diff --git "a/tmp-checkpoint-40467/trainer_state.json" "b/tmp-checkpoint-40467/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/tmp-checkpoint-40467/trainer_state.json"
@@ -0,0 +1,24717 @@
+{
+  "best_metric": 0.9808588624000549,
+  "best_model_checkpoint": "swinv2-base-patch4-window12-192-22k-ConcreteClassifier-PVT\\checkpoint-1927",
+  "epoch": 21.0,
+  "eval_steps": 500,
+  "global_step": 40467,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7298045320878742e-06,
+      "loss": 2.3066,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4596090641757485e-06,
+      "loss": 2.131,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.1894135962636225e-06,
+      "loss": 2.0104,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.919218128351497e-06,
+      "loss": 1.8938,
+      "step": 40
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.649022660439371e-06,
+      "loss": 1.7743,
+      "step": 50
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0378827192527245e-05,
+      "loss": 1.73,
+      "step": 60
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2108631724615119e-05,
+      "loss": 1.6552,
+      "step": 70
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3838436256702994e-05,
+      "loss": 1.4119,
+      "step": 80
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5568240788790867e-05,
+      "loss": 1.1716,
+      "step": 90
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7298045320878743e-05,
+      "loss": 1.2431,
+      "step": 100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9027849852966615e-05,
+      "loss": 1.1115,
+      "step": 110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.075765438505449e-05,
+      "loss": 1.0365,
+      "step": 120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.2487458917142365e-05,
+      "loss": 1.2179,
+      "step": 130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.4217263449230237e-05,
+      "loss": 1.1192,
+      "step": 140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.5947067981318112e-05,
+      "loss": 0.8697,
+      "step": 150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.7676872513405988e-05,
+      "loss": 1.0286,
+      "step": 160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.940667704549386e-05,
+      "loss": 0.9186,
+      "step": 170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.1136481577581735e-05,
+      "loss": 1.0909,
+      "step": 180
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.286628610966961e-05,
+      "loss": 1.0519,
+      "step": 190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.4596090641757486e-05,
+      "loss": 1.1043,
+      "step": 200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.632589517384536e-05,
+      "loss": 0.8634,
+      "step": 210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.805569970593323e-05,
+      "loss": 0.86,
+      "step": 220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.97855042380211e-05,
+      "loss": 0.8325,
+      "step": 230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.151530877010898e-05,
+      "loss": 1.1083,
+      "step": 240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.324511330219685e-05,
+      "loss": 1.0759,
+      "step": 250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.497491783428473e-05,
+      "loss": 0.9178,
+      "step": 260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.67047223663726e-05,
+      "loss": 1.1132,
+      "step": 270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.8434526898460474e-05,
+      "loss": 1.0484,
+      "step": 280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 5.0164331430548346e-05,
+      "loss": 0.8629,
+      "step": 290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.1894135962636225e-05,
+      "loss": 0.6426,
+      "step": 300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5.36239404947241e-05,
+      "loss": 1.0117,
+      "step": 310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.5353745026811976e-05,
+      "loss": 0.9769,
+      "step": 320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.708354955889985e-05,
+      "loss": 0.9459,
+      "step": 330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.881335409098772e-05,
+      "loss": 1.3208,
+      "step": 340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 6.054315862307559e-05,
+      "loss": 0.9631,
+      "step": 350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 6.227296315516347e-05,
+      "loss": 1.3533,
+      "step": 360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 6.400276768725135e-05,
+      "loss": 1.2228,
+      "step": 370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.573257221933921e-05,
+      "loss": 0.9195,
+      "step": 380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.746237675142709e-05,
+      "loss": 1.0466,
+      "step": 390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6.919218128351497e-05,
+      "loss": 0.6537,
+      "step": 400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 7.092198581560285e-05,
+      "loss": 1.0766,
+      "step": 410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.265179034769071e-05,
+      "loss": 1.3196,
+      "step": 420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.438159487977858e-05,
+      "loss": 1.1902,
+      "step": 430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.611139941186646e-05,
+      "loss": 0.8251,
+      "step": 440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.784120394395434e-05,
+      "loss": 0.6196,
+      "step": 450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.95710084760422e-05,
+      "loss": 1.1941,
+      "step": 460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.130081300813008e-05,
+      "loss": 0.9509,
+      "step": 470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.303061754021796e-05,
+      "loss": 1.0669,
+      "step": 480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.476042207230583e-05,
+      "loss": 0.8493,
+      "step": 490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.64902266043937e-05,
+      "loss": 1.0381,
+      "step": 500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.822003113648158e-05,
+      "loss": 0.6853,
+      "step": 510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.994983566856946e-05,
+      "loss": 1.0025,
+      "step": 520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.167964020065733e-05,
+      "loss": 0.9758,
+      "step": 530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.34094447327452e-05,
+      "loss": 1.2263,
+      "step": 540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.513924926483307e-05,
+      "loss": 1.0245,
+      "step": 550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.686905379692095e-05,
+      "loss": 0.9136,
+      "step": 560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.859885832900881e-05,
+      "loss": 1.0586,
+      "step": 570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00010032866286109669,
+      "loss": 1.029,
+      "step": 580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00010205846739318457,
+      "loss": 0.9973,
+      "step": 590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00010378827192527245,
+      "loss": 0.7652,
+      "step": 600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00010551807645736032,
+      "loss": 1.2198,
+      "step": 610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001072478809894482,
+      "loss": 1.3727,
+      "step": 620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00010897768552153607,
+      "loss": 0.9157,
+      "step": 630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00011070749005362395,
+      "loss": 1.4063,
+      "step": 640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00011243729458571182,
+      "loss": 0.9697,
+      "step": 650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001141670991177997,
+      "loss": 1.0678,
+      "step": 660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00011589690364988756,
+      "loss": 1.1373,
+      "step": 670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00011762670818197544,
+      "loss": 1.1127,
+      "step": 680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001193565127140633,
+      "loss": 0.8828,
+      "step": 690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00012108631724615118,
+      "loss": 1.0497,
+      "step": 700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00012281612177823906,
+      "loss": 1.0252,
+      "step": 710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00012454592631032694,
+      "loss": 1.0175,
+      "step": 720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00012627573084241482,
+      "loss": 1.1303,
+      "step": 730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001280055353745027,
+      "loss": 0.8751,
+      "step": 740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00012973533990659055,
+      "loss": 0.9646,
+      "step": 750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013146514443867843,
+      "loss": 1.0412,
+      "step": 760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001331949489707663,
+      "loss": 1.3243,
+      "step": 770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013492475350285418,
+      "loss": 1.0182,
+      "step": 780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013665455803494206,
+      "loss": 1.3814,
+      "step": 790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013838436256702994,
+      "loss": 0.9131,
+      "step": 800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00014011416709911782,
+      "loss": 1.0305,
+      "step": 810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001418439716312057,
+      "loss": 0.9296,
+      "step": 820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00014357377616329355,
+      "loss": 1.1304,
+      "step": 830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00014530358069538143,
+      "loss": 0.964,
+      "step": 840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001470333852274693,
+      "loss": 0.9223,
+      "step": 850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00014876318975955716,
+      "loss": 0.8419,
+      "step": 860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00015049299429164504,
+      "loss": 0.7896,
+      "step": 870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00015222279882373292,
+      "loss": 1.2753,
+      "step": 880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001539526033558208,
+      "loss": 1.01,
+      "step": 890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015568240788790867,
+      "loss": 0.8729,
+      "step": 900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00015741221241999653,
+      "loss": 1.2337,
+      "step": 910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591420169520844,
+      "loss": 1.0598,
+      "step": 920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016087182148417228,
+      "loss": 0.9216,
+      "step": 930
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00016260162601626016,
+      "loss": 1.2614,
+      "step": 940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00016433143054834804,
+      "loss": 0.9602,
+      "step": 950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00016606123508043592,
+      "loss": 1.1126,
+      "step": 960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001677910396125238,
+      "loss": 0.7103,
+      "step": 970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00016952084414461165,
+      "loss": 1.0462,
+      "step": 980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00017125064867669953,
+      "loss": 0.9843,
+      "step": 990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001729804532087874,
+      "loss": 1.2419,
+      "step": 1000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00017471025774087529,
+      "loss": 1.3823,
+      "step": 1010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017644006227296316,
+      "loss": 0.9726,
+      "step": 1020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00017816986680505104,
+      "loss": 1.1238,
+      "step": 1030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00017989967133713892,
+      "loss": 0.934,
+      "step": 1040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001816294758692268,
+      "loss": 1.5093,
+      "step": 1050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00018335928040131465,
+      "loss": 1.0234,
+      "step": 1060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00018508908493340253,
+      "loss": 0.7998,
+      "step": 1070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001868188894654904,
+      "loss": 1.0191,
+      "step": 1080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001885486939975783,
+      "loss": 1.1171,
+      "step": 1090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00019027849852966614,
+      "loss": 1.1417,
+      "step": 1100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00019200830306175402,
+      "loss": 0.9539,
+      "step": 1110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001937381075938419,
+      "loss": 1.0955,
+      "step": 1120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00019546791212592978,
+      "loss": 1.3748,
+      "step": 1130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00019719771665801763,
+      "loss": 1.1126,
+      "step": 1140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001989275211901055,
+      "loss": 1.0674,
+      "step": 1150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00020065732572219339,
+      "loss": 1.1236,
+      "step": 1160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020238713025428126,
+      "loss": 1.0824,
+      "step": 1170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00020411693478636914,
+      "loss": 1.1181,
+      "step": 1180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00020584673931845702,
+      "loss": 1.0881,
+      "step": 1190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002075765438505449,
+      "loss": 1.1874,
+      "step": 1200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00020930634838263278,
+      "loss": 1.0377,
+      "step": 1210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00021103615291472063,
+      "loss": 0.8027,
+      "step": 1220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002127659574468085,
+      "loss": 0.8027,
+      "step": 1230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002144957619788964,
+      "loss": 1.2303,
+      "step": 1240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00021622556651098427,
+      "loss": 1.2914,
+      "step": 1250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00021795537104307214,
+      "loss": 1.2018,
+      "step": 1260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00021968517557516002,
+      "loss": 0.994,
+      "step": 1270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002214149801072479,
+      "loss": 1.1917,
+      "step": 1280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00022314478463933578,
+      "loss": 0.9846,
+      "step": 1290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00022487458917142363,
+      "loss": 1.1561,
+      "step": 1300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002266043937035115,
+      "loss": 1.3345,
+      "step": 1310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002283341982355994,
+      "loss": 1.1663,
+      "step": 1320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00023006400276768727,
+      "loss": 1.0486,
+      "step": 1330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00023179380729977512,
+      "loss": 1.2478,
+      "step": 1340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000233523611831863,
+      "loss": 1.1334,
+      "step": 1350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00023525341636395088,
+      "loss": 1.1248,
+      "step": 1360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00023698322089603876,
+      "loss": 1.2183,
+      "step": 1370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002387130254281266,
+      "loss": 1.4618,
+      "step": 1380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00024044282996021449,
+      "loss": 1.5796,
+      "step": 1390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00024217263449230237,
+      "loss": 1.4877,
+      "step": 1400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00024390243902439024,
+      "loss": 0.9612,
+      "step": 1410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002456322435564781,
+      "loss": 1.3634,
+      "step": 1420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000247362048088566,
+      "loss": 1.1377,
+      "step": 1430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002490918526206539,
+      "loss": 1.1365,
+      "step": 1440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00025082165715274176,
+      "loss": 1.2433,
+      "step": 1450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00025255146168482964,
+      "loss": 1.2164,
+      "step": 1460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002542812662169175,
+      "loss": 1.0668,
+      "step": 1470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002560110707490054,
+      "loss": 1.174,
+      "step": 1480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002577408752810932,
+      "loss": 1.1307,
+      "step": 1490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002594706798131811,
+      "loss": 1.2406,
+      "step": 1500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.000261200484345269,
+      "loss": 1.2542,
+      "step": 1510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00026293028887735685,
+      "loss": 0.9126,
+      "step": 1520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00026466009340944473,
+      "loss": 1.1105,
+      "step": 1530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002663898979415326,
+      "loss": 0.8491,
+      "step": 1540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002681197024736205,
+      "loss": 1.4498,
+      "step": 1550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00026984950700570837,
+      "loss": 1.3299,
+      "step": 1560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00027157931153779625,
+      "loss": 1.2222,
+      "step": 1570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002733091160698841,
+      "loss": 0.9812,
+      "step": 1580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.000275038920601972,
+      "loss": 1.2824,
+      "step": 1590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002767687251340599,
+      "loss": 1.1105,
+      "step": 1600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00027849852966614776,
+      "loss": 1.0978,
+      "step": 1610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00028022833419823564,
+      "loss": 1.1159,
+      "step": 1620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002819581387303235,
+      "loss": 1.1441,
+      "step": 1630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002836879432624114,
+      "loss": 1.2566,
+      "step": 1640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002854177477944992,
+      "loss": 1.1327,
+      "step": 1650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002871475523265871,
+      "loss": 1.232,
+      "step": 1660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000288877356858675,
+      "loss": 1.0974,
+      "step": 1670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00029060716139076286,
+      "loss": 1.0083,
+      "step": 1680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00029233696592285074,
+      "loss": 1.5513,
+      "step": 1690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002940667704549386,
+      "loss": 1.3731,
+      "step": 1700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00029579657498702644,
+      "loss": 1.1877,
+      "step": 1710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002975263795191143,
+      "loss": 1.199,
+      "step": 1720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002992561840512022,
+      "loss": 1.4768,
+      "step": 1730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003009859885832901,
+      "loss": 1.7352,
+      "step": 1740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00030271579311537796,
+      "loss": 1.3652,
+      "step": 1750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00030444559764746583,
+      "loss": 1.1335,
+      "step": 1760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003061754021795537,
+      "loss": 1.2259,
+      "step": 1770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003079052067116416,
+      "loss": 1.3024,
+      "step": 1780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00030963501124372947,
+      "loss": 1.4433,
+      "step": 1790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00031136481577581735,
+      "loss": 1.1668,
+      "step": 1800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003130946203079052,
+      "loss": 1.3298,
+      "step": 1810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00031482442483999305,
+      "loss": 1.2601,
+      "step": 1820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00031655422937208093,
+      "loss": 1.1204,
+      "step": 1830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003182840339041688,
+      "loss": 1.2142,
+      "step": 1840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003200138384362567,
+      "loss": 0.93,
+      "step": 1850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00032174364296834457,
+      "loss": 1.4335,
+      "step": 1860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00032347344750043245,
+      "loss": 1.3081,
+      "step": 1870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003252032520325203,
+      "loss": 1.2692,
+      "step": 1880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003269330565646082,
+      "loss": 1.2172,
+      "step": 1890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003286628610966961,
+      "loss": 1.2254,
+      "step": 1900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00033039266562878396,
+      "loss": 1.1448,
+      "step": 1910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00033212247016087184,
+      "loss": 1.2473,
+      "step": 1920
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": {
+        "accuracy": 0.5909208819714656
+      },
+      "eval_f1": {
+        "f1": 0.5910435702573883
+      },
+      "eval_loss": 0.9808588624000549,
+      "eval_precision": {
+        "precision": 0.6661654194259725
+      },
+      "eval_recall": {
+        "recall": 0.6016064778916844
+      },
+      "eval_runtime": 243.6104,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 7.914,
+      "step": 1927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003338522746929597,
+      "loss": 1.1687,
+      "step": 1930
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003355820792250476,
+      "loss": 1.4294,
+      "step": 1940
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003373118837571355,
+      "loss": 1.2866,
+      "step": 1950
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0003390416882892233,
+      "loss": 1.0858,
+      "step": 1960
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0003407714928213112,
+      "loss": 1.1465,
+      "step": 1970
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00034250129735339906,
+      "loss": 1.2876,
+      "step": 1980
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00034423110188548694,
+      "loss": 1.0266,
+      "step": 1990
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0003459609064175748,
+      "loss": 1.1494,
+      "step": 2000
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0003476907109496627,
+      "loss": 1.5705,
+      "step": 2010
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00034942051548175057,
+      "loss": 0.9916,
+      "step": 2020
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00035115032001383845,
+      "loss": 1.2183,
+      "step": 2030
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00035288012454592633,
+      "loss": 1.1186,
+      "step": 2040
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0003546099290780142,
+      "loss": 1.1386,
+      "step": 2050
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0003563397336101021,
+      "loss": 1.413,
+      "step": 2060
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00035806953814218997,
+      "loss": 1.4434,
+      "step": 2070
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00035979934267427784,
+      "loss": 1.3144,
+      "step": 2080
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003615291472063657,
+      "loss": 1.4108,
+      "step": 2090
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0003632589517384536,
+      "loss": 1.2305,
+      "step": 2100
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0003649887562705415,
+      "loss": 1.2335,
+      "step": 2110
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0003667185608026293,
+      "loss": 1.3617,
+      "step": 2120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0003684483653347172,
+      "loss": 1.1457,
+      "step": 2130
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00037017816986680506,
+      "loss": 1.0253,
+      "step": 2140
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00037190797439889294,
+      "loss": 1.1486,
+      "step": 2150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0003736377789309808,
+      "loss": 1.2603,
+      "step": 2160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003753675834630687,
+      "loss": 1.25,
+      "step": 2170
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003770973879951566,
+      "loss": 1.5531,
+      "step": 2180
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003788271925272444,
+      "loss": 1.3442,
+      "step": 2190
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003805569970593323,
+      "loss": 1.1317,
+      "step": 2200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00038228680159142016,
+      "loss": 1.3271,
+      "step": 2210
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00038401660612350804,
+      "loss": 1.3165,
+      "step": 2220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0003857464106555959,
+      "loss": 1.2377,
+      "step": 2230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0003874762151876838,
+      "loss": 1.2677,
+      "step": 2240
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0003892060197197717,
+      "loss": 1.2756,
+      "step": 2250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00039093582425185955,
+      "loss": 1.0193,
+      "step": 2260
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00039266562878394743,
+      "loss": 1.239,
+      "step": 2270
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00039439543331603526,
+      "loss": 1.6044,
+      "step": 2280
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00039612523784812313,
+      "loss": 1.2396,
+      "step": 2290
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000397855042380211,
