333333333334,
+      "grad_norm": 56512.19140625,
+      "learning_rate": 7.850767434035177e-06,
+      "loss": 0.3434,
+      "step": 1139
+    },
+    {
+      "epoch": 190.0,
+      "grad_norm": 88472.5625,
+      "learning_rate": 7.59612349389599e-06,
+      "loss": 0.3397,
+      "step": 1140
+    },
+    {
+      "epoch": 190.0,
+      "eval_accuracy": 0.8413163897034864,
+      "eval_best_threshold": 0.47650080919265747,
+      "eval_f1": 0.8582241630276565,
+      "eval_loss": 0.3567776381969452,
+      "eval_pr_auc": 0.9300562636053985,
+      "eval_precision": 0.8485895221646517,
+      "eval_recall": 0.8680800942285041,
+      "eval_roc_auc": 0.9194770696034894,
+      "eval_runtime": 4.4139,
+      "eval_samples_per_second": 695.303,
+      "eval_steps_per_second": 1.359,
+      "step": 1140
+    },
+    {
+      "epoch": 190.16666666666666,
+      "grad_norm": 80563.6796875,
+      "learning_rate": 7.345646072995315e-06,
+      "loss": 0.3177,
+      "step": 1141
+    },
+    {
+      "epoch": 190.33333333333334,
+      "grad_norm": 52460.4921875,
+      "learning_rate": 7.099337290770169e-06,
+      "loss": 0.3426,
+      "step": 1142
+    },
+    {
+      "epoch": 190.5,
+      "grad_norm": 60113.8359375,
+      "learning_rate": 6.8571992313842815e-06,
+      "loss": 0.3162,
+      "step": 1143
+    },
+    {
+      "epoch": 190.66666666666666,
+      "grad_norm": 86581.2109375,
+      "learning_rate": 6.61923394371039e-06,
+      "loss": 0.3342,
+      "step": 1144
+    },
+    {
+      "epoch": 190.83333333333334,
+      "grad_norm": 90135.9375,
+      "learning_rate": 6.385443441312977e-06,
+      "loss": 0.3307,
+      "step": 1145
+    },
+    {
+      "epoch": 191.0,
+      "grad_norm": 75371.625,
+      "learning_rate": 6.15582970243117e-06,
+      "loss": 0.3468,
+      "step": 1146
+    },
+    {
+      "epoch": 191.0,
+      "eval_accuracy": 0.8426197458455523,
+      "eval_best_threshold": 0.46230238676071167,
+      "eval_f1": 0.8587306229891781,
+      "eval_loss": 0.3580075204372406,
+      "eval_pr_auc": 0.9295473297541057,
+      "eval_precision": 0.852992446252179,
+      "eval_recall": 0.8645465253239105,
+      "eval_roc_auc": 0.9190067002927028,
+      "eval_runtime": 4.3965,
+      "eval_samples_per_second": 698.062,
+      "eval_steps_per_second": 1.365,
+      "step": 1146
+    },
+    {
+      "epoch": 191.16666666666666,
+      "grad_norm": 66957.3984375,
+      "learning_rate": 5.9303946699620355e-06,
+      "loss": 0.3281,
+      "step": 1147
+    },
+    {
+      "epoch": 191.33333333333334,
+      "grad_norm": 106900.6796875,
+      "learning_rate": 5.709140251444201e-06,
+      "loss": 0.3299,
+      "step": 1148
+    },
+    {
+      "epoch": 191.5,
+      "grad_norm": 69170.3203125,
+      "learning_rate": 5.492068319041588e-06,
+      "loss": 0.3284,
+      "step": 1149
+    },
+    {
+      "epoch": 191.66666666666666,
+      "grad_norm": 85578.3359375,
+      "learning_rate": 5.279180709527765e-06,
+      "loss": 0.3441,
+      "step": 1150
+    },
+    {
+      "epoch": 191.83333333333334,
+      "grad_norm": 90991.1484375,
+      "learning_rate": 5.070479224270119e-06,
+      "loss": 0.3408,
+      "step": 1151
+    },
+    {
+      "epoch": 192.0,
+      "grad_norm": 80094.2109375,
+      "learning_rate": 4.865965629214819e-06,
+      "loss": 0.3497,
+      "step": 1152
+    },
+    {
+      "epoch": 192.0,
+      "eval_accuracy": 0.8432714239165852,