+      "loss": 1.0345,
+      "step": 2300
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0003995848469122989,
+      "loss": 1.3125,
+      "step": 2310
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00040131465144438677,
+      "loss": 1.4382,
+      "step": 2320
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00040304445597647465,
+      "loss": 1.1232,
+      "step": 2330
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00040477426050856253,
+      "loss": 1.3415,
+      "step": 2340
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0004065040650406504,
+      "loss": 1.3517,
+      "step": 2350
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0004082338695727383,
+      "loss": 1.1871,
+      "step": 2360
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00040996367410482616,
+      "loss": 1.541,
+      "step": 2370
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00041169347863691404,
+      "loss": 1.0871,
+      "step": 2380
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0004134232831690019,
+      "loss": 1.0727,
+      "step": 2390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0004151530877010898,
+      "loss": 1.2664,
+      "step": 2400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0004168828922331777,
+      "loss": 0.9518,
+      "step": 2410
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00041861269676526556,
+      "loss": 1.5851,
+      "step": 2420
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0004203425012973534,
+      "loss": 1.3433,
+      "step": 2430
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00042207230582944126,
+      "loss": 1.1357,
+      "step": 2440
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00042380211036152914,
+      "loss": 1.3124,
+      "step": 2450
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.000425531914893617,
+      "loss": 1.3198,
+      "step": 2460
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0004272617194257049,
+      "loss": 1.31,
+      "step": 2470
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0004289915239577928,
+      "loss": 1.2563,
+      "step": 2480
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00043072132848988065,
+      "loss": 1.2553,
+      "step": 2490
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00043245113302196853,
+      "loss": 1.3314,
+      "step": 2500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0004341809375540564,
+      "loss": 1.8077,
+      "step": 2510
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0004359107420861443,
+      "loss": 1.7739,
+      "step": 2520
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00043764054661823217,
+      "loss": 1.3835,
+      "step": 2530
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00043937035115032005,
+      "loss": 1.3689,
+      "step": 2540
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0004411001556824079,
+      "loss": 1.4259,
+      "step": 2550
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0004428299602144958,
+      "loss": 1.0997,
+      "step": 2560
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0004445597647465837,
+      "loss": 0.944,
+      "step": 2570
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00044628956927867156,
+      "loss": 1.431,
+      "step": 2580
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0004480193738107594,
+      "loss": 1.492,
+      "step": 2590
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00044974917834284726,
+      "loss": 1.0205,
+      "step": 2600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00045147898287493514,
+      "loss": 1.2334,
+      "step": 2610
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.000453208787407023,
+      "loss": 1.0434,
+      "step": 2620
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0004549385919391109,
+      "loss": 1.4912,
+      "step": 2630
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0004566683964711988,
+      "loss": 1.2939,
+      "step": 2640
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00045839820100328666,
+      "loss": 1.0389,
+      "step": 2650
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00046012800553537454,
+      "loss": 1.1925,
+      "step": 2660
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0004618578100674624,
+      "loss": 1.5455,
+      "step": 2670
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00046358761459955024,
+      "loss": 1.2916,
+      "step": 2680
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0004653174191316381,
+      "loss": 1.253,
+      "step": 2690
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000467047223663726,
+      "loss": 1.377,
+      "step": 2700
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0004687770281958139,
+      "loss": 1.4184,
+      "step": 2710
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00047050683272790175,
+      "loss": 1.3838,
+      "step": 2720
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00047223663725998963,
+      "loss": 1.1912,
+      "step": 2730
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0004739664417920775,
+      "loss": 1.0304,
+      "step": 2740
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00047569624632416534,
+      "loss": 1.3583,
+      "step": 2750
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0004774260508562532,
+      "loss": 1.3856,
+      "step": 2760
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0004791558553883411,
+      "loss": 1.5254,
+      "step": 2770
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00048088565992042897,
+      "loss": 1.2068,
+      "step": 2780
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00048261546445251685,
+      "loss": 1.3795,
+      "step": 2790
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00048434526898460473,
+      "loss": 1.5103,
+      "step": 2800
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0004860750735166926,
+      "loss": 1.6531,
+      "step": 2810
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0004878048780487805,
+      "loss": 1.3675,
+      "step": 2820
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0004895346825808684,
+      "loss": 1.6244,
+      "step": 2830
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0004912644871129562,
+      "loss": 1.8395,
+      "step": 2840
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0004929942916450441,
+      "loss": 1.6552,
+      "step": 2850
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.000494724096177132,
+      "loss": 1.6432,
+      "step": 2860
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0004964539007092199,
+      "loss": 1.4839,
+      "step": 2870
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0004981837052413078,
+      "loss": 1.403,
+      "step": 2880
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0004999135097733956,
+      "loss": 1.8624,
+      "step": 2890
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0005016433143054835,
+      "loss": 1.4894,
+      "step": 2900
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0005033731188375714,
+      "loss": 2.1144,
+      "step": 2910
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0005051029233696593,
+      "loss": 1.7605,
+      "step": 2920
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000506832727901747,
+      "loss": 1.6352,
+      "step": 2930
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.000508562532433835,
+      "loss": 1.6826,
+      "step": 2940
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0005102923369659228,
+      "loss": 1.8023,
+      "step": 2950
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0005120221414980108,
+      "loss": 1.6957,
+      "step": 2960
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0005137519460300986,
+      "loss": 2.1667,
+      "step": 2970
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0005154817505621864,
+      "loss": 2.1662,
+      "step": 2980
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0005172115550942743,
+      "loss": 1.8724,
+      "step": 2990
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0005189413596263622,
+      "loss": 1.7312,
+      "step": 3000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0005206711641584501,
+      "loss": 2.0953,
+      "step": 3010
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.000522400968690538,
+      "loss": 1.8218,
+      "step": 3020
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0005241307732226258,
+      "loss": 1.9732,
+      "step": 3030
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0005258605777547137,
+      "loss": 2.229,
+      "step": 3040
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0005275903822868016,
+      "loss": 1.9946,
+      "step": 3050
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0005293201868188895,
+      "loss": 2.162,
+      "step": 3060
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0005310499913509773,
+      "loss": 2.0647,
+      "step": 3070
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0005327797958830652,
+      "loss": 2.0291,
+      "step": 3080
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0005345096004151531,
+      "loss": 2.0622,
+      "step": 3090
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000536239404947241,
+      "loss": 1.9414,
+      "step": 3100
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0005379692094793289,
+      "loss": 2.0906,
+      "step": 3110
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0005396990140114167,
+      "loss": 2.0018,
+      "step": 3120
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0005414288185435046,
+      "loss": 1.9754,
+      "step": 3130
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0005431586230755925,
+      "loss": 1.949,
+      "step": 3140
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0005448884276076803,
+      "loss": 1.918,
+      "step": 3150
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0005466182321397683,
+      "loss": 2.1205,
+      "step": 3160
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000548348036671856,
+      "loss": 1.8933,
+      "step": 3170
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.000550077841203944,
+      "loss": 1.9666,
+      "step": 3180
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0005518076457360318,
+      "loss": 1.9286,
+      "step": 3190
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0005535374502681198,
+      "loss": 1.9417,
+      "step": 3200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0005552672548002075,
+      "loss": 1.8296,
+      "step": 3210
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0005569970593322955,
+      "loss": 1.9005,
+      "step": 3220
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0005587268638643833,
+      "loss": 1.954,
+      "step": 3230
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0005604566683964713,
+      "loss": 1.9252,
+      "step": 3240
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000562186472928559,
+      "loss": 1.9827,
+      "step": 3250
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000563916277460647,
+      "loss": 1.7931,
+      "step": 3260
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0005656460819927348,
+      "loss": 1.8842,
+      "step": 3270
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0005673758865248228,
+      "loss": 1.8838,
+      "step": 3280
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0005691056910569106,
+      "loss": 1.7902,
+      "step": 3290
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0005708354955889984,
+      "loss": 1.9054,
+      "step": 3300
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0005725653001210863,
+      "loss": 1.972,
+      "step": 3310
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0005742951046531742,
+      "loss": 1.999,
+      "step": 3320
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0005760249091852621,
+      "loss": 1.9566,
+      "step": 3330
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00057775471371735,
+      "loss": 1.8273,
+      "step": 3340
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0005794845182494378,
+      "loss": 1.9554,
+      "step": 3350
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0005812143227815257,
+      "loss": 2.0318,
+      "step": 3360
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0005829441273136136,
+      "loss": 1.9838,
+      "step": 3370
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0005846739318457015,
+      "loss": 1.9402,
+      "step": 3380
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0005864037363777894,
+      "loss": 1.9904,
+      "step": 3390
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0005881335409098772,
+      "loss": 1.9198,
+      "step": 3400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0005898633454419651,
+      "loss": 1.9405,
+      "step": 3410
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0005915931499740529,
+      "loss": 1.8445,
+      "step": 3420
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0005933229545061409,
+      "loss": 1.9201,
+      "step": 3430
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0005950527590382286,
+      "loss": 1.8237,
+      "step": 3440
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0005967825635703165,
+      "loss": 1.7255,
+      "step": 3450
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0005985123681024044,
+      "loss": 1.844,
+      "step": 3460
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0006002421726344923,
+      "loss": 2.0433,
+      "step": 3470
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0006019719771665802,
+      "loss": 1.7753,
+      "step": 3480
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.000603701781698668,
+      "loss": 1.858,
+      "step": 3490
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0006054315862307559,
+      "loss": 1.9348,
+      "step": 3500
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0006071613907628438,
+      "loss": 1.8196,
+      "step": 3510
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0006088911952949317,
+      "loss": 2.0001,
+      "step": 3520
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0006106209998270195,
+      "loss": 1.806,
+      "step": 3530
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0006123508043591074,
+      "loss": 1.8369,
+      "step": 3540
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0006140806088911953,
+      "loss": 1.8824,
+      "step": 3550
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0006158104134232832,
+      "loss": 1.8376,
+      "step": 3560
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0006175402179553711,
+      "loss": 1.7554,
+      "step": 3570
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0006192700224874589,
+      "loss": 1.835,
+      "step": 3580
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0006209998270195468,
+      "loss": 1.7473,
+      "step": 3590
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0006227296315516347,
+      "loss": 1.8827,
+      "step": 3600
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0006244594360837226,
+      "loss": 1.7536,
+      "step": 3610
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0006261892406158103,
+      "loss": 1.8997,
+      "step": 3620
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0006279190451478983,
+      "loss": 1.813,
+      "step": 3630
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0006296488496799861,
+      "loss": 1.8474,
+      "step": 3640
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0006313786542120741,
+      "loss": 1.8263,
+      "step": 3650
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0006331084587441619,
+      "loss": 1.8817,
+      "step": 3660
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0006348382632762499,
+      "loss": 1.9347,
+      "step": 3670
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0006365680678083376,
+      "loss": 1.786,
+      "step": 3680
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0006382978723404256,
+      "loss": 2.197,
+      "step": 3690
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0006400276768725134,
+      "loss": 1.9695,
+      "step": 3700
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0006417574814046014,
+      "loss": 1.8939,
+      "step": 3710
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0006434872859366891,
+      "loss": 1.8662,
+      "step": 3720
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0006452170904687771,
+      "loss": 1.8054,
+      "step": 3730
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0006469468950008649,
+      "loss": 1.8351,
+      "step": 3740
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0006486766995329529,
+      "loss": 1.9102,
+      "step": 3750
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0006504065040650406,
+      "loss": 1.8462,
+      "step": 3760
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0006521363085971285,
+      "loss": 1.894,
+      "step": 3770
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0006538661131292164,
+      "loss": 1.7559,
+      "step": 3780
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0006555959176613043,
+      "loss": 1.8255,
+      "step": 3790
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0006573257221933922,
+      "loss": 1.7118,
+      "step": 3800
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00065905552672548,
+      "loss": 1.8048,
+      "step": 3810
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0006607853312575679,
+      "loss": 1.7759,
+      "step": 3820
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0006625151357896558,
+      "loss": 1.7974,
+      "step": 3830
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0006642449403217437,
+      "loss": 1.7673,
+      "step": 3840
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0006659747448538316,
+      "loss": 1.7959,
+      "step": 3850
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": {
+        "accuracy": 0.26848249027237353
+      },
+      "eval_f1": {
+        "f1": 0.17635435227079216
+      },
+      "eval_loss": 1.8579649925231934,
+      "eval_precision": {
+        "precision": 0.15600300783137744
+      },
+      "eval_recall": {
+        "recall": 0.2748312652287413
+      },
+      "eval_runtime": 115.151,
+      "eval_samples_per_second": 33.478,
+      "eval_steps_per_second": 16.743,
+      "step": 3854
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0006677045493859194,
+      "loss": 1.9251,
+      "step": 3860
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0006694343539180073,
+      "loss": 1.7743,
+      "step": 3870
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0006711641584500952,
+      "loss": 1.8658,
+      "step": 3880
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000672893962982183,
+      "loss": 1.8682,
+      "step": 3890
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.000674623767514271,
+      "loss": 1.9279,
+      "step": 3900
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0006763535720463587,
+      "loss": 2.0145,
+      "step": 3910
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.0006780833765784466,
+      "loss": 2.0323,
+      "step": 3920
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0006798131811105345,
+      "loss": 1.9978,
+      "step": 3930
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0006815429856426224,
+      "loss": 1.9258,
+      "step": 3940
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0006832727901747102,
+      "loss": 1.9777,
+      "step": 3950
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0006850025947067981,
+      "loss": 1.9442,
+      "step": 3960
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.000686732399238886,
+      "loss": 1.9772,
+      "step": 3970
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0006884622037709739,
+      "loss": 1.9368,
+      "step": 3980
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.0006901920083030618,
+      "loss": 1.8948,
+      "step": 3990
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0006919218128351496,
+      "loss": 1.9485,
+      "step": 4000
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0006936516173672375,
+      "loss": 1.8472,
+      "step": 4010
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0006953814218993254,
+      "loss": 1.7604,
+      "step": 4020
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0006971112264314133,
+      "loss": 1.8371,
+      "step": 4030
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0006988410309635011,
+      "loss": 1.7566,
+      "step": 4040
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.000700570835495589,
+      "loss": 1.9021,
+      "step": 4050
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0007023006400276769,
+      "loss": 1.744,
+      "step": 4060
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0007040304445597648,
+      "loss": 1.8639,
+      "step": 4070
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0007057602490918527,
+      "loss": 1.857,
+      "step": 4080
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0007074900536239404,
+      "loss": 1.8229,
+      "step": 4090
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0007092198581560284,
+      "loss": 1.8733,
+      "step": 4100
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0007109496626881162,
+      "loss": 1.8442,
+      "step": 4110
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0007126794672202042,
+      "loss": 1.8209,
+      "step": 4120
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0007144092717522919,
+      "loss": 1.7018,
+      "step": 4130
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0007161390762843799,
+      "loss": 1.8013,
+      "step": 4140
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0007178688808164677,
+      "loss": 1.9447,
+      "step": 4150
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0007195986853485557,
+      "loss": 1.8768,
+      "step": 4160
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0007213284898806435,
+      "loss": 1.822,
+      "step": 4170
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0007230582944127314,
+      "loss": 1.77,
+      "step": 4180
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0007247880989448192,
+      "loss": 1.8653,
+      "step": 4190
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0007265179034769072,
+      "loss": 1.7695,
+      "step": 4200
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.000728247708008995,
+      "loss": 1.8426,
+      "step": 4210
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.000729977512541083,
+      "loss": 1.8379,
+      "step": 4220
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0007317073170731707,
+      "loss": 1.8384,
+      "step": 4230
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0007334371216052586,
+      "loss": 1.9245,
+      "step": 4240
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0007351669261373465,
+      "loss": 1.8017,
+      "step": 4250
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0007368967306694344,
+      "loss": 1.8151,
+      "step": 4260
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0007386265352015222,
+      "loss": 1.8356,
+      "step": 4270
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0007403563397336101,
+      "loss": 1.9064,
+      "step": 4280
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.000742086144265698,
+      "loss": 1.8325,
+      "step": 4290
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.0007438159487977859,
+      "loss": 1.8079,
+      "step": 4300
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0007455457533298738,
+      "loss": 1.7581,
+      "step": 4310
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0007472755578619616,
+      "loss": 1.9346,
+      "step": 4320
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0007490053623940495,
+      "loss": 1.8706,
+      "step": 4330
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0007507351669261374,
+      "loss": 1.8314,
+      "step": 4340
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0007524649714582253,
+      "loss": 1.8881,
+      "step": 4350
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0007541947759903132,
+      "loss": 1.8785,
+      "step": 4360
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.000755924580522401,
+      "loss": 1.7916,
+      "step": 4370
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0007576543850544888,
+      "loss": 1.8089,
+      "step": 4380
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0007593841895865767,
+      "loss": 1.7736,
+      "step": 4390
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0007611139941186646,
+      "loss": 1.8184,
+      "step": 4400
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0007628437986507524,
+      "loss": 1.8084,
+      "step": 4410
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0007645736031828403,
+      "loss": 2.0347,
+      "step": 4420
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0007663034077149282,
+      "loss": 1.7984,
+      "step": 4430
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0007680332122470161,
+      "loss": 1.7799,
+      "step": 4440
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.000769763016779104,
+      "loss": 1.7603,
+      "step": 4450
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0007714928213111918,
+      "loss": 1.6663,
+      "step": 4460
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0007732226258432797,
+      "loss": 1.9232,
+      "step": 4470
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0007749524303753676,
+      "loss": 1.8086,
+      "step": 4480
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0007766822349074555,
+      "loss": 1.7616,
+      "step": 4490
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0007784120394395433,
+      "loss": 1.8018,
+      "step": 4500
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0007801418439716312,
+      "loss": 1.8413,
+      "step": 4510
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.0007818716485037191,
+      "loss": 2.1126,
+      "step": 4520
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.000783601453035807,
+      "loss": 1.9105,
+      "step": 4530
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0007853312575678949,
+      "loss": 1.8485,
+      "step": 4540
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0007870610620999827,
+      "loss": 1.8835,
+      "step": 4550
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0007887908666320705,
+      "loss": 1.7881,
+      "step": 4560
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0007905206711641585,
+      "loss": 1.7013,
+      "step": 4570
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0007922504756962463,
+      "loss": 1.7657,
+      "step": 4580
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0007939802802283343,
+      "loss": 1.9326,
+      "step": 4590
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.000795710084760422,
+      "loss": 1.6888,
+      "step": 4600
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00079743988929251,
+      "loss": 1.8114,
+      "step": 4610
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0007991696938245978,
+      "loss": 1.9245,
+      "step": 4620
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0008008994983566858,
+      "loss": 1.8287,
+      "step": 4630
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0008026293028887735,
+      "loss": 1.8763,
+      "step": 4640
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0008043591074208615,
+      "loss": 1.6896,
+      "step": 4650
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0008060889119529493,
+      "loss": 1.9996,
+      "step": 4660
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0008078187164850373,
+      "loss": 1.7345,
+      "step": 4670
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0008095485210171251,
+      "loss": 1.7908,
+      "step": 4680
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.000811278325549213,
+      "loss": 1.8212,
+      "step": 4690
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0008130081300813008,
+      "loss": 1.8434,
+      "step": 4700
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0008147379346133887,
+      "loss": 1.9086,
+      "step": 4710
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0008164677391454766,
+      "loss": 1.9098,
+      "step": 4720
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0008181975436775644,
+      "loss": 1.7695,
+      "step": 4730
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0008199273482096523,
+      "loss": 1.7415,
+      "step": 4740
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0008216571527417402,
+      "loss": 1.8291,
+      "step": 4750
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0008233869572738281,
+      "loss": 1.7842,
+      "step": 4760
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.000825116761805916,
+      "loss": 1.8745,
+      "step": 4770
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.0008268465663380038,
+      "loss": 1.7075,
+      "step": 4780
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0008285763708700917,
+      "loss": 1.8133,
+      "step": 4790
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0008303061754021796,
+      "loss": 1.7318,
+      "step": 4800
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0008320359799342675,
+      "loss": 1.874,
+      "step": 4810
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0008337657844663554,
+      "loss": 1.8895,
+      "step": 4820
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0008354955889984432,
+      "loss": 1.8485,
+      "step": 4830
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0008372253935305311,
+      "loss": 1.7643,
+      "step": 4840
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.000838955198062619,
+      "loss": 1.8477,
+      "step": 4850
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0008406850025947068,
+      "loss": 1.8611,
+      "step": 4860
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0008424148071267946,
+      "loss": 1.8908,
+      "step": 4870
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0008441446116588825,
+      "loss": 1.8017,
+      "step": 4880
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0008458744161909704,
+      "loss": 1.9186,
+      "step": 4890
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0008476042207230583,
+      "loss": 1.8141,
+      "step": 4900
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.0008493340252551462,
+      "loss": 1.8298,
+      "step": 4910
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000851063829787234,
+      "loss": 1.7711,
+      "step": 4920
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0008527936343193219,
+      "loss": 1.8265,
+      "step": 4930
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0008545234388514098,
+      "loss": 1.7649,
+      "step": 4940
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0008562532433834977,
+      "loss": 1.9159,
+      "step": 4950
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0008579830479155855,
+      "loss": 1.7968,
+      "step": 4960
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0008597128524476734,
+      "loss": 1.807,
+      "step": 4970
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0008614426569797613,
+      "loss": 1.7633,
+      "step": 4980
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0008631724615118492,
+      "loss": 1.8019,
+      "step": 4990
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0008649022660439371,
+      "loss": 1.8697,
+      "step": 5000
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0008666320705760249,
+      "loss": 1.8348,
+      "step": 5010
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0008683618751081128,
+      "loss": 1.805,
+      "step": 5020
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0008700916796402006,
+      "loss": 1.7346,
+      "step": 5030
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0008718214841722886,
+      "loss": 1.9219,
+      "step": 5040
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.0008735512887043763,
+      "loss": 2.0954,
+      "step": 5050
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0008752810932364643,
+      "loss": 1.9262,
+      "step": 5060
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0008770108977685521,
+      "loss": 1.8491,
+      "step": 5070
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0008787407023006401,
+      "loss": 2.0113,
+      "step": 5080
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0008804705068327279,
+      "loss": 1.9341,
+      "step": 5090
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0008822003113648159,
+      "loss": 1.8968,
+      "step": 5100
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0008839301158969036,
+      "loss": 2.0009,
+      "step": 5110
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0008856599204289916,
+      "loss": 1.9956,
+      "step": 5120
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0008873897249610794,
+      "loss": 1.9083,
+      "step": 5130
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0008891195294931674,
+      "loss": 1.9274,
+      "step": 5140
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0008908493340252551,
+      "loss": 1.8263,
+      "step": 5150
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0008925791385573431,
+      "loss": 1.8235,
+      "step": 5160
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0008943089430894309,
+      "loss": 1.8786,
+      "step": 5170
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0008960387476215188,
+      "loss": 1.8952,
+      "step": 5180
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0008977685521536067,
+      "loss": 1.7448,
+      "step": 5190
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0008994983566856945,
+      "loss": 1.9518,
+      "step": 5200
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0009012281612177824,
+      "loss": 1.8636,
+      "step": 5210
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0009029579657498703,
+      "loss": 1.9309,
+      "step": 5220
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.0009046877702819582,
+      "loss": 1.8714,
+      "step": 5230
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.000906417574814046,
+      "loss": 1.7624,
+      "step": 5240
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0009081473793461339,
+      "loss": 1.9421,
+      "step": 5250
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0009098771838782218,
+      "loss": 1.9204,
+      "step": 5260
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0009116069884103097,
+      "loss": 1.8561,
+      "step": 5270
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0009133367929423976,
+      "loss": 1.8766,
+      "step": 5280
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0009150665974744854,
+      "loss": 1.8905,
+      "step": 5290
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0009167964020065733,
+      "loss": 1.8466,
+      "step": 5300
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0009185262065386612,
+      "loss": 1.8735,
+      "step": 5310
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0009202560110707491,
+      "loss": 2.0758,
+      "step": 5320
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0009219858156028368,
+      "loss": 1.9582,
+      "step": 5330
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0009237156201349248,
+      "loss": 1.9317,
+      "step": 5340
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0009254454246670126,
+      "loss": 1.8719,
+      "step": 5350
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0009271752291991005,
+      "loss": 1.8377,
+      "step": 5360
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0009289050337311884,
+      "loss": 1.9785,
+      "step": 5370
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0009306348382632762,
+      "loss": 1.9259,
+      "step": 5380
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0009323646427953641,
+      "loss": 1.9097,
+      "step": 5390
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.000934094447327452,
+      "loss": 1.9034,
+      "step": 5400
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0009358242518595399,
+      "loss": 1.8983,
+      "step": 5410
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0009375540563916278,
+      "loss": 1.8115,
+      "step": 5420
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0009392838609237156,
+      "loss": 1.8606,
+      "step": 5430
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0009410136654558035,
+      "loss": 1.8883,
+      "step": 5440
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0009427434699878914,
+      "loss": 1.7868,
+      "step": 5450
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0009444732745199793,
+      "loss": 1.9107,
+      "step": 5460
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0009462030790520671,
+      "loss": 1.9427,
+      "step": 5470
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.000947932883584155,
+      "loss": 1.8159,
+      "step": 5480
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0009496626881162429,
+      "loss": 1.909,
+      "step": 5490
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0009513924926483307,
+      "loss": 1.8716,
+      "step": 5500
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0009531222971804187,
+      "loss": 1.9116,
+      "step": 5510
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0009548521017125064,
+      "loss": 1.8846,
+      "step": 5520
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0009565819062445944,
+      "loss": 1.8776,
+      "step": 5530
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0009583117107766822,
+      "loss": 1.828,
+      "step": 5540
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.0009600415153087702,
+      "loss": 1.9392,
+      "step": 5550
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0009617713198408579,
+      "loss": 1.8823,
+      "step": 5560
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0009635011243729459,
+      "loss": 1.8782,
+      "step": 5570
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0009652309289050337,
+      "loss": 1.8898,
+      "step": 5580
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0009669607334371217,
+      "loss": 1.8107,
+      "step": 5590
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0009686905379692095,
+      "loss": 1.8026,
+      "step": 5600
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0009704203425012974,
+      "loss": 1.8616,
+      "step": 5610
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0009721501470333852,
+      "loss": 1.8443,
+      "step": 5620
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.0009738799515654732,
+      "loss": 1.8465,
+      "step": 5630
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000975609756097561,
+      "loss": 1.8806,
+      "step": 5640
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0009773395606296489,
+      "loss": 1.8441,
+      "step": 5650
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0009790693651617367,
+      "loss": 1.8194,
+      "step": 5660
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0009807991696938246,
+      "loss": 1.9228,
+      "step": 5670
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0009825289742259125,
+      "loss": 1.792,
+      "step": 5680
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0009842587787580004,
+      "loss": 1.8211,
+      "step": 5690
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.0009859885832900882,
+      "loss": 1.81,
+      "step": 5700
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.0009877183878221761,
+      "loss": 1.8862,
+      "step": 5710
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.000989448192354264,
+      "loss": 1.8373,
+      "step": 5720
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0009911779968863519,
+      "loss": 1.8639,
+      "step": 5730
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0009929078014184398,
+      "loss": 1.8718,
+      "step": 5740
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0009946376059505276,
+      "loss": 1.8258,
+      "step": 5750
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0009963674104826155,
+      "loss": 1.8564,
+      "step": 5760
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0009980972150147034,
+      "loss": 1.9462,
+      "step": 5770
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0009998270195467913,
+      "loss": 1.8437,
+      "step": 5780
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": {
+        "accuracy": 0.1743190661478599
+      },
+      "eval_f1": {
+        "f1": 0.1111482743173735
+      },
+      "eval_loss": 1.9242523908615112,
+      "eval_precision": {
+        "precision": 0.1516816437642854
+      },
+      "eval_recall": {
+        "recall": 0.1887752678966561
+      },
+      "eval_runtime": 118.0197,
+      "eval_samples_per_second": 32.664,
+      "eval_steps_per_second": 16.336,
+      "step": 5781
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.000999999926169841,
+      "loss": 1.8482,
+      "step": 5790
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0009999996709545039,
+      "loss": 1.8564,
+      "step": 5800
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0009999992334425987,
+      "loss": 1.8474,
+      "step": 5810
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0009999986136342847,
+      "loss": 1.8997,
+      "step": 5820
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.000999997811529788,
+      "loss": 1.8401,
+      "step": 5830
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.000999996827129401,
+      "loss": 1.8567,
+      "step": 5840
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0009999956604334825,
+      "loss": 1.9804,
+      "step": 5850
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.000999994311442458,
+      "loss": 1.8427,
+      "step": 5860
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0009999927801568192,
+      "loss": 1.8681,
+      "step": 5870
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0009999910665771247,
+      "loss": 1.8561,
+      "step": 5880
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.000999989170703999,
+      "loss": 1.9098,
+      "step": 5890
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0009999870925381333,
+      "loss": 1.935,
+      "step": 5900
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0009999848320802853,
+      "loss": 2.0045,
+      "step": 5910
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0009999823893312794,
+      "loss": 1.9466,
+      "step": 5920
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0009999797642920058,
+      "loss": 1.9047,
+      "step": 5930
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0009999769569634218,
+      "loss": 1.9901,
+      "step": 5940
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.000999973967346551,
+      "loss": 1.9159,
+      "step": 5950
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0009999707954424836,
+      "loss": 1.9826,
+      "step": 5960
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0009999674412523754,
+      "loss": 1.9556,
+      "step": 5970
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0009999639047774498,
+      "loss": 1.9591,
+      "step": 5980
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.000999960186018996,
+      "loss": 1.9551,
+      "step": 5990
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0009999562849783699,
+      "loss": 1.937,
+      "step": 6000
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0009999522016569938,
+      "loss": 1.9167,
+      "step": 6010
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0009999479360563565,
+      "loss": 1.9296,
+      "step": 6020
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.000999943488178013,
+      "loss": 1.9602,
+      "step": 6030
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0009999388580235853,
+      "loss": 1.9229,
+      "step": 6040
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0009999340455947613,
+      "loss": 1.9508,
+      "step": 6050
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0009999290508932957,
+      "loss": 1.9451,
+      "step": 6060
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0009999238739210093,
+      "loss": 1.9379,
+      "step": 6070
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00099991851467979,
+      "loss": 1.9622,
+      "step": 6080
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0009999129731715912,
+      "loss": 1.9257,
+      "step": 6090
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0009999072493984337,
+      "loss": 1.936,
+      "step": 6100
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0009999013433624041,
+      "loss": 1.939,
+      "step": 6110
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.000999895255065656,
+      "loss": 1.9445,
+      "step": 6120
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0009998889845104092,
+      "loss": 1.945,
+      "step": 6130
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0009998825316989493,
+      "loss": 1.989,
+      "step": 6140
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0009998758966336297,
+      "loss": 1.9469,
+      "step": 6150
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0009998690793168688,
+      "loss": 1.9297,
+      "step": 6160
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0009998620797511525,
+      "loss": 1.9461,
+      "step": 6170
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0009998548979390331,
+      "loss": 1.9486,
+      "step": 6180
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0009998475338831286,
+      "loss": 1.9473,
+      "step": 6190
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.000999839987586124,
+      "loss": 1.9557,
+      "step": 6200
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0009998322590507706,
+      "loss": 1.9447,
+      "step": 6210
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0009998243482798863,
+      "loss": 1.9628,
+      "step": 6220
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0009998162552763551,
+      "loss": 1.9514,
+      "step": 6230
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0009998079800431282,
+      "loss": 1.9163,
+      "step": 6240
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.000999799522583222,
+      "loss": 1.9278,
+      "step": 6250
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00099979088289972,
+      "loss": 1.9194,
+      "step": 6260
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0009997820609957732,
+      "loss": 1.9528,
+      "step": 6270
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0009997730568745966,
+      "loss": 1.9465,
+      "step": 6280
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0009997638705394741,
+      "loss": 1.9105,
+      "step": 6290
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0009997545019937546,
+      "loss": 1.9432,
+      "step": 6300
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0009997449512408538,
+      "loss": 1.9348,
+      "step": 6310
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0009997352182842538,
+      "loss": 1.9364,
+      "step": 6320
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0009997253031275035,
+      "loss": 1.9734,
+      "step": 6330
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0009997152057742173,
+      "loss": 1.9606,
+      "step": 6340
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0009997049262280772,
+      "loss": 1.9483,
+      "step": 6350
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0009996944644928307,
+      "loss": 1.9196,
+      "step": 6360
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0009996838205722923,
+      "loss": 1.9586,
+      "step": 6370
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0009996729944703427,
+      "loss": 1.9599,
+      "step": 6380
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0009996619861909288,
+      "loss": 1.9236,
+      "step": 6390
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0009996507957380643,
+      "loss": 1.9563,
+      "step": 6400
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.000999639423115829,
+      "loss": 1.9324,
+      "step": 6410
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0009996278683283696,
+      "loss": 1.9386,
+      "step": 6420
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0009996161313798985,
+      "loss": 1.9506,
+      "step": 6430
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0009996042122746955,
+      "loss": 1.9184,
+      "step": 6440
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0009995921110171058,
+      "loss": 1.955,
+      "step": 6450
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0009995798276115414,
+      "loss": 1.9215,
+      "step": 6460
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0009995673620624807,
+      "loss": 1.9362,
+      "step": 6470
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0009995547143744688,
+      "loss": 1.939,
+      "step": 6480
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.000999541884552117,
+      "loss": 1.8658,
+      "step": 6490
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0009995288726001028,
+      "loss": 1.9591,
+      "step": 6500
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0009995156785231702,
+      "loss": 1.9814,
+      "step": 6510
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0009995023023261298,
+      "loss": 1.9623,
+      "step": 6520
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0009994887440138588,
+      "loss": 1.9069,
+      "step": 6530
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0009994750035912997,
+      "loss": 1.8249,
+      "step": 6540
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0009994610810634627,
+      "loss": 1.8952,
+      "step": 6550
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000999446976435424,
+      "loss": 1.9932,
+      "step": 6560
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0009994326897123256,
+      "loss": 1.8771,
+      "step": 6570
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0009994182208993764,
+      "loss": 2.0021,
+      "step": 6580
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.000999403570001852,
+      "loss": 1.9962,
+      "step": 6590
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0009993887370250939,
+      "loss": 1.9772,
+      "step": 6600
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00099937372197451,
+      "loss": 1.951,
+      "step": 6610
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0009993585248555745,
+      "loss": 1.9398,
+      "step": 6620
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0009993431456738284,
+      "loss": 1.9512,
+      "step": 6630
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.000999327584434879,
+      "loss": 1.9954,
+      "step": 6640
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0009993118411443995,
+      "loss": 1.9426,
+      "step": 6650
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00099929591580813,
+      "loss": 1.9368,
+      "step": 6660
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0009992798084318766,
+      "loss": 1.9514,
+      "step": 6670
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0009992635190215122,
+      "loss": 1.9406,
+      "step": 6680
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0009992470475829755,
+      "loss": 1.9547,
+      "step": 6690
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0009992303941222722,
+      "loss": 1.9348,
+      "step": 6700
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.000999213558645474,
+      "loss": 1.9717,
+      "step": 6710
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0009991965411587187,
+      "loss": 1.9505,
+      "step": 6720
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.000999179341668211,
+      "loss": 1.925,
+      "step": 6730
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0009991619601802217,
+      "loss": 1.9599,
+      "step": 6740
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.000999144396701088,
+      "loss": 1.9617,
+      "step": 6750
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0009991266512372132,
+      "loss": 1.9616,
+      "step": 6760
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0009991087237950676,
+      "loss": 1.9458,
+      "step": 6770
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0009990906143811868,
+      "loss": 1.9476,
+      "step": 6780
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0009990723230021742,
+      "loss": 1.919,
+      "step": 6790
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.000999053849664698,
+      "loss": 1.9499,
+      "step": 6800
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.000999035194375494,
+      "loss": 1.9444,
+      "step": 6810
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0009990163571413633,
+      "loss": 1.9498,
+      "step": 6820
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0009989973379691739,
+      "loss": 1.9729,
+      "step": 6830
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0009989781368658604,
+      "loss": 2.0205,
+      "step": 6840
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0009989587538384232,
+      "loss": 1.9512,
+      "step": 6850
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0009989391888939295,
+      "loss": 1.9329,
+      "step": 6860
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.000998919442039512,
+      "loss": 1.9252,
+      "step": 6870
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0009988995132823705,
+      "loss": 1.9392,
+      "step": 6880
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0009988794026297711,
+      "loss": 1.9263,
+      "step": 6890
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.000998859110089046,
+      "loss": 1.9332,
+      "step": 6900