+      "eval_best_threshold": 0.486149400472641,
+      "eval_f1": 0.8591508052708638,
+      "eval_loss": 0.3572799861431122,
+      "eval_pr_auc": 0.9297294809940666,
+      "eval_precision": 0.8543972044263249,
+      "eval_recall": 0.8639575971731449,
+      "eval_roc_auc": 0.9192713098775837,
+      "eval_runtime": 4.4784,
+      "eval_samples_per_second": 685.294,
+      "eval_steps_per_second": 1.34,
+      "step": 1152
+    },
+    {
+      "epoch": 192.16666666666666,
+      "grad_norm": 50093.2265625,
+      "learning_rate": 4.665641654871989e-06,
+      "loss": 0.323,
+      "step": 1153
+    },
+    {
+      "epoch": 192.33333333333334,
+      "grad_norm": 116718.796875,
+      "learning_rate": 4.469508996300664e-06,
+      "loss": 0.3508,
+      "step": 1154
+    },
+    {
+      "epoch": 192.5,
+      "grad_norm": 53607.64453125,
+      "learning_rate": 4.277569313094809e-06,
+      "loss": 0.3174,
+      "step": 1155
+    },
+    {
+      "epoch": 192.66666666666666,
+      "grad_norm": 54493.2421875,
+      "learning_rate": 4.089824229369155e-06,
+      "loss": 0.3451,
+      "step": 1156
+    },
+    {
+      "epoch": 192.83333333333334,
+      "grad_norm": 82322.3359375,
+      "learning_rate": 3.906275333745435e-06,
+      "loss": 0.3356,
+      "step": 1157
+    },
+    {
+      "epoch": 193.0,
+      "grad_norm": 93312.859375,
+      "learning_rate": 3.7269241793390084e-06,
+      "loss": 0.3567,
+      "step": 1158
+    },
+    {
+      "epoch": 193.0,
+      "eval_accuracy": 0.8432714239165852,
+      "eval_best_threshold": 0.4834613502025604,
+      "eval_f1": 0.8591508052708638,
+      "eval_loss": 0.35751256346702576,
+      "eval_pr_auc": 0.9295732860994348,
+      "eval_precision": 0.8543972044263249,
+      "eval_recall": 0.8639575971731449,
+      "eval_roc_auc": 0.9191643491849939,
+      "eval_runtime": 4.2553,
+      "eval_samples_per_second": 721.211,
+      "eval_steps_per_second": 1.41,
+      "step": 1158
+    },
+    {
+      "epoch": 193.16666666666666,
+      "grad_norm": 53519.34765625,
+      "learning_rate": 3.551772283745536e-06,
+      "loss": 0.3178,
+      "step": 1159
+    },
+    {
+      "epoch": 193.33333333333334,
+      "grad_norm": 48680.87109375,
+      "learning_rate": 3.3808211290284885e-06,
+      "loss": 0.327,
+      "step": 1160
+    },
+    {
+      "epoch": 193.5,
+      "grad_norm": 65628.5859375,
+      "learning_rate": 3.2140721617062717e-06,
+      "loss": 0.348,
+      "step": 1161
+    },
+    {
+      "epoch": 193.66666666666666,
+      "grad_norm": 87672.9296875,
+      "learning_rate": 3.0515267927400113e-06,
+      "loss": 0.3446,
+      "step": 1162
+    },
+    {
+      "epoch": 193.83333333333334,
+      "grad_norm": 95234.8671875,
+      "learning_rate": 2.8931863975218966e-06,
+      "loss": 0.3429,
+      "step": 1163
+    },
+    {
+      "epoch": 194.0,
+      "grad_norm": 61562.625,
+      "learning_rate": 2.739052315863355e-06,
+      "loss": 0.3158,
+      "step": 1164
+    },
+    {
+      "epoch": 194.0,
+      "eval_accuracy": 0.8432714239165852,
+      "eval_best_threshold": 0.4742875397205353,
+      "eval_f1": 0.8593155893536122,
+      "eval_loss": 0.3577769100666046,
+      "eval_pr_auc": 0.9295315138035642,
+      "eval_precision": 0.8535735037768739,
+      "eval_recall": 0.8651354534746761,
+      "eval_roc_auc": 0.919044931222986,
+      "eval_runtime": 4.2171,
+      "eval_samples_per_second": 727.752,
+      "eval_steps_per_second": 1.423,