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0009988386356675932,
+      "loss": 1.9138,
+      "step": 6910
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.000998817979372878,
+      "loss": 1.9538,
+      "step": 6920
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0009987971412124318,
+      "loss": 1.9492,
+      "step": 6930
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0009987761211938514,
+      "loss": 1.9498,
+      "step": 6940
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.000998754919324801,
+      "loss": 1.9517,
+      "step": 6950
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0009987335356130107,
+      "loss": 1.916,
+      "step": 6960
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0009987119700662767,
+      "loss": 1.9625,
+      "step": 6970
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0009986902226924615,
+      "loss": 1.9502,
+      "step": 6980
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.000998668293499494,
+      "loss": 1.9368,
+      "step": 6990
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0009986461824953701,
+      "loss": 1.9518,
+      "step": 7000
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0009986238896881508,
+      "loss": 1.9244,
+      "step": 7010
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0009986014150859637,
+      "loss": 1.9348,
+      "step": 7020
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0009985787586970034,
+      "loss": 1.9646,
+      "step": 7030
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0009985559205295301,
+      "loss": 1.9216,
+      "step": 7040
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0009985329005918703,
+      "loss": 1.9065,
+      "step": 7050
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0009985096988924169,
+      "loss": 1.9527,
+      "step": 7060
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0009984863154396293,
+      "loss": 1.9143,
+      "step": 7070
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0009984627502420329,
+      "loss": 1.9082,
+      "step": 7080
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0009984390033082194,
+      "loss": 1.9009,
+      "step": 7090
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0009984150746468465,
+      "loss": 1.766,
+      "step": 7100
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0009983909642666388,
+      "loss": 1.8211,
+      "step": 7110
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.000998366672176387,
+      "loss": 1.8446,
+      "step": 7120
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.000998342198384947,
+      "loss": 1.8342,
+      "step": 7130
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0009983175429012425,
+      "loss": 1.8967,
+      "step": 7140
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0009982927057342623,
+      "loss": 1.8948,
+      "step": 7150
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0009982676868930623,
+      "loss": 1.9191,
+      "step": 7160
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0009982424863867639,
+      "loss": 1.9185,
+      "step": 7170
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0009982171042245552,
+      "loss": 1.8958,
+      "step": 7180
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0009981915404156901,
+      "loss": 1.9371,
+      "step": 7190
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0009981657949694892,
+      "loss": 1.8752,
+      "step": 7200
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.000998139867895339,
+      "loss": 1.8434,
+      "step": 7210
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0009981137592026926,
+      "loss": 1.8518,
+      "step": 7220
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0009980874689010687,
+      "loss": 1.7907,
+      "step": 7230
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.000998060997000053,
+      "loss": 1.8275,
+      "step": 7240
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0009980343435092966,
+      "loss": 1.9524,
+      "step": 7250
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0009980075084385174,
+      "loss": 1.8303,
+      "step": 7260
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.000997980491797499,
+      "loss": 1.8564,
+      "step": 7270
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.000997953293596092,
+      "loss": 1.8884,
+      "step": 7280
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.000997925913844212,
+      "loss": 1.862,
+      "step": 7290
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0009978983525518422,
+      "loss": 1.8021,
+      "step": 7300
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0009978706097290308,
+      "loss": 1.9126,
+      "step": 7310
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0009978426853858929,
+      "loss": 1.9132,
+      "step": 7320
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0009978145795326093,
+      "loss": 1.8305,
+      "step": 7330
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0009977862921794272,
+      "loss": 1.871,
+      "step": 7340
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0009977578233366603,
+      "loss": 1.8928,
+      "step": 7350
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0009977291730146876,
+      "loss": 1.8817,
+      "step": 7360
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0009977003412239556,
+      "loss": 1.8307,
+      "step": 7370
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0009976713279749754,
+      "loss": 1.8866,
+      "step": 7380
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0009976421332783256,
+      "loss": 1.8442,
+      "step": 7390
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00099761275714465,
+      "loss": 1.8367,
+      "step": 7400
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0009975831995846591,
+      "loss": 1.8936,
+      "step": 7410
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0009975534606091296,
+      "loss": 1.807,
+      "step": 7420
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0009975235402289039,
+      "loss": 1.8231,
+      "step": 7430
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0009974934384548906,
+      "loss": 1.8122,
+      "step": 7440
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.000997463155298065,
+      "loss": 1.8028,
+      "step": 7450
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.000997432690769468,
+      "loss": 1.8177,
+      "step": 7460
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.0009974020448802066,
+      "loss": 1.9069,
+      "step": 7470
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.0009973712176414543,
+      "loss": 1.8245,
+      "step": 7480
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0009973402090644504,
+      "loss": 1.8876,
+      "step": 7490
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0009973090191605005,
+      "loss": 1.8971,
+      "step": 7500
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.000997277647940976,
+      "loss": 1.7917,
+      "step": 7510
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.000997246095417315,
+      "loss": 1.8608,
+      "step": 7520
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.000997214361601021,
+      "loss": 1.7863,
+      "step": 7530
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.0009971824465036642,
+      "loss": 1.8487,
+      "step": 7540
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0009971503501368805,
+      "loss": 1.8698,
+      "step": 7550
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0009971180725123722,
+      "loss": 1.8424,
+      "step": 7560
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.000997085613641907,
+      "loss": 1.9235,
+      "step": 7570
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0009970529735373198,
+      "loss": 1.7802,
+      "step": 7580
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.0009970201522105106,
+      "loss": 1.8938,
+      "step": 7590
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.000996987149673446,
+      "loss": 1.7626,
+      "step": 7600
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0009969539659381583,
+      "loss": 1.8598,
+      "step": 7610
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0009969206010167465,
+      "loss": 1.8503,
+      "step": 7620
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0009968870549213749,
+      "loss": 1.9692,
+      "step": 7630
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.0009968533276642742,
+      "loss": 1.8035,
+      "step": 7640
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000996819419257741,
+      "loss": 1.8945,
+      "step": 7650
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0009967853297141384,
+      "loss": 1.8432,
+      "step": 7660
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.0009967510590458953,
+      "loss": 1.8781,
+      "step": 7670
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.000996716607265506,
+      "loss": 1.8669,
+      "step": 7680
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0009966819743855318,
+      "loss": 1.8776,
+      "step": 7690
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0009966471604185997,
+      "loss": 1.803,
+      "step": 7700
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": {
+        "accuracy": 0.1846952010376135
+      },
+      "eval_f1": {
+        "f1": 0.1316379146314846
+      },
+      "eval_loss": 1.9003419876098633,
+      "eval_precision": {
+        "precision": 0.1460659091286986
+      },
+      "eval_recall": {
+        "recall": 0.1928584239028949
+      },
+      "eval_runtime": 112.0693,
+      "eval_samples_per_second": 34.398,
+      "eval_steps_per_second": 17.204,
+      "step": 7708
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0009966121653774023,
+      "loss": 1.8044,
+      "step": 7710
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.000996576989274699,
+      "loss": 1.856,
+      "step": 7720
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0009965416321233143,
+      "loss": 1.7691,
+      "step": 7730
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.0009965060939361393,
+      "loss": 1.8381,
+      "step": 7740
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.0009964703747261312,
+      "loss": 1.9571,
+      "step": 7750
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.000996434474506313,
+      "loss": 1.8362,
+      "step": 7760
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0009963983932897735,
+      "loss": 1.776,
+      "step": 7770
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0009963621310896678,
+      "loss": 1.8468,
+      "step": 7780
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0009963256879192165,
+      "loss": 1.7727,
+      "step": 7790
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0009962890637917072,
+      "loss": 1.8783,
+      "step": 7800
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.0009962522587204923,
+      "loss": 1.8325,
+      "step": 7810
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0009962152727189908,
+      "loss": 1.8977,
+      "step": 7820
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.0009961781058006876,
+      "loss": 1.7941,
+      "step": 7830
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0009961407579791334,
+      "loss": 1.7944,
+      "step": 7840
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0009961032292679453,
+      "loss": 1.7098,
+      "step": 7850
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0009960655196808055,
+      "loss": 1.9654,
+      "step": 7860
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0009960276292314632,
+      "loss": 1.755,
+      "step": 7870
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0009959895579337325,
+      "loss": 1.7779,
+      "step": 7880
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0009959513058014945,
+      "loss": 2.0262,
+      "step": 7890
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.000995912872848695,
+      "loss": 1.9572,
+      "step": 7900
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.000995874259089347,
+      "loss": 1.7426,
+      "step": 7910
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0009958354645375287,
+      "loss": 1.7812,
+      "step": 7920
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.000995796489207384,
+      "loss": 1.8882,
+      "step": 7930
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0009957573331131236,
+      "loss": 1.8841,
+      "step": 7940
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.000995717996269023,
+      "loss": 1.8111,
+      "step": 7950
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.0009956784786894243,
+      "loss": 1.7847,
+      "step": 7960
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0009956387803887355,
+      "loss": 1.8502,
+      "step": 7970
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0009955989013814303,
+      "loss": 1.8171,
+      "step": 7980
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0009955588416820482,
+      "loss": 1.9437,
+      "step": 7990
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0009955186013051948,
+      "loss": 1.8288,
+      "step": 8000
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.0009954781802655418,
+      "loss": 1.7936,
+      "step": 8010
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.0009954375785778257,
+      "loss": 1.7904,
+      "step": 8020
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0009953967962568501,
+      "loss": 1.743,
+      "step": 8030
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0009953558333174839,
+      "loss": 1.7847,
+      "step": 8040
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0009953146897746618,
+      "loss": 1.8605,
+      "step": 8050
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.0009952733656433848,
+      "loss": 1.7898,
+      "step": 8060
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.000995231860938719,
+      "loss": 1.9566,
+      "step": 8070
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.000995190175675797,
+      "loss": 1.8443,
+      "step": 8080
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0009951483098698168,
+      "loss": 1.8683,
+      "step": 8090
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0009951062635360425,
+      "loss": 1.786,
+      "step": 8100
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.000995064036689804,
+      "loss": 1.8314,
+      "step": 8110
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0009950216293464966,
+      "loss": 1.8071,
+      "step": 8120
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.000994979041521582,
+      "loss": 1.7505,
+      "step": 8130
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0009949362732305875,
+      "loss": 1.7527,
+      "step": 8140
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0009948933244891059,
+      "loss": 1.91,
+      "step": 8150
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0009948501953127961,
+      "loss": 1.7183,
+      "step": 8160
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0009948068857173829,
+      "loss": 1.9057,
+      "step": 8170
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0009947633957186567,
+      "loss": 1.7652,
+      "step": 8180
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0009947197253324731,
+      "loss": 1.8544,
+      "step": 8190
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0009946758745747548,
+      "loss": 1.9037,
+      "step": 8200
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.000994631843461489,
+      "loss": 1.8158,
+      "step": 8210
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.0009945876320087293,
+      "loss": 1.8054,
+      "step": 8220
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.000994543240232595,
+      "loss": 1.9452,
+      "step": 8230
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0009944986681492708,
+      "loss": 1.8333,
+      "step": 8240
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0009944539157750077,
+      "loss": 1.7858,
+      "step": 8250
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0009944089831261218,
+      "loss": 1.8569,
+      "step": 8260
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0009943638702189954,
+      "loss": 1.9051,
+      "step": 8270
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0009943185770700767,
+      "loss": 1.8262,
+      "step": 8280
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0009942731036958784,
+      "loss": 1.7873,
+      "step": 8290
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0009942274501129808,
+      "loss": 1.9258,
+      "step": 8300
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.0009941816163380284,
+      "loss": 1.7442,
+      "step": 8310
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0009941356023877317,
+      "loss": 1.7348,
+      "step": 8320
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0009940894082788674,
+      "loss": 1.8439,
+      "step": 8330
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0009940430340282776,
+      "loss": 1.9081,
+      "step": 8340
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.0009939964796528698,
+      "loss": 1.8382,
+      "step": 8350
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0009939497451696175,
+      "loss": 1.754,
+      "step": 8360
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00099390283059556,
+      "loss": 1.7955,
+      "step": 8370
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.0009938557359478017,
+      "loss": 1.6966,
+      "step": 8380
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000993808461243513,
+      "loss": 1.7841,
+      "step": 8390
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0009937610064999305,
+      "loss": 1.9377,
+      "step": 8400
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0009937133717343553,
+      "loss": 2.106,
+      "step": 8410
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0009936655569641548,
+      "loss": 1.874,
+      "step": 8420
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0009936175622067623,
+      "loss": 1.9231,
+      "step": 8430
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0009935693874796759,
+      "loss": 1.8391,
+      "step": 8440
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00099352103280046,
+      "loss": 1.8697,
+      "step": 8450
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.0009934724981867447,
+      "loss": 1.8482,
+      "step": 8460
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0009934237836562248,
+      "loss": 1.7992,
+      "step": 8470
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.000993374889226662,
+      "loss": 1.8635,
+      "step": 8480
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.000993325814915882,
+      "loss": 1.9378,
+      "step": 8490
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0009932765607417778,
+      "loss": 1.7518,
+      "step": 8500
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.0009932271267223064,
+      "loss": 1.8512,
+      "step": 8510
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.000993177512875492,
+      "loss": 1.8,
+      "step": 8520
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0009931277192194228,
+      "loss": 1.9404,
+      "step": 8530
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0009930777457722534,
+      "loss": 1.8171,
+      "step": 8540
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.000993027592552204,
+      "loss": 1.853,
+      "step": 8550
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.0009929772595775597,
+      "loss": 1.7074,
+      "step": 8560
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000992926746866672,
+      "loss": 1.886,
+      "step": 8570
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.0009928760544379571,
+      "loss": 1.786,
+      "step": 8580
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0009928251823098976,
+      "loss": 1.7788,
+      "step": 8590
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.0009927741305010408,
+      "loss": 1.7721,
+      "step": 8600
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0009927228990299999,
+      "loss": 1.8738,
+      "step": 8610
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.0009926714879154538,
+      "loss": 1.8784,
+      "step": 8620
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.0009926198971761462,
+      "loss": 1.7806,
+      "step": 8630
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.000992568126830887,
+      "loss": 1.7354,
+      "step": 8640
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0009925161768985515,
+      "loss": 1.8456,
+      "step": 8650
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.00099246404739808,
+      "loss": 1.9295,
+      "step": 8660
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0009924117383484788,
+      "loss": 1.7736,
+      "step": 8670
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0009923592497688193,
+      "loss": 1.9097,
+      "step": 8680
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0009923065816782382,
+      "loss": 1.7617,
+      "step": 8690
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0009922537340959386,
+      "loss": 1.8676,
+      "step": 8700
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.000992200707041188,
+      "loss": 1.8388,
+      "step": 8710
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.0009921475005333195,
+      "loss": 1.8154,
+      "step": 8720
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.0009920941145917322,
+      "loss": 1.9333,
+      "step": 8730
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00099204054923589,
+      "loss": 1.843,
+      "step": 8740
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.0009919868044853228,
+      "loss": 1.9432,
+      "step": 8750
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.0009919328803596255,
+      "loss": 1.7557,
+      "step": 8760
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.0009918787768784581,
+      "loss": 1.805,
+      "step": 8770
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0009918244940615468,
+      "loss": 1.8165,
+      "step": 8780
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0009917700319286827,
+      "loss": 1.8896,
+      "step": 8790
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.000991715390499722,
+      "loss": 1.8423,
+      "step": 8800
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.000991660569794587,
+      "loss": 1.9545,
+      "step": 8810
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0009916055698332646,
+      "loss": 1.8081,
+      "step": 8820
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0009915503906358079,
+      "loss": 1.7959,
+      "step": 8830
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.0009914950322223344,
+      "loss": 1.7619,
+      "step": 8840
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.0009914394946130276,
+      "loss": 1.6856,
+      "step": 8850
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0009913837778281363,
+      "loss": 1.8222,
+      "step": 8860
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.000991327881887974,
+      "loss": 1.7221,
+      "step": 8870
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0009912718068129207,
+      "loss": 1.7537,
+      "step": 8880
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0009912155526234203,
+      "loss": 1.7016,
+      "step": 8890
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0009911591193399831,
+      "loss": 1.84,
+      "step": 8900
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0009911025069831842,
+      "loss": 1.8604,
+      "step": 8910
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0009910457155736642,
+      "loss": 1.7591,
+      "step": 8920
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0009909887451321287,
+      "loss": 1.8894,
+      "step": 8930
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.000990931595679349,
+      "loss": 1.8903,
+      "step": 8940
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.000990874267236161,
+      "loss": 1.8954,
+      "step": 8950
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0009908167598234666,
+      "loss": 1.7949,
+      "step": 8960
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.0009907590734622325,
+      "loss": 1.8333,
+      "step": 8970
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.000990701208173491,
+      "loss": 1.8182,
+      "step": 8980
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.000990643163978339,
+      "loss": 1.8232,
+      "step": 8990
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.000990584940897939,
+      "loss": 1.6914,
+      "step": 9000
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0009905265389535192,
+      "loss": 1.8622,
+      "step": 9010
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0009904679581663723,
+      "loss": 1.8321,
+      "step": 9020
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0009904091985578564,
+      "loss": 1.8133,
+      "step": 9030
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.0009903502601493952,
+      "loss": 1.825,
+      "step": 9040
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.000990291142962477,
+      "loss": 1.8074,
+      "step": 9050
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0009902318470186554,
+      "loss": 1.7792,
+      "step": 9060
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0009901723723395495,
+      "loss": 1.8017,
+      "step": 9070
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0009901127189468433,
+      "loss": 1.8355,
+      "step": 9080
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0009900528868622862,
+      "loss": 1.7781,
+      "step": 9090
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.0009899928761076923,
+      "loss": 1.7382,
+      "step": 9100
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.0009899326867049414,
+      "loss": 1.7697,
+      "step": 9110
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.000989872318675978,
+      "loss": 1.7791,
+      "step": 9120
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0009898117720428118,
+      "loss": 1.853,
+      "step": 9130
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.0009897510468275179,
+      "loss": 1.8826,
+      "step": 9140
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0009896901430522362,
+      "loss": 1.9778,
+      "step": 9150
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0009896290607391718,
+      "loss": 1.9321,
+      "step": 9160
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.000989567799910595,
+      "loss": 1.9429,
+      "step": 9170
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.000989506360588841,
+      "loss": 1.9588,
+      "step": 9180
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0009894447427963103,
+      "loss": 1.9601,
+      "step": 9190
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0009893829465554681,
+      "loss": 1.9501,
+      "step": 9200
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0009893209718888454,
+      "loss": 1.9414,
+      "step": 9210
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.0009892588188190371,
+      "loss": 2.0087,
+      "step": 9220
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0009891964873687041,
+      "loss": 1.9263,
+      "step": 9230
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.000989133977560572,
+      "loss": 1.9342,
+      "step": 9240
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0009890712894174316,
+      "loss": 1.9505,
+      "step": 9250
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0009890084229621384,
+      "loss": 1.9658,
+      "step": 9260
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.0009889453782176132,
+      "loss": 1.9293,
+      "step": 9270
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0009888821552068417,
+      "loss": 1.946,
+      "step": 9280
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0009888187539528746,
+      "loss": 2.0008,
+      "step": 9290
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0009887551744788272,
+      "loss": 1.9339,
+      "step": 9300
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0009886914168078808,
+      "loss": 1.9194,
+      "step": 9310
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0009886274809632807,
+      "loss": 1.9403,
+      "step": 9320
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0009885633669683375,
+      "loss": 1.9332,
+      "step": 9330
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0009884990748464267,
+      "loss": 1.9838,
+      "step": 9340
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0009884346046209887,
+      "loss": 1.9355,
+      "step": 9350
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0009883699563155292,
+      "loss": 1.9767,
+      "step": 9360
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0009883051299536183,
+      "loss": 1.9497,
+      "step": 9370
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0009882401255588912,
+      "loss": 1.9496,
+      "step": 9380
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0009881749431550485,
+      "loss": 1.9549,
+      "step": 9390
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.000988109582765855,
+      "loss": 1.937,
+      "step": 9400
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.0009880440444151404,
+      "loss": 1.9494,
+      "step": 9410
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.0009879783281267999,
+      "loss": 1.9397,
+      "step": 9420
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.000987912433924793,
+      "loss": 1.9257,
+      "step": 9430
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0009878463618331445,
+      "loss": 1.8816,
+      "step": 9440
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0009877801118759438,
+      "loss": 1.9802,
+      "step": 9450
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0009877136840773452,
+      "loss": 1.9163,
+      "step": 9460
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.0009876470784615677,
+      "loss": 1.8752,
+      "step": 9470
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.0009875802950528954,
+      "loss": 1.9372,
+      "step": 9480
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.000987513333875677,
+      "loss": 1.8707,
+      "step": 9490
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.0009874461949543262,
+      "loss": 1.7964,
+      "step": 9500
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0009873788783133213,
+      "loss": 1.8374,
+      "step": 9510
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0009873113839772058,
+      "loss": 1.7562,
+      "step": 9520
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0009872437119705871,
+      "loss": 1.7843,
+      "step": 9530
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.0009871758623181384,
+      "loss": 1.7316,
+      "step": 9540
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0009871078350445971,
+      "loss": 1.8084,
+      "step": 9550
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0009870396301747658,
+      "loss": 1.8954,
+      "step": 9560
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000986971247733511,
+      "loss": 1.7907,
+      "step": 9570
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.0009869026877457649,
+      "loss": 1.9981,
+      "step": 9580
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.0009868339502365237,
+      "loss": 1.8354,
+      "step": 9590
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.000986765035230849,
+      "loss": 1.8247,
+      "step": 9600
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.0009866959427538666,
+      "loss": 1.7769,
+      "step": 9610
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.000986626672830767,
+      "loss": 1.8071,
+      "step": 9620
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.0009865572254868058,
+      "loss": 1.8121,
+      "step": 9630
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": {
+        "accuracy": 0.33229571984435796
+      },
+      "eval_f1": {
+        "f1": 0.24134878979912178
+      },
+      "eval_loss": 1.7385518550872803,
+      "eval_precision": {
+        "precision": 0.2476900084001273
+      },
+      "eval_recall": {
+        "recall": 0.3378007007551934
+      },
+      "eval_runtime": 102.5366,
+      "eval_samples_per_second": 37.596,
+      "eval_steps_per_second": 18.803,
+      "step": 9635
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.000986487600747303,
+      "loss": 1.842,
+      "step": 9640
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0009864177986376428,
+      "loss": 1.6668,
+      "step": 9650
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0009863478191832755,
+      "loss": 1.7779,
+      "step": 9660
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0009862776624097146,
+      "loss": 1.9484,
+      "step": 9670
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.0009862073283425387,
+      "loss": 1.7389,
+      "step": 9680
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0009861368170073915,
+      "loss": 1.7946,
+      "step": 9690
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.0009860661284299807,
+      "loss": 1.9001,
+      "step": 9700
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.000985995262636079,
+      "loss": 1.8473,
+      "step": 9710
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.0009859242196515235,
+      "loss": 1.8383,
+      "step": 9720
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0009858529995022162,
+      "loss": 1.7071,
+      "step": 9730
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0009857816022141232,
+      "loss": 1.7298,
+      "step": 9740
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0009857100278132759,
+      "loss": 1.7207,
+      "step": 9750
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0009856382763257694,
+      "loss": 1.7967,
+      "step": 9760
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.0009855663477777638,
+      "loss": 1.8998,
+      "step": 9770
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0009854942421954843,
+      "loss": 1.8676,
+      "step": 9780
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.0009854219596052197,
+      "loss": 1.9189,
+      "step": 9790
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.000985349500033324,
+      "loss": 1.7721,
+      "step": 9800
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.000985276863506215,
+      "loss": 1.8258,
+      "step": 9810
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0009852040500503761,
+      "loss": 1.7623,
+      "step": 9820
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0009851310596923541,
+      "loss": 1.7993,
+      "step": 9830
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0009850578924587613,
+      "loss": 1.7903,
+      "step": 9840
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.0009849845483762736,
+      "loss": 1.8658,
+      "step": 9850
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.000984911027471632,
+      "loss": 1.8133,
+      "step": 9860
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.0009848373297716413,
+      "loss": 1.842,
+      "step": 9870
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.000984763455303172,
+      "loss": 1.7339,
+      "step": 9880
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.0009846894040931573,
+      "loss": 1.7764,
+      "step": 9890
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0009846151761685965,
+      "loss": 1.832,
+      "step": 9900
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0009845407715565523,
+      "loss": 1.8243,
+      "step": 9910
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0009844661902841522,
+      "loss": 1.8842,
+      "step": 9920
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.0009843914323785878,
+      "loss": 1.8539,
+      "step": 9930
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.0009843164978671159,
+      "loss": 1.7264,
+      "step": 9940
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.0009842413867770566,
+      "loss": 1.8868,
+      "step": 9950
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.000984166099135795,
+      "loss": 1.7069,
+      "step": 9960
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.0009840906349707808,
+      "loss": 1.7897,
+      "step": 9970
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0009840149943095274,
+      "loss": 1.7677,
+      "step": 9980
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.0009839391771796129,
+      "loss": 1.643,
+      "step": 9990
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.0009838631836086797,
+      "loss": 1.7819,
+      "step": 10000
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.000983787013624435,
+      "loss": 1.8601,
+      "step": 10010
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0009837106672546492,
+      "loss": 1.7895,
+      "step": 10020
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0009836341445271582,
+      "loss": 1.8085,
+      "step": 10030
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.0009835574454698612,
+      "loss": 1.787,
+      "step": 10040
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0009834805701107229,
+      "loss": 1.8519,
+      "step": 10050
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0009834035184777708,
+      "loss": 1.7236,
+      "step": 10060
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0009833262905990978,
+      "loss": 1.8543,
+      "step": 10070
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0009832488865028606,
+      "loss": 1.6843,
+      "step": 10080
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0009831713062172802,
+      "loss": 1.6678,
+      "step": 10090
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.000983093549770642,
+      "loss": 1.9228,
+      "step": 10100
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.000983015617191295,
+      "loss": 1.7595,
+      "step": 10110
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.0009829375085076534,
+      "loss": 1.7518,
+      "step": 10120
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0009828592237481953,
+      "loss": 1.7682,
+      "step": 10130
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0009827807629414623,
+      "loss": 1.7051,
+      "step": 10140
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0009827021261160607,
+      "loss": 1.7019,
+      "step": 10150
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0009826233133006612,
+      "loss": 1.6837,
+      "step": 10160
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.0009825443245239985,
+      "loss": 1.7594,
+      "step": 10170
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.000982465159814871,
+      "loss": 1.867,
+      "step": 10180
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.000982385819202142,
+      "loss": 1.6618,
+      "step": 10190
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0009823063027147383,
+      "loss": 1.6971,
+      "step": 10200
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0009822266103816516,
+      "loss": 1.9753,
+      "step": 10210
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0009821467422319364,
+      "loss": 1.7991,
+      "step": 10220
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0009820666982947126,
+      "loss": 1.8229,
+      "step": 10230
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0009819864785991638,
+      "loss": 1.8833,
+      "step": 10240
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0009819060831745372,
+      "loss": 1.8254,
+      "step": 10250
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.0009818255120501447,
+      "loss": 1.7899,
+      "step": 10260
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0009817447652553618,
+      "loss": 1.7448,
+      "step": 10270
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.0009816638428196286,
+      "loss": 1.7687,
+      "step": 10280
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0009815827447724483,
+      "loss": 1.7017,
+      "step": 10290
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0009815014711433891,
+      "loss": 1.6916,
+      "step": 10300
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.000981420021962083,
+      "loss": 1.6695,
+      "step": 10310
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0009813383972582254,
+      "loss": 1.916,
+      "step": 10320
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.0009812565970615766,
+      "loss": 1.8299,
+      "step": 10330
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0009811746214019599,
+      "loss": 1.7731,
+      "step": 10340
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0009810924703092634,
+      "loss": 1.8146,
+      "step": 10350
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0009810101438134388,
+      "loss": 1.794,
+      "step": 10360
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.0009809276419445018,
+      "loss": 1.7443,
+      "step": 10370
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.000980844964732532,
+      "loss": 1.6457,
+      "step": 10380
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.0009807621122076729,
+      "loss": 1.7514,
+      "step": 10390
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.000980679084400132,
+      "loss": 1.7748,
+      "step": 10400
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.0009805958813401809,
+      "loss": 1.8627,
+      "step": 10410
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.0009805125030581546,
+      "loss": 1.9099,
+      "step": 10420
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.0009804289495844525,
+      "loss": 1.8258,
+      "step": 10430
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0009803452209495374,
+      "loss": 1.8714,
+      "step": 10440
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0009802613171839363,
+      "loss": 1.7998,
+      "step": 10450
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0009801772383182402,
+      "loss": 1.7625,
+      "step": 10460
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.0009800929843831034,
+      "loss": 1.8561,
+      "step": 10470
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0009800085554092443,
+      "loss": 1.7429,
+      "step": 10480
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.0009799239514274455,
+      "loss": 1.8117,
+      "step": 10490
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0009798391724685526,
+      "loss": 1.5906,
+      "step": 10500
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.0009797542185634757,
+      "loss": 1.7367,
+      "step": 10510
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0009796690897431886,
+      "loss": 1.8763,
+      "step": 10520
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0009795837860387285,
+      "loss": 1.7159,
+      "step": 10530
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0009794983074811962,
+      "loss": 1.6452,
+      "step": 10540
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0009794126541017574,
+      "loss": 1.7642,
+      "step": 10550
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.0009793268259316401,
+      "loss": 1.7403,
+      "step": 10560
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0009792408230021369,
+      "loss": 1.8482,
+      "step": 10570
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.000979154645344604,
+      "loss": 1.8208,
+      "step": 10580
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0009790682929904611,
+      "loss": 1.8134,
+      "step": 10590
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0009789817659711917,
+      "loss": 1.7651,
+      "step": 10600
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.000978895064318343,
+      "loss": 1.7863,
+      "step": 10610
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.000978808188063526,
+      "loss": 1.7277,
+      "step": 10620
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0009787211372384147,
+      "loss": 1.777,
+      "step": 10630
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0009786339118747478,
+      "loss": 1.7441,
+      "step": 10640
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.000978546512004327,
+      "loss": 1.9068,
+      "step": 10650
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0009784589376590175,
+      "loss": 1.7786,
+      "step": 10660
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.0009783711888707485,
+      "loss": 1.7149,
+      "step": 10670
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.0009782832656715123,
+      "loss": 1.7479,
+      "step": 10680
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0009781951680933657,
+      "loss": 1.7153,
+      "step": 10690
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0009781068961684282,
+      "loss": 1.8753,
+      "step": 10700
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.000978018449928883,
+      "loss": 1.7972,
+      "step": 10710
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.000977929829406977,
+      "loss": 1.6859,
+      "step": 10720
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0009778410346350209,
+      "loss": 1.8016,
+      "step": 10730
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0009777520656453886,
+      "loss": 1.6856,
+      "step": 10740
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.0009776629224705175,
+      "loss": 1.7482,
+      "step": 10750
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.000977573605142909,
+      "loss": 1.7562,
+      "step": 10760
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.000977484113695127,
+      "loss": 1.7686,
+      "step": 10770
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.0009773944481598,
+      "loss": 1.8105,
+      "step": 10780
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0009773046085696192,
+      "loss": 1.8326,
+      "step": 10790
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0009772145949573394,
+      "loss": 1.7717,
+      "step": 10800
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0009771244073557793,
+      "loss": 1.8094,
+      "step": 10810
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0009770340457978205,
+      "loss": 1.7605,
+      "step": 10820
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0009769435103164084,
+      "loss": 1.7474,
+      "step": 10830
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0009768528009445512,
+      "loss": 1.7731,
+      "step": 10840
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.0009767619177153215,
+      "loss": 1.8068,
+      "step": 10850
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.0009766708606618543,
+      "loss": 1.8979,
+      "step": 10860
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.0009765796298173487,
+      "loss": 1.7847,
+      "step": 10870
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0009764882252150666,
+      "loss": 1.8401,
+      "step": 10880
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.0009763966468883338,
+      "loss": 1.5723,
+      "step": 10890
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0009763048948705389,
+      "loss": 1.7514,
+      "step": 10900
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0009762129691951343,
+      "loss": 1.7905,
+      "step": 10910
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0009761208698956352,
+      "loss": 1.907,
+      "step": 10920
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0009760285970056205,
+      "loss": 1.8021,
+      "step": 10930
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0009759361505587324,
+      "loss": 1.6455,
+      "step": 10940
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0009758435305886763,
+      "loss": 1.8394,
+      "step": 10950
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.0009757507371292207,
+      "loss": 1.7937,
+      "step": 10960
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.0009756577702141973,
+      "loss": 1.7578,
+      "step": 10970
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0009755646298775016,
+      "loss": 1.7708,
+      "step": 10980
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0009754713161530919,
+      "loss": 1.8778,
+      "step": 10990
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0009753778290749895,
+      "loss": 1.843,
+      "step": 11000
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.0009752841686772793,
+      "loss": 1.8345,
+      "step": 11010
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0009751903349941094,
+      "loss": 1.6698,
+      "step": 11020
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0009750963280596907,
+      "loss": 1.7422,
+      "step": 11030
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0009750021479082977,
+      "loss": 1.8949,
+      "step": 11040
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0009749077945742677,
+      "loss": 1.806,
+      "step": 11050
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.0009748132680920014,
+      "loss": 1.7548,
+      "step": 11060
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.0009747185684959625,
+      "loss": 1.795,
+      "step": 11070
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0009746236958206779,
+      "loss": 1.7781,
+      "step": 11080
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0009745286501007376,
+      "loss": 1.787,
+      "step": 11090
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.0009744334313707944,
+      "loss": 1.8119,
+      "step": 11100
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0009743380396655646,
+      "loss": 1.7497,
+      "step": 11110
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.0009742424750198273,
+      "loss": 1.7348,
+      "step": 11120
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0009741467374684247,
+      "loss": 1.7012,
+      "step": 11130
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0009740508270462625,
+      "loss": 1.816,
+      "step": 11140
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.0009739547437883084,
+      "loss": 1.7649,
+      "step": 11150
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.000973858487729594,
+      "loss": 1.8383,
+      "step": 11160
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0009737620589052136,
+      "loss": 1.7866,
+      "step": 11170
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0009736654573503247,
+      "loss": 1.7176,
+      "step": 11180
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0009735686831001473,
+      "loss": 1.7426,
+      "step": 11190
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.0009734717361899646,
+      "loss": 1.7987,
+      "step": 11200
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0009733746166551232,
+      "loss": 1.7344,
+      "step": 11210
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0009732773245310318,
+      "loss": 1.809,
+      "step": 11220
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0009731798598531628,
+      "loss": 1.7074,
+      "step": 11230
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.0009730822226570509,
+      "loss": 1.6944,
+      "step": 11240
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0009729844129782942,
+      "loss": 1.7243,
+      "step": 11250
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.0009728864308525534,
+      "loss": 1.7516,
+      "step": 11260
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.000972788276315552,
+      "loss": 1.7105,
+      "step": 11270
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.0009726899494030768,
+      "loss": 1.6982,
+      "step": 11280
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0009725914501509767,
+      "loss": 1.7656,
+      "step": 11290
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0009724927785951643,
+      "loss": 1.8973,
+      "step": 11300
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0009723939347716143,
+      "loss": 1.9031,
+      "step": 11310
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0009722949187163646,
+      "loss": 1.8425,
+      "step": 11320
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0009721957304655159,
+      "loss": 1.8458,
+      "step": 11330
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.0009720963700552315,
+      "loss": 1.7749,
+      "step": 11340
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.0009719968375217374,
+      "loss": 1.9048,
+      "step": 11350
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0009718971329013229,
+      "loss": 1.7444,
+      "step": 11360
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0009717972562303391,
+      "loss": 1.6944,
+      "step": 11370
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0009716972075452009,
+      "loss": 1.7349,
+      "step": 11380
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.0009715969868823849,
+      "loss": 1.8686,
+      "step": 11390
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0009714965942784313,
+      "loss": 1.6679,
+      "step": 11400
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0009713960297699424,
+      "loss": 1.7048,
+      "step": 11410
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0009712952933935832,
+      "loss": 1.8649,
+      "step": 11420
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0009711943851860819,
+      "loss": 1.7515,
+      "step": 11430
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0009710933051842288,
+      "loss": 1.9001,
+      "step": 11440
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.0009709920534248769,
+      "loss": 1.8789,
+      "step": 11450
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0009708906299449418,
+      "loss": 1.9457,
+      "step": 11460
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0009707890347814022,
+      "loss": 1.8666,
+      "step": 11470
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0009706872679712986,
+      "loss": 1.7263,
+      "step": 11480
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.0009705853295517348,
+      "loss": 1.8461,
+      "step": 11490