+      "step": 1164
+    },
+    {
+      "epoch": 194.16666666666666,
+      "grad_norm": 65050.80859375,
+      "learning_rate": 2.589125851983509e-06,
+      "loss": 0.3344,
+      "step": 1165
+    },
+    {
+      "epoch": 194.33333333333334,
+      "grad_norm": 55117.078125,
+      "learning_rate": 2.4434082744984598e-06,
+      "loss": 0.3308,
+      "step": 1166
+    },
+    {
+      "epoch": 194.5,
+      "grad_norm": 97287.5,
+      "learning_rate": 2.3019008164105737e-06,
+      "loss": 0.3409,
+      "step": 1167
+    },
+    {
+      "epoch": 194.66666666666666,
+      "grad_norm": 59091.03125,
+      "learning_rate": 2.1646046750978256e-06,
+      "loss": 0.3445,
+      "step": 1168
+    },
+    {
+      "epoch": 194.83333333333334,
+      "grad_norm": 52100.3828125,
+      "learning_rate": 2.0315210123035833e-06,
+      "loss": 0.3412,
+      "step": 1169
+    },
+    {
+      "epoch": 195.0,
+      "grad_norm": 89186.1171875,
+      "learning_rate": 1.9026509541272275e-06,
+      "loss": 0.3369,
+      "step": 1170
+    },
+    {
+      "epoch": 195.0,
+      "eval_accuracy": 0.8442489410231345,
+      "eval_best_threshold": 0.48313555121421814,
+      "eval_f1": 0.8595769682726204,
+      "eval_loss": 0.35788410902023315,
+      "eval_pr_auc": 0.9295530985016898,
+      "eval_precision": 0.8575615474794842,
+      "eval_recall": 0.8616018845700825,
+      "eval_roc_auc": 0.919023453172265,
+      "eval_runtime": 4.2448,
+      "eval_samples_per_second": 723.009,
+      "eval_steps_per_second": 1.414,
+      "step": 1170
+    },
+    {
+      "epoch": 195.16666666666666,
+      "grad_norm": 74249.125,
+      "learning_rate": 1.7779955910142142e-06,
+      "loss": 0.3233,
+      "step": 1171
+    },
+    {
+      "epoch": 195.33333333333334,
+      "grad_norm": 83374.2265625,
+      "learning_rate": 1.6575559777469718e-06,
+      "loss": 0.3331,
+      "step": 1172
+    },
+    {
+      "epoch": 195.5,
+      "grad_norm": 59120.72265625,
+      "learning_rate": 1.541333133436018e-06,
+      "loss": 0.3472,
+      "step": 1173
+    },
+    {
+      "epoch": 195.66666666666666,
+      "grad_norm": 64595.88671875,
+      "learning_rate": 1.429328041511302e-06,
+      "loss": 0.3523,
+      "step": 1174
+    },
+    {
+      "epoch": 195.83333333333334,
+      "grad_norm": 77457.6640625,
+      "learning_rate": 1.3215416497138755e-06,
+      "loss": 0.3406,
+      "step": 1175
+    },
+    {
+      "epoch": 196.0,
+      "grad_norm": 60436.64453125,
+      "learning_rate": 1.2179748700879012e-06,
+      "loss": 0.3301,
+      "step": 1176
+    },
+    {
+      "epoch": 196.0,
+      "eval_accuracy": 0.8449006190941675,
+      "eval_best_threshold": 0.4860289394855499,
+      "eval_f1": 0.8602466236054023,
+      "eval_loss": 0.35769322514533997,
+      "eval_pr_auc": 0.9295719105862945,
+      "eval_precision": 0.857728337236534,
+      "eval_recall": 0.8627797408716137,
+      "eval_roc_auc": 0.9190900351294997,
+      "eval_runtime": 4.5203,
+      "eval_samples_per_second": 678.935,
+      "eval_steps_per_second": 1.327,
+      "step": 1176
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 1200,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 200,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 20,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 0
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.880234183022346e+17,
+  "train_batch_size": 1024,
+  "trial_name": null,
+  "trial_params": null
+}