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.0009704832195598769,
+      "loss": 1.7918,
+      "step": 11500
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.0009703809380329534,
+      "loss": 1.8426,
+      "step": 11510
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.0009702784850082553,
+      "loss": 1.7251,
+      "step": 11520
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.0009701758605231368,
+      "loss": 1.6913,
+      "step": 11530
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0009700730646150135,
+      "loss": 1.5939,
+      "step": 11540
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0009699700973213645,
+      "loss": 1.8592,
+      "step": 11550
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0009698669586797311,
+      "loss": 1.7984,
+      "step": 11560
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": {
+        "accuracy": 0.3014267185473411
+      },
+      "eval_f1": {
+        "f1": 0.2074421859419775
+      },
+      "eval_loss": 1.7321857213974,
+      "eval_precision": {
+        "precision": 0.17325128260235972
+      },
+      "eval_recall": {
+        "recall": 0.31234162186318526
+      },
+      "eval_runtime": 103.9228,
+      "eval_samples_per_second": 37.095,
+      "eval_steps_per_second": 18.552,
+      "step": 11562
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.0009697636487277166,
+      "loss": 1.7792,
+      "step": 11570
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0009696601675029873,
+      "loss": 1.6447,
+      "step": 11580
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.0009695565150432717,
+      "loss": 1.8134,
+      "step": 11590
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.0009694526913863608,
+      "loss": 1.7022,
+      "step": 11600
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.0009693486965701082,
+      "loss": 1.8088,
+      "step": 11610
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0009692445306324296,
+      "loss": 1.852,
+      "step": 11620
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0009691401936113032,
+      "loss": 1.8435,
+      "step": 11630
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.0009690356855447695,
+      "loss": 1.8285,
+      "step": 11640
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0009689310064709316,
+      "loss": 1.7503,
+      "step": 11650
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.0009688261564279548,
+      "loss": 1.8343,
+      "step": 11660
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0009687211354540664,
+      "loss": 1.7397,
+      "step": 11670
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0009686159435875569,
+      "loss": 1.8581,
+      "step": 11680
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0009685105808667782,
+      "loss": 1.9103,
+      "step": 11690
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0009684050473301449,
+      "loss": 1.7767,
+      "step": 11700
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0009682993430161339,
+      "loss": 1.7691,
+      "step": 11710
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0009681934679632842,
+      "loss": 1.8959,
+      "step": 11720
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0009680874222101972,
+      "loss": 1.7943,
+      "step": 11730
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.0009679812057955365,
+      "loss": 1.8252,
+      "step": 11740
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0009678748187580278,
+      "loss": 1.8658,
+      "step": 11750
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.0009677682611364593,
+      "loss": 1.9736,
+      "step": 11760
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.000967661532969681,
+      "loss": 1.7876,
+      "step": 11770
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.0009675546342966054,
+      "loss": 1.7609,
+      "step": 11780
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0009674475651562071,
+      "loss": 1.8479,
+      "step": 11790
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0009673403255875225,
+      "loss": 1.756,
+      "step": 11800
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.0009672329156296508,
+      "loss": 1.7295,
+      "step": 11810
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.0009671253353217528,
+      "loss": 2.0492,
+      "step": 11820
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0009670175847030515,
+      "loss": 1.7373,
+      "step": 11830
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0009669096638128322,
+      "loss": 1.9517,
+      "step": 11840
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.000966801572690442,
+      "loss": 1.7751,
+      "step": 11850
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0009666933113752905,
+      "loss": 1.8541,
+      "step": 11860
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.0009665848799068486,
+      "loss": 1.8422,
+      "step": 11870
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0009664762783246501,
+      "loss": 1.7967,
+      "step": 11880
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0009663675066682902,
+      "loss": 1.761,
+      "step": 11890
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0009662585649774264,
+      "loss": 1.7239,
+      "step": 11900
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.000966149453291778,
+      "loss": 1.8602,
+      "step": 11910
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0009660401716511266,
+      "loss": 1.7893,
+      "step": 11920
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.0009659307200953156,
+      "loss": 1.7777,
+      "step": 11930
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0009658210986642499,
+      "loss": 1.8514,
+      "step": 11940
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.0009657113073978972,
+      "loss": 1.7697,
+      "step": 11950
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0009656013463362865,
+      "loss": 1.7849,
+      "step": 11960
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0009654912155195088,
+      "loss": 1.7234,
+      "step": 11970
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0009653809149877172,
+      "loss": 1.7891,
+      "step": 11980
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.0009652704447811265,
+      "loss": 1.6848,
+      "step": 11990
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0009651598049400132,
+      "loss": 1.8787,
+      "step": 12000
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.0009650489955047161,
+      "loss": 1.8163,
+      "step": 12010
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0009649380165156356,
+      "loss": 1.6971,
+      "step": 12020
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0009648268680132338,
+      "loss": 1.7703,
+      "step": 12030
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0009647155500380348,
+      "loss": 1.8627,
+      "step": 12040
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0009646040626306243,
+      "loss": 1.831,
+      "step": 12050
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.0009644924058316498,
+      "loss": 1.6914,
+      "step": 12060
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.0009643805796818208,
+      "loss": 1.8321,
+      "step": 12070
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0009642685842219084,
+      "loss": 1.7376,
+      "step": 12080
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.0009641564194927451,
+      "loss": 1.7676,
+      "step": 12090
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0009640440855352257,
+      "loss": 1.88,
+      "step": 12100
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0009639315823903065,
+      "loss": 1.7956,
+      "step": 12110
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.000963818910099005,
+      "loss": 1.8877,
+      "step": 12120
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.0009637060687024012,
+      "loss": 1.8099,
+      "step": 12130
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.000963593058241636,
+      "loss": 1.7973,
+      "step": 12140
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0009634798787579124,
+      "loss": 1.7283,
+      "step": 12150
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.000963366530292495,
+      "loss": 1.8744,
+      "step": 12160
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0009632530128867099,
+      "loss": 1.7395,
+      "step": 12170
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0009631393265819444,
+      "loss": 1.7538,
+      "step": 12180
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.0009630254714196483,
+      "loss": 1.8393,
+      "step": 12190
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.000962911447441332,
+      "loss": 1.7348,
+      "step": 12200
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.0009627972546885682,
+      "loss": 1.8174,
+      "step": 12210
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.0009626828932029907,
+      "loss": 1.7659,
+      "step": 12220
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0009625683630262951,
+      "loss": 1.7505,
+      "step": 12230
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0009624536642002379,
+      "loss": 1.7804,
+      "step": 12240
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.0009623387967666382,
+      "loss": 1.71,
+      "step": 12250
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.0009622237607673752,
+      "loss": 1.6895,
+      "step": 12260
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0009621085562443908,
+      "loss": 1.8456,
+      "step": 12270
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.0009619931832396873,
+      "loss": 1.8019,
+      "step": 12280
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0009618776417953294,
+      "loss": 1.6919,
+      "step": 12290
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.0009617619319534428,
+      "loss": 1.7653,
+      "step": 12300
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.000961646053756214,
+      "loss": 1.7869,
+      "step": 12310
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0009615300072458919,
+      "loss": 1.8052,
+      "step": 12320
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0009614137924647861,
+      "loss": 1.7572,
+      "step": 12330
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0009612974094552678,
+      "loss": 1.6216,
+      "step": 12340
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0009611808582597694,
+      "loss": 1.8323,
+      "step": 12350
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.0009610641389207848,
+      "loss": 1.8704,
+      "step": 12360
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.000960947251480869,
+      "loss": 1.8894,
+      "step": 12370
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.0009608301959826384,
+      "loss": 1.8785,
+      "step": 12380
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0009607129724687708,
+      "loss": 1.8482,
+      "step": 12390
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0009605955809820049,
+      "loss": 1.904,
+      "step": 12400
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.000960478021565141,
+      "loss": 1.8867,
+      "step": 12410
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0009603602942610403,
+      "loss": 1.907,
+      "step": 12420
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.0009602423991126258,
+      "loss": 1.9181,
+      "step": 12430
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.0009601243361628809,
+      "loss": 1.8593,
+      "step": 12440
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.0009600061054548506,
+      "loss": 1.7883,
+      "step": 12450
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0009598877070316412,
+      "loss": 1.7691,
+      "step": 12460
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0009597691409364199,
+      "loss": 1.8437,
+      "step": 12470
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0009596504072124153,
+      "loss": 1.8041,
+      "step": 12480
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0009595315059029166,
+      "loss": 1.8488,
+      "step": 12490
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0009594124370512746,
+      "loss": 1.8085,
+      "step": 12500
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0009592932007009009,
+      "loss": 1.7688,
+      "step": 12510
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0009591737968952685,
+      "loss": 1.8884,
+      "step": 12520
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.0009590542256779111,
+      "loss": 1.9053,
+      "step": 12530
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0009589344870924237,
+      "loss": 1.9115,
+      "step": 12540
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.000958814581182462,
+      "loss": 1.8114,
+      "step": 12550
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0009586945079917429,
+      "loss": 1.8891,
+      "step": 12560
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0009585742675640445,
+      "loss": 1.8384,
+      "step": 12570
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0009584538599432056,
+      "loss": 1.859,
+      "step": 12580
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0009583332851731258,
+      "loss": 1.7339,
+      "step": 12590
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.0009582125432977661,
+      "loss": 1.8824,
+      "step": 12600
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.0009580916343611481,
+      "loss": 1.751,
+      "step": 12610
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0009579705584073544,
+      "loss": 1.6991,
+      "step": 12620
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.0009578493154805285,
+      "loss": 1.7719,
+      "step": 12630
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0009577279056248749,
+      "loss": 1.8559,
+      "step": 12640
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0009576063288846585,
+      "loss": 1.7681,
+      "step": 12650
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0009574845853042057,
+      "loss": 1.7717,
+      "step": 12660
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.0009573626749279032,
+      "loss": 1.8828,
+      "step": 12670
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.0009572405978001988,
+      "loss": 1.7839,
+      "step": 12680
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.000957118353965601,
+      "loss": 1.714,
+      "step": 12690
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.0009569959434686792,
+      "loss": 1.8501,
+      "step": 12700
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0009568733663540634,
+      "loss": 1.7382,
+      "step": 12710
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.0009567506226664441,
+      "loss": 1.8327,
+      "step": 12720
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0009566277124505733,
+      "loss": 1.8203,
+      "step": 12730
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.000956504635751263,
+      "loss": 1.7329,
+      "step": 12740
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0009563813926133862,
+      "loss": 1.7699,
+      "step": 12750
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0009562579830818765,
+      "loss": 1.8813,
+      "step": 12760
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0009561344072017283,
+      "loss": 1.9075,
+      "step": 12770
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0009560106650179964,
+      "loss": 1.7317,
+      "step": 12780
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.0009558867565757965,
+      "loss": 1.6895,
+      "step": 12790
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.0009557626819203047,
+      "loss": 1.6617,
+      "step": 12800
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0009556384410967578,
+      "loss": 1.7724,
+      "step": 12810
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.0009555140341504534,
+      "loss": 1.8285,
+      "step": 12820
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.000955389461126749,
+      "loss": 1.8063,
+      "step": 12830
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.0009552647220710636,
+      "loss": 1.725,
+      "step": 12840
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.000955139817028876,
+      "loss": 1.6705,
+      "step": 12850
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.0009550147460457256,
+      "loss": 1.7254,
+      "step": 12860
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.0009548895091672127,
+      "loss": 1.8107,
+      "step": 12870
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.0009547641064389978,
+      "loss": 1.6768,
+      "step": 12880
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0009546385379068018,
+      "loss": 1.7145,
+      "step": 12890
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0009545128036164064,
+      "loss": 1.7159,
+      "step": 12900
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.000954386903613653,
+      "loss": 1.7503,
+      "step": 12910
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.0009542608379444444,
+      "loss": 1.7583,
+      "step": 12920
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.000954134606654743,
+      "loss": 1.8885,
+      "step": 12930
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.000954008209790572,
+      "loss": 1.6543,
+      "step": 12940
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.0009538816473980149,
+      "loss": 1.8198,
+      "step": 12950
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.0009537549195232156,
+      "loss": 1.7566,
+      "step": 12960
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000953628026212378,
+      "loss": 1.7466,
+      "step": 12970
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0009535009675117665,
+      "loss": 1.6681,
+      "step": 12980
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.0009533737434677063,
+      "loss": 1.8596,
+      "step": 12990
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0009532463541265822,
+      "loss": 1.8017,
+      "step": 13000
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0009531187995348394,
+      "loss": 1.7382,
+      "step": 13010
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0009529910797389835,
+      "loss": 1.7205,
+      "step": 13020
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.0009528631947855803,
+      "loss": 1.9236,
+      "step": 13030
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.000952735144721256,
+      "loss": 1.596,
+      "step": 13040
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0009526069295926966,
+      "loss": 1.8309,
+      "step": 13050
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0009524785494466485,
+      "loss": 1.7308,
+      "step": 13060
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.0009523500043299184,
+      "loss": 1.7496,
+      "step": 13070
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0009522212942893728,
+      "loss": 1.639,
+      "step": 13080
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.0009520924193719387,
+      "loss": 1.8481,
+      "step": 13090
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0009519633796246028,
+      "loss": 1.5754,
+      "step": 13100
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0009518341750944124,
+      "loss": 1.91,
+      "step": 13110
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0009517048058284745,
+      "loss": 1.7873,
+      "step": 13120
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.0009515752718739564,
+      "loss": 1.9317,
+      "step": 13130
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.0009514455732780851,
+      "loss": 1.8516,
+      "step": 13140
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.000951315710088148,
+      "loss": 1.8624,
+      "step": 13150
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0009511856823514923,
+      "loss": 1.7967,
+      "step": 13160
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 0.0009510554901155252,
+      "loss": 1.6538,
+      "step": 13170
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0009509251334277141,
+      "loss": 1.7719,
+      "step": 13180
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.000950794612335586,
+      "loss": 1.7984,
+      "step": 13190
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0009506639268867283,
+      "loss": 1.9263,
+      "step": 13200
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0009505330771287877,
+      "loss": 1.838,
+      "step": 13210
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.0009504020631094715,
+      "loss": 1.7548,
+      "step": 13220
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.000950270884876546,
+      "loss": 1.8724,
+      "step": 13230
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.0009501395424778385,
+      "loss": 1.7411,
+      "step": 13240
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0009500080359612353,
+      "loss": 1.7446,
+      "step": 13250
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.0009498763653746829,
+      "loss": 1.7895,
+      "step": 13260
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.0009497445307661875,
+      "loss": 1.6295,
+      "step": 13270
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.000949612532183815,
+      "loss": 1.6633,
+      "step": 13280
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0009494803696756913,
+      "loss": 2.1007,
+      "step": 13290
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0009493480432900021,
+      "loss": 1.7611,
+      "step": 13300
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.0009492155530749925,
+      "loss": 1.847,
+      "step": 13310
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.0009490828990789677,
+      "loss": 1.7704,
+      "step": 13320
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0009489500813502923,
+      "loss": 1.7631,
+      "step": 13330
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0009488170999373911,
+      "loss": 1.8335,
+      "step": 13340
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0009486839548887479,
+      "loss": 1.852,
+      "step": 13350
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0009485506462529067,
+      "loss": 1.609,
+      "step": 13360
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0009484171740784708,
+      "loss": 1.7785,
+      "step": 13370
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0009482835384141034,
+      "loss": 1.7499,
+      "step": 13380
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.0009481497393085272,
+      "loss": 1.8266,
+      "step": 13390
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.0009480157768105242,
+      "loss": 1.7917,
+      "step": 13400
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0009478816509689367,
+      "loss": 1.753,
+      "step": 13410
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0009477473618326656,
+      "loss": 1.7479,
+      "step": 13420
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.0009476129094506722,
+      "loss": 1.7644,
+      "step": 13430
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.0009474782938719769,
+      "loss": 1.6127,
+      "step": 13440
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0009473435151456593,
+      "loss": 1.817,
+      "step": 13450
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0009472085733208592,
+      "loss": 1.6814,
+      "step": 13460
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.0009470734684467754,
+      "loss": 1.6956,
+      "step": 13470
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0009469382005726661,
+      "loss": 1.7385,
+      "step": 13480
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": {
+        "accuracy": 0.3120622568093385
+      },
+      "eval_f1": {
+        "f1": 0.2220020856519651
+      },
+      "eval_loss": 1.6964426040649414,
+      "eval_precision": {
+        "precision": 0.24821744277779434
+      },
+      "eval_recall": {
+        "recall": 0.3239827067963708
+      },
+      "eval_runtime": 102.2041,
+      "eval_samples_per_second": 37.719,
+      "eval_steps_per_second": 18.864,
+      "step": 13489
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.0009468027697478494,
+      "loss": 1.867,
+      "step": 13490
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0009466671760217023,
+      "loss": 1.7677,
+      "step": 13500
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0009465314194436611,
+      "loss": 1.8369,
+      "step": 13510
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.0009463955000632222,
+      "loss": 1.8289,
+      "step": 13520
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.0009462594179299406,
+      "loss": 1.819,
+      "step": 13530
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0009461231730934312,
+      "loss": 1.8158,
+      "step": 13540
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0009459867656033676,
+      "loss": 1.7489,
+      "step": 13550
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.0009458501955094835,
+      "loss": 1.7089,
+      "step": 13560
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.0009457134628615712,
+      "loss": 1.9333,
+      "step": 13570
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0009455765677094825,
+      "loss": 1.8292,
+      "step": 13580
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0009454395101031283,
+      "loss": 1.8648,
+      "step": 13590
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0009453022900924795,
+      "loss": 1.7783,
+      "step": 13600
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 0.0009451649077275651,
+      "loss": 1.7176,
+      "step": 13610
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0009450273630584738,
+      "loss": 1.6828,
+      "step": 13620
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0009448896561353536,
+      "loss": 1.8494,
+      "step": 13630
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0009447517870084116,
+      "loss": 1.7528,
+      "step": 13640
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.000944613755727914,
+      "loss": 1.7617,
+      "step": 13650
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0009444755623441858,
+      "loss": 1.8195,
+      "step": 13660
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.0009443372069076118,
+      "loss": 1.8131,
+      "step": 13670
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.000944198689468635,
+      "loss": 1.6968,
+      "step": 13680
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.0009440600100777583,
+      "loss": 1.7565,
+      "step": 13690
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.0009439211687855433,
+      "loss": 1.7608,
+      "step": 13700
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.0009437821656426104,
+      "loss": 1.7052,
+      "step": 13710
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0009436430006996392,
+      "loss": 1.8568,
+      "step": 13720
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0009435036740073687,
+      "loss": 1.9427,
+      "step": 13730
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0009433641856165961,
+      "loss": 1.7566,
+      "step": 13740
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.0009432245355781782,
+      "loss": 1.7976,
+      "step": 13750
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.0009430847239430302,
+      "loss": 1.7797,
+      "step": 13760
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0009429447507621268,
+      "loss": 1.6913,
+      "step": 13770
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 0.0009428046160865009,
+      "loss": 1.7375,
+      "step": 13780
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0009426643199672451,
+      "loss": 1.7487,
+      "step": 13790
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.0009425238624555101,
+      "loss": 1.7797,
+      "step": 13800
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0009423832436025062,
+      "loss": 1.9005,
+      "step": 13810
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0009422424634595017,
+      "loss": 1.7843,
+      "step": 13820
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.0009421015220778242,
+      "loss": 1.7961,
+      "step": 13830
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00094196041950886,
+      "loss": 1.741,
+      "step": 13840
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.0009418191558040543,
+      "loss": 1.724,
+      "step": 13850
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.000941677731014911,
+      "loss": 1.7876,
+      "step": 13860
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0009415361451929922,
+      "loss": 1.7415,
+      "step": 13870
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0009413943983899194,
+      "loss": 1.6685,
+      "step": 13880
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.0009412524906573726,
+      "loss": 1.6868,
+      "step": 13890
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.0009411104220470903,
+      "loss": 1.8981,
+      "step": 13900
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00094096819261087,
+      "loss": 1.7498,
+      "step": 13910
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0009408258024005674,
+      "loss": 1.8692,
+      "step": 13920
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0009406832514680971,
+      "loss": 1.8906,
+      "step": 13930
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0009405405398654321,
+      "loss": 1.7531,
+      "step": 13940
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.0009403976676446044,
+      "loss": 1.7559,
+      "step": 13950
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.0009402546348577039,
+      "loss": 1.7564,
+      "step": 13960
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.0009401114415568797,
+      "loss": 1.8532,
+      "step": 13970
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 0.0009399680877943389,
+      "loss": 1.8019,
+      "step": 13980
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.0009398245736223476,
+      "loss": 1.8442,
+      "step": 13990
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0009396808990932298,
+      "loss": 1.8377,
+      "step": 14000
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0009395370642593686,
+      "loss": 1.8058,
+      "step": 14010
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.000939393069173205,
+      "loss": 1.6893,
+      "step": 14020
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.0009392489138872386,
+      "loss": 1.7234,
+      "step": 14030
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.0009391045984540278,
+      "loss": 1.7425,
+      "step": 14040
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.0009389601229261888,
+      "loss": 1.7706,
+      "step": 14050
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0009388154873563964,
+      "loss": 1.8947,
+      "step": 14060
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.000938670691797384,
+      "loss": 1.8376,
+      "step": 14070
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.0009385257363019428,
+      "loss": 1.6913,
+      "step": 14080
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.0009383806209229226,
+      "loss": 1.7429,
+      "step": 14090
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.0009382353457132319,
+      "loss": 1.7054,
+      "step": 14100
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.0009380899107258367,
+      "loss": 1.7288,
+      "step": 14110
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0009379443160137617,
+      "loss": 1.7682,
+      "step": 14120
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00093779856163009,
+      "loss": 1.7532,
+      "step": 14130
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0009376526476279625,
+      "loss": 1.6605,
+      "step": 14140
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.0009375065740605785,
+      "loss": 1.8331,
+      "step": 14150
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.0009373603409811954,
+      "loss": 1.7324,
+      "step": 14160
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.0009372139484431288,
+      "loss": 1.7655,
+      "step": 14170
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0009370673964997527,
+      "loss": 1.8016,
+      "step": 14180
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.0009369206852044985,
+      "loss": 1.7753,
+      "step": 14190
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0009367738146108566,
+      "loss": 1.6695,
+      "step": 14200
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0009366267847723751,
+      "loss": 1.7622,
+      "step": 14210
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0009364795957426596,
+      "loss": 1.6401,
+      "step": 14220
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.0009363322475753746,
+      "loss": 1.7551,
+      "step": 14230
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0009361847403242423,
+      "loss": 1.78,
+      "step": 14240
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.0009360370740430428,
+      "loss": 1.7698,
+      "step": 14250
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.0009358892487856142,
+      "loss": 1.7255,
+      "step": 14260
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0009357412646058526,
+      "loss": 1.7899,
+      "step": 14270
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.000935593121557712,
+      "loss": 1.717,
+      "step": 14280
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0009354448196952045,
+      "loss": 1.7246,
+      "step": 14290
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.0009352963590723999,
+      "loss": 1.7834,
+      "step": 14300
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0009351477397434261,
+      "loss": 1.9131,
+      "step": 14310
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.0009349989617624685,
+      "loss": 1.6523,
+      "step": 14320
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.000934850025183771,
+      "loss": 1.771,
+      "step": 14330
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.0009347009300616343,
+      "loss": 1.6129,
+      "step": 14340
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0009345516764504179,
+      "loss": 1.7177,
+      "step": 14350
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.0009344022644045384,
+      "loss": 1.7367,
+      "step": 14360
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0009342526939784706,
+      "loss": 1.6732,
+      "step": 14370
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.000934102965226747,
+      "loss": 1.6573,
+      "step": 14380
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.0009339530782039575,
+      "loss": 1.7023,
+      "step": 14390
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00093380303296475,
+      "loss": 1.7721,
+      "step": 14400
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00093365282956383,
+      "loss": 1.8415,
+      "step": 14410
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.0009335024680559608,
+      "loss": 1.7607,
+      "step": 14420
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.000933351948495963,
+      "loss": 1.8485,
+      "step": 14430
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.000933201270938715,
+      "loss": 1.7508,
+      "step": 14440
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0009330504354391531,
+      "loss": 1.8016,
+      "step": 14450
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0009328994420522707,
+      "loss": 1.7386,
+      "step": 14460
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.000932748290833119,
+      "loss": 1.8136,
+      "step": 14470
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.000932596981836807,
+      "loss": 1.7424,
+      "step": 14480
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.0009324455151185008,
+      "loss": 1.6428,
+      "step": 14490
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.0009322938907334238,
+      "loss": 1.7389,
+      "step": 14500
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.0009321421087368578,
+      "loss": 1.6418,
+      "step": 14510
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0009319901691841413,
+      "loss": 1.774,
+      "step": 14520
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.0009318380721306703,
+      "loss": 1.7014,
+      "step": 14530
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.0009316858176318985,
+      "loss": 1.6742,
+      "step": 14540
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.000931533405743337,
+      "loss": 1.6953,
+      "step": 14550
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.000931380836520554,
+      "loss": 1.6567,
+      "step": 14560
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0009312281100191752,
+      "loss": 1.7334,
+      "step": 14570
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0009310752262948838,
+      "loss": 1.8719,
+      "step": 14580
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.0009309221854034202,
+      "loss": 1.8223,
+      "step": 14590
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.000930768987400582,
+      "loss": 1.8442,
+      "step": 14600
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.0009306156323422244,
+      "loss": 1.7612,
+      "step": 14610
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.0009304621202842593,
+      "loss": 1.8426,
+      "step": 14620
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.0009303084512826562,
+      "loss": 1.7055,
+      "step": 14630
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.0009301546253934422,
+      "loss": 1.6697,
+      "step": 14640
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.0009300006426727008,
+      "loss": 1.7494,
+      "step": 14650
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0009298465031765732,
+      "loss": 1.6976,
+      "step": 14660
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0009296922069612578,
+      "loss": 1.8057,
+      "step": 14670
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0009295377540830097,
+      "loss": 1.9329,
+      "step": 14680
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.0009293831445981415,
+      "loss": 1.7457,
+      "step": 14690
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.000929228378563023,
+      "loss": 1.7327,
+      "step": 14700
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0009290734560340806,
+      "loss": 1.8215,
+      "step": 14710
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0009289183770677981,
+      "loss": 1.5827,
+      "step": 14720
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.0009287631417207163,
+      "loss": 1.54,
+      "step": 14730
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.000928607750049433,
+      "loss": 1.7683,
+      "step": 14740
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.000928452202110603,
+      "loss": 1.8053,
+      "step": 14750
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0009282964979609379,
+      "loss": 1.8536,
+      "step": 14760
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.0009281406376572066,
+      "loss": 1.7002,
+      "step": 14770
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.0009279846212562344,
+      "loss": 1.7629,
+      "step": 14780
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.0009278284488149042,
+      "loss": 1.677,
+      "step": 14790
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.0009276721203901555,
+      "loss": 1.7235,
+      "step": 14800
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0009275156360389842,
+      "loss": 1.6409,
+      "step": 14810
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.0009273589958184437,
+      "loss": 1.7063,
+      "step": 14820
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.000927202199785644,
+      "loss": 1.7886,
+      "step": 14830
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.0009270452479977519,
+      "loss": 1.9833,
+      "step": 14840
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0009268881405119909,
+      "loss": 1.7168,
+      "step": 14850
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.0009267308773856413,
+      "loss": 1.7221,
+      "step": 14860
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.0009265734586760405,
+      "loss": 1.7475,
+      "step": 14870
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.0009264158844405818,
+      "loss": 1.6445,
+      "step": 14880
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.0009262581547367164,
+      "loss": 1.7001,
+      "step": 14890
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.000926100269621951,
+      "loss": 1.8752,
+      "step": 14900
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.0009259422291538496,
+      "loss": 1.6439,
+      "step": 14910
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.0009257840333900325,
+      "loss": 1.7411,
+      "step": 14920
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0009256256823881772,
+      "loss": 1.5836,
+      "step": 14930
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.0009254671762060171,
+      "loss": 1.7718,
+      "step": 14940
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0009253085149013429,
+      "loss": 1.7117,
+      "step": 14950
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0009251496985320011,
+      "loss": 1.8424,
+      "step": 14960
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0009249907271558953,
+      "loss": 1.7725,
+      "step": 14970
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0009248316008309854,
+      "loss": 1.6359,
+      "step": 14980
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0009246723196152876,
+      "loss": 1.8313,
+      "step": 14990
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.000924512883566875,
+      "loss": 1.8976,
+      "step": 15000
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0009243532927438767,
+      "loss": 1.7728,
+      "step": 15010
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.0009241935472044787,
+      "loss": 1.8352,
+      "step": 15020
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0009240336470069231,
+      "loss": 1.7677,
+      "step": 15030
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.0009238735922095083,
+      "loss": 1.9427,
+      "step": 15040
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.0009237133828705893,
+      "loss": 1.8624,
+      "step": 15050
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0009235530190485776,
+      "loss": 1.8505,
+      "step": 15060
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0009233925008019405,
+      "loss": 1.827,
+      "step": 15070
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.000923231828189202,
+      "loss": 1.8163,
+      "step": 15080
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.0009230710012689421,
+      "loss": 1.8301,
+      "step": 15090
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0009229100200997976,
+      "loss": 1.7922,
+      "step": 15100
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.0009227488847404608,
+      "loss": 1.7587,
+      "step": 15110
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0009225875952496807,
+      "loss": 1.9375,
+      "step": 15120
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0009224261516862625,
+      "loss": 1.8747,
+      "step": 15130
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0009222645541090673,
+      "loss": 1.9106,
+      "step": 15140
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0009221028025770126,
+      "loss": 1.8543,
+      "step": 15150
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.0009219408971490719,
+      "loss": 1.8744,
+      "step": 15160
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.0009217788378842748,
+      "loss": 1.8751,
+      "step": 15170
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.0009216166248417072,
+      "loss": 1.7944,
+      "step": 15180
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.0009214542580805108,
+      "loss": 1.8621,
+      "step": 15190
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0009212917376598832,
+      "loss": 1.8727,
+      "step": 15200
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0009211290636390787,
+      "loss": 1.8287,
+      "step": 15210
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.0009209662360774071,
+      "loss": 1.8699,
+      "step": 15220
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.000920803255034234,
+      "loss": 1.8996,
+      "step": 15230
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.0009206401205689811,
+      "loss": 1.8532,
+      "step": 15240
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.0009204768327411267,
+      "loss": 1.8261,
+      "step": 15250
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.000920313391610204,
+      "loss": 1.8188,
+      "step": 15260
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0009201497972358028,
+      "loss": 1.7931,
+      "step": 15270
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0009199860496775683,
+      "loss": 1.6533,
+      "step": 15280
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0009198221489952019,
+      "loss": 1.705,
+      "step": 15290
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0009196580952484607,
+      "loss": 1.6474,
+      "step": 15300
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 0.0009194938884971576,
+      "loss": 1.7114,
+      "step": 15310
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0009193295288011614,
+      "loss": 1.9186,
+      "step": 15320
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0009191650162203963,
+      "loss": 1.795,
+      "step": 15330
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0009190003508148428,
+      "loss": 1.6633,
+      "step": 15340
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0009188355326445366,
+      "loss": 1.7603,
+      "step": 15350
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.0009186705617695695,
+      "loss": 1.5689,
+      "step": 15360
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.0009185054382500884,
+      "loss": 1.8437,
+      "step": 15370
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.0009183401621462966,
+      "loss": 1.8561,
+      "step": 15380
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0009181747335184527,
+      "loss": 1.754,
+      "step": 15390
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0009180091524268707,
+      "loss": 1.7165,
+      "step": 15400
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0009178434189319205,
+      "loss": 1.578,
+      "step": 15410
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": {
+        "accuracy": 0.2827496757457847
+      },
+      "eval_f1": {
+        "f1": 0.20468330816120264
+      },
+      "eval_loss": 1.6264866590499878,
+      "eval_precision": {
+        "precision": 0.23605832742279897
+      },
+      "eval_recall": {
+        "recall": 0.2983260879145022
+      },
+      "eval_runtime": 106.6464,
+      "eval_samples_per_second": 36.148,
+      "eval_steps_per_second": 18.078,
+      "step": 15416
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.0009176775330940274,
+      "loss": 1.7684,
+      "step": 15420
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.0009175114949736722,
+      "loss": 1.6885,
+      "step": 15430
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.0009173453046313914,
+      "loss": 1.6669,
+      "step": 15440
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.0009171789621277766,
+      "loss": 1.7721,
+      "step": 15450
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.0009170124675234757,
+      "loss": 1.7607,
+      "step": 15460
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0009168458208791911,
+      "loss": 1.7391,
+      "step": 15470
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0009166790222556812,
+      "loss": 1.8084,
+      "step": 15480
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0009165120717137598,
+      "loss": 1.7343,
+      "step": 15490
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0009163449693142957,
+      "loss": 1.6727,
+      "step": 15500
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0009161777151182136,
+      "loss": 1.7554,
+      "step": 15510
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.0009160103091864931,
+      "loss": 1.7161,
+      "step": 15520
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0009158427515801695,
+      "loss": 1.7783,
+      "step": 15530
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0009156750423603328,
+      "loss": 1.9211,
+      "step": 15540
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.0009155071815881291,
+      "loss": 1.687,
+      "step": 15550
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.000915339169324759,
+      "loss": 1.7848,
+      "step": 15560
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.0009151710056314788,
+      "loss": 1.8129,
+      "step": 15570
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.0009150026905695998,
+      "loss": 1.7993,
+      "step": 15580
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.0009148342242004888,
+      "loss": 1.9364,
+      "step": 15590
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.0009146656065855672,
+      "loss": 1.7503,
+      "step": 15600
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.0009144968377863121,
+      "loss": 1.8689,
+      "step": 15610
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0009143279178642552,
+      "loss": 1.7269,
+      "step": 15620
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.000914158846880984,
+      "loss": 1.7695,
+      "step": 15630
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0009139896248981403,
+      "loss": 1.9229,
+      "step": 15640
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.0009138202519774215,
+      "loss": 1.6619,
+      "step": 15650
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0009136507281805799,
+      "loss": 1.736,
+      "step": 15660
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.0009134810535694225,
+      "loss": 1.7193,
+      "step": 15670
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0009133112282058117,
+      "loss": 1.8024,
+      "step": 15680
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0009131412521516648,
+      "loss": 1.8519,
+      "step": 15690
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0009129711254689536,
+      "loss": 1.6348,
+      "step": 15700
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0009128008482197057,
+      "loss": 1.5877,
+      "step": 15710
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.0009126304204660025,
+      "loss": 1.6835,
+      "step": 15720
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.0009124598422699812,
+      "loss": 1.7916,
+      "step": 15730
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0009122891136938334,
+      "loss": 1.8504,
+      "step": 15740
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 0.0009121182347998055,
+      "loss": 1.7855,
+      "step": 15750
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.000911947205650199,
+      "loss": 1.7365,
+      "step": 15760
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0009117760263073697,
+      "loss": 1.8042,
+      "step": 15770
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.0009116046968337287,
+      "loss": 1.7053,
+      "step": 15780
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.0009114332172917416,
+      "loss": 1.8252,
+      "step": 15790
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0009112615877439285,
+      "loss": 1.7089,
+      "step": 15800
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.0009110898082528647,
+      "loss": 1.7229,
+      "step": 15810
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.0009109178788811797,
+      "loss": 1.6567,
+      "step": 15820
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.0009107457996915576,
+      "loss": 1.8491,
+      "step": 15830
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.0009105735707467375,
+      "loss": 1.6803,
+      "step": 15840
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0009104011921095131,
+      "loss": 1.7887,
+      "step": 15850
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0009102286638427324,
+      "loss": 1.7155,
+      "step": 15860
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0009100559860092981,
+      "loss": 1.6415,
+      "step": 15870
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.0009098831586721672,
+      "loss": 1.7925,
+      "step": 15880
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.0009097101818943516,
+      "loss": 1.8212,
+      "step": 15890
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.0009095370557389174,
+      "loss": 1.7778,
+      "step": 15900
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0009093637802689853,
+      "loss": 1.7496,
+      "step": 15910
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0009091903555477303,
+      "loss": 1.6661,
+      "step": 15920
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.0009090167816383821,
+      "loss": 1.646,
+      "step": 15930
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.0009088430586042242,
+      "loss": 1.76,
+      "step": 15940
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.0009086691865085954,
+      "loss": 1.8181,
+      "step": 15950
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.000908495165414888,
+      "loss": 1.6909,
+      "step": 15960
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.0009083209953865489,
+      "loss": 1.8341,
+      "step": 15970
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.0009081466764870793,
+      "loss": 1.7918,
+      "step": 15980
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0009079722087800352,
+      "loss": 1.6997,
+      "step": 15990
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0009077975923290258,
+      "loss": 1.6834,
+      "step": 16000
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.0009076228271977153,
+      "loss": 1.62,
+      "step": 16010
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 0.0009074479134498219,
+      "loss": 1.7986,
+      "step": 16020
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0009072728511491182,
+      "loss": 1.742,
+      "step": 16030
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.0009070976403594304,
+      "loss": 1.6723,
+      "step": 16040
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0009069222811446394,
+      "loss": 1.5897,
+      "step": 16050
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.0009067467735686799,
+      "loss": 1.6807,
+      "step": 16060
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0009065711176955409,
+      "loss": 1.7275,
+      "step": 16070
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0009063953135892654,
+      "loss": 1.7292,
+      "step": 16080
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0009062193613139503,
+      "loss": 1.7459,
+      "step": 16090
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0009060432609337468,
+      "loss": 1.7126,
+      "step": 16100
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.0009058670125128597,
+      "loss": 1.8328,
+      "step": 16110
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.000905690616115548,
+      "loss": 1.6626,
+      "step": 16120
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.000905514071806125,
+      "loss": 1.5575,
+      "step": 16130
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0009053373796489573,
+      "loss": 1.737,
+      "step": 16140
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.0009051605397084659,
+      "loss": 1.6842,
+      "step": 16150
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0009049835520491252,
+      "loss": 1.6591,
+      "step": 16160
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0009048064167354637,
+      "loss": 1.6706,
+      "step": 16170
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.0009046291338320642,
+      "loss": 1.6511,
+      "step": 16180
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.0009044517034035624,
+      "loss": 1.7473,
+      "step": 16190
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0009042741255146485,
+      "loss": 1.6796,
+      "step": 16200
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0009040964002300663,
+      "loss": 1.7314,
+      "step": 16210
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0009039185276146129,
+      "loss": 1.9037,
+      "step": 16220
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.0009037405077331399,
+      "loss": 1.6408,
+      "step": 16230
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0009035623406505519,
+      "loss": 1.5448,
+      "step": 16240
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0009033840264318076,
+      "loss": 1.6984,
+      "step": 16250
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.0009032055651419191,
+      "loss": 1.749,
+      "step": 16260
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.0009030269568459521,
+      "loss": 1.726,
+      "step": 16270
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0009028482016090262,
+      "loss": 1.7653,
+      "step": 16280
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0009026692994963142,
+      "loss": 1.8845,
+      "step": 16290
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0009024902505730428,
+      "loss": 1.7547,
+      "step": 16300
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0009023110549044919,
+      "loss": 1.6612,
+      "step": 16310
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.0009021317125559951,
+      "loss": 1.7741,
+      "step": 16320
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.0009019522235929394,
+      "loss": 1.6434,
+      "step": 16330
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0009017725880807655,
+      "loss": 1.737,
+      "step": 16340
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.0009015928060849671,
+      "loss": 1.8481,
+      "step": 16350
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.0009014128776710916,
+      "loss": 1.7002,
+      "step": 16360
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0009012328029047396,
+      "loss": 1.6447,
+      "step": 16370
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 0.0009010525818515654,
+      "loss": 1.6343,
+      "step": 16380
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0009008722145772761,
+      "loss": 1.7652,
+      "step": 16390
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0009006917011476326,
+      "loss": 1.8056,
+      "step": 16400
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.0009005110416284488,
+      "loss": 1.6906,
+      "step": 16410
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.0009003302360855922,
+      "loss": 1.7852,
+      "step": 16420
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.000900149284584983,
+      "loss": 1.7675,
+      "step": 16430
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.0008999681871925951,
+      "loss": 1.7707,
+      "step": 16440
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.0008997869439744555,
+      "loss": 1.7354,
+      "step": 16450
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.000899605554996644,
+      "loss": 1.7349,
+      "step": 16460
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0008994240203252939,
+      "loss": 1.8627,
+      "step": 16470
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0008992423400265917,
+      "loss": 1.8491,
+      "step": 16480
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0008990605141667767,
+      "loss": 1.7826,
+      "step": 16490
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0008988785428121416,
+      "loss": 1.6465,
+      "step": 16500
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.0008986964260290315,
+      "loss": 1.7663,
+      "step": 16510
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.0008985141638838454,
+      "loss": 1.7625,
+      "step": 16520
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0008983317564430349,
+      "loss": 1.7147,
+      "step": 16530
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.0008981492037731042,
+      "loss": 1.8479,
+      "step": 16540
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.0008979665059406111,
+      "loss": 1.8151,
+      "step": 16550
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.0008977836630121657,
+      "loss": 1.6965,
+      "step": 16560
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0008976006750544314,
+      "loss": 1.7078,
+      "step": 16570
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0008974175421341249,
+      "loss": 1.7454,
+      "step": 16580
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0008972342643180145,
+      "loss": 1.6888,
+      "step": 16590
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.0008970508416729227,
+      "loss": 1.5654,
+      "step": 16600
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0008968672742657238,
+      "loss": 1.9123,
+      "step": 16610
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.0008966835621633454,
+      "loss": 1.7758,
+      "step": 16620
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.0008964997054327677,
+      "loss": 1.7509,
+      "step": 16630
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0008963157041410237,
+      "loss": 1.6829,
+      "step": 16640
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.000896131558355199,
+      "loss": 1.7555,
+      "step": 16650
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.0008959472681424318,
+      "loss": 1.6942,
+      "step": 16660
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.0008957628335699133,
+      "loss": 1.631,
+      "step": 16670
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.0008955782547048869,
+      "loss": 1.6741,
+      "step": 16680
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.0008953935316146491,
+      "loss": 1.6885,
+      "step": 16690
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0008952086643665487,
+      "loss": 1.732,
+      "step": 16700
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.0008950236530279868,
+      "loss": 1.8046,
+      "step": 16710
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0008948384976664174,
+      "loss": 1.7123,
+      "step": 16720
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0008946531983493472,
+      "loss": 1.5934,
+      "step": 16730
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0008944677551443348,
+      "loss": 1.7552,
+      "step": 16740
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.0008942821681189916,
+      "loss": 1.6654,
+      "step": 16750
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.0008940964373409817,
+      "loss": 1.78,
+      "step": 16760
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.0008939105628780208,
+      "loss": 1.6564,
+      "step": 16770
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0008937245447978779,
+      "loss": 1.5775,
+      "step": 16780
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0008935383831683737,
+      "loss": 1.8037,
+      "step": 16790
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0008933520780573818,
+      "loss": 1.7375,
+      "step": 16800
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0008931656295328275,
+      "loss": 1.7594,
+      "step": 16810
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.0008929790376626887,
+      "loss": 1.7031,
+      "step": 16820
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.000892792302514996,
+      "loss": 1.6636,
+      "step": 16830
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0008926054241578315,
+      "loss": 1.805,
+      "step": 16840
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.0008924184026593295,
+      "loss": 1.6382,
+      "step": 16850
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0008922312380876772,
+      "loss": 1.7755,
+      "step": 16860
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0008920439305111136,
+      "loss": 1.694,
+      "step": 16870
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0008918564799979296,
+      "loss": 1.7392,
+      "step": 16880
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0008916688866164684,
+      "loss": 1.6026,
+      "step": 16890
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.0008914811504351255,
+      "loss": 1.6491,
+      "step": 16900
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.000891293271522348,
+      "loss": 1.7105,
+      "step": 16910
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.0008911052499466355,
+      "loss": 1.549,
+      "step": 16920
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0008909170857765395,
+      "loss": 1.6778,
+      "step": 16930
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0008907287790806633,
+      "loss": 1.7083,
+      "step": 16940
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0008905403299276622,
+      "loss": 1.6005,
+      "step": 16950
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.0008903517383862435,
+      "loss": 1.7597,
+      "step": 16960
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0008901630045251665,
+      "loss": 1.7905,
+      "step": 16970
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.0008899741284132426,
+      "loss": 1.6071,
+      "step": 16980
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0008897851101193341,
+      "loss": 1.7545,
+      "step": 16990
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.0008895959497123565,
+      "loss": 1.5481,
+      "step": 17000
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0008894066472612761,
+      "loss": 1.7643,
+      "step": 17010
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.0008892172028351113,
+      "loss": 1.7423,
+      "step": 17020
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0008890276165029324,
+      "loss": 1.6081,
+      "step": 17030
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0008888378883338613,
+      "loss": 1.7366,
+      "step": 17040
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0008886480183970715,
+      "loss": 1.5514,
+      "step": 17050
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.0008884580067617885,
+      "loss": 1.6213,
+      "step": 17060
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0008882678534972892,
+      "loss": 1.6877,
+      "step": 17070
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0008880775586729023,
+      "loss": 1.7817,
+      "step": 17080
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0008878871223580078,
+      "loss": 1.8002,
+      "step": 17090
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 0.0008876965446220379,
+      "loss": 1.7922,
+      "step": 17100
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0008875058255344757,
+      "loss": 1.8084,
+      "step": 17110
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0008873149651648561,
+      "loss": 1.9055,
+      "step": 17120
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.0008871239635827656,
+      "loss": 1.6935,
+      "step": 17130
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.0008869328208578422,
+      "loss": 1.7536,
+      "step": 17140
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0008867415370597752,
+      "loss": 1.6415,
+      "step": 17150
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0008865501122583052,
+      "loss": 1.7195,
+      "step": 17160
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.0008863585465232249,
+      "loss": 1.6616,
+      "step": 17170
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0008861668399243774,
+      "loss": 1.7208,
+      "step": 17180
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0008859749925316579,
+      "loss": 1.6244,
+      "step": 17190
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.0008857830044150129,
+      "loss": 1.7088,
+      "step": 17200
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.0008855908756444395,
+      "loss": 1.7205,
+      "step": 17210
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0008853986062899868,
+      "loss": 1.7547,
+      "step": 17220
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0008852061964217551,
+      "loss": 1.7007,
+      "step": 17230
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0008850136461098954,
+      "loss": 1.7271,
+      "step": 17240
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0008848209554246108,
+      "loss": 1.6084,
+      "step": 17250
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.0008846281244361545,
+      "loss": 1.7916,
+      "step": 17260
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 0.0008844351532148316,
+      "loss": 1.6409,
+      "step": 17270
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0008842420418309983,
+      "loss": 1.737,
+      "step": 17280
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.0008840487903550616,
+      "loss": 1.7597,
+      "step": 17290
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0008838553988574798,
+      "loss": 1.7017,
+      "step": 17300
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.000883661867408762,
+      "loss": 1.6752,
+      "step": 17310
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.0008834681960794684,
+      "loss": 1.761,
+      "step": 17320
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.0008832743849402106,
+      "loss": 1.7054,
+      "step": 17330
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0008830804340616507,
+      "loss": 1.7397,
+      "step": 17340
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": {
+        "accuracy": 0.41686121919584956
+      },
+      "eval_f1": {
+        "f1": 0.38005456072209515
+      },
+      "eval_loss": 1.454464077949524,
+      "eval_precision": {
+        "precision": 0.38606801407175106
+      },
+      "eval_recall": {
+        "recall": 0.43686617410103795
+      },
+      "eval_runtime": 187.9837,
+      "eval_samples_per_second": 20.507,
+      "eval_steps_per_second": 10.256,
+      "step": 17343
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.0008828863435145018,
+      "loss": 1.7464,
+      "step": 17350
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0008826921133695283,
+      "loss": 1.7127,
+      "step": 17360
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0008824977436975447,
+      "loss": 1.7265,
+      "step": 17370
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.0008823032345694176,
+      "loss": 1.7485,
+      "step": 17380
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.0008821085860560632,
+      "loss": 1.6013,
+      "step": 17390
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0008819137982284493,
+      "loss": 1.6759,
+      "step": 17400
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.0008817188711575943,
+      "loss": 1.6814,
+      "step": 17410
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.0008815238049145671,
+      "loss": 1.5589,
+      "step": 17420
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0008813285995704877,
+      "loss": 1.6493,
+      "step": 17430
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0008811332551965267,
+      "loss": 1.6929,
+      "step": 17440
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0008809377718639052,
+      "loss": 1.8371,
+      "step": 17450
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.0008807421496438955,
+      "loss": 1.6704,
+      "step": 17460
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.0008805463886078196,
+      "loss": 1.6997,
+      "step": 17470
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.000880350488827051,
+      "loss": 1.6588,
+      "step": 17480
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0008801544503730137,
+      "loss": 1.4844,
+      "step": 17490
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.0008799582733171816,
+      "loss": 1.6334,
+      "step": 17500
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0008797619577310798,
+      "loss": 1.6067,
+      "step": 17510
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0008795655036862836,
+      "loss": 1.6338,
+      "step": 17520
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.0008793689112544188,
+      "loss": 1.7329,
+      "step": 17530
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.0008791721805071616,
+      "loss": 1.8056,
+      "step": 17540
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.000878975311516239,
+      "loss": 1.8037,
+      "step": 17550
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.000878778304353428,
+      "loss": 1.7212,
+      "step": 17560
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0008785811590905562,
+      "loss": 1.7048,
+      "step": 17570
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.0008783838757995011,
+      "loss": 1.7173,
+      "step": 17580
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.0008781864545521912,
+      "loss": 1.5471,
+      "step": 17590
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.000877988895420605,
+      "loss": 1.6644,
+      "step": 17600
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.0008777911984767711,
+      "loss": 1.8331,
+      "step": 17610
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.0008775933637927687,
+      "loss": 1.6078,
+      "step": 17620
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.0008773953914407266,
+      "loss": 1.6736,
+      "step": 17630
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.0008771972814928246,
+      "loss": 1.7015,
+      "step": 17640
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0008769990340212923,
+      "loss": 1.5992,
+      "step": 17650
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0008768006490984093,
+      "loss": 1.6792,
+      "step": 17660
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.0008766021267965052,
+      "loss": 1.9294,
+      "step": 17670
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.0008764034671879602,
+      "loss": 1.6635,
+      "step": 17680
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.0008762046703452043,
+      "loss": 1.7352,
+      "step": 17690
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0008760057363407173,
+      "loss": 1.8641,
+      "step": 17700
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.0008758066652470294,
+      "loss": 1.6782,
+      "step": 17710
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0008756074571367203,
+      "loss": 1.7397,
+      "step": 17720
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0008754081120824204,
+      "loss": 1.8227,
+      "step": 17730
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.0008752086301568094,
+      "loss": 1.776,
+      "step": 17740
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.000875009011432617,
+      "loss": 1.6865,
+      "step": 17750
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.0008748092559826228,
+      "loss": 1.8389,
+      "step": 17760
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.0008746093638796566,
+      "loss": 1.7185,
+      "step": 17770
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0008744093351965976,
+      "loss": 1.7295,
+      "step": 17780
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0008742091700063748,
+      "loss": 1.8283,
+      "step": 17790
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0008740088683819674,
+      "loss": 1.769,
+      "step": 17800
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.0008738084303964036,
+      "loss": 1.7363,
+      "step": 17810
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.0008736078561227623,
+      "loss": 1.579,
+      "step": 17820
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.0008734071456341712,
+      "loss": 1.8071,
+      "step": 17830
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0008732062990038079,
+      "loss": 1.6405,
+      "step": 17840
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.0008730053163049001,
+      "loss": 1.7147,
+      "step": 17850
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0008728041976107246,
+      "loss": 1.7477,
+      "step": 17860
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0008726029429946078,
+      "loss": 1.8003,
+      "step": 17870
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0008724015525299261,
+      "loss": 1.808,
+      "step": 17880
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.0008722000262901049,
+      "loss": 1.6048,
+      "step": 17890
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0008719983643486195,
+      "loss": 1.6745,
+      "step": 17900
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0008717965667789943,
+      "loss": 1.7592,
+      "step": 17910
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.0008715946336548035,
+      "loss": 1.6439,
+      "step": 17920
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.0008713925650496706,
+      "loss": 1.6228,
+      "step": 17930
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0008711903610372684,
+      "loss": 1.6367,
+      "step": 17940
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.0008709880216913192,
+      "loss": 1.6975,
+      "step": 17950
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.0008707855470855945,
+      "loss": 1.8129,
+      "step": 17960
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0008705829372939154,
+      "loss": 1.6713,
+      "step": 17970
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 0.0008703801923901518,
+      "loss": 1.7297,
+      "step": 17980
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0008701773124482233,
+      "loss": 1.7521,
+      "step": 17990
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.0008699742975420985,
+      "loss": 1.6632,
+      "step": 18000
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0008697711477457955,
+      "loss": 1.6765,
+      "step": 18010
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0008695678631333812,
+      "loss": 1.6975,
+      "step": 18020
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0008693644437789719,
+      "loss": 1.7092,
+      "step": 18030
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0008691608897567328,
+      "loss": 1.7355,
+      "step": 18040
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.0008689572011408785,
+      "loss": 1.7607,
+      "step": 18050
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.0008687533780056724,
+      "loss": 1.6662,
+      "step": 18060
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.0008685494204254274,
+      "loss": 1.7526,
+      "step": 18070
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.0008683453284745046,
+      "loss": 1.7676,
+      "step": 18080
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.000868141102227315,
+      "loss": 1.6147,
+      "step": 18090
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.0008679367417583176,
+      "loss": 1.6999,
+      "step": 18100
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.0008677322471420214,
+      "loss": 1.6868,
+      "step": 18110
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.0008675276184529837,
+      "loss": 1.71,
+      "step": 18120
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0008673228557658103,
+      "loss": 1.8405,
+      "step": 18130
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0008671179591551571,
+      "loss": 1.6646,
+      "step": 18140
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0008669129286957275,
+      "loss": 1.6315,
+      "step": 18150
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.0008667077644622743,
+      "loss": 1.7164,
+      "step": 18160
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.000866502466529599,
+      "loss": 1.736,
+      "step": 18170
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.0008662970349725521,
+      "loss": 1.6481,
+      "step": 18180
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0008660914698660325,
+      "loss": 1.917,
+      "step": 18190
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0008658857712849876,
+      "loss": 1.7755,
+      "step": 18200
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.0008656799393044141,
+      "loss": 1.8172,
+      "step": 18210
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.0008654739739993568,
+      "loss": 1.7424,
+      "step": 18220
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.0008652678754449093,
+      "loss": 1.5829,
+      "step": 18230
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0008650616437162139,
+      "loss": 1.7492,
+      "step": 18240
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.0008648552788884611,
+      "loss": 1.7261,
+      "step": 18250
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0008646487810368904,
+      "loss": 1.803,
+      "step": 18260
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0008644421502367893,
+      "loss": 1.7519,
+      "step": 18270
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0008642353865634941,
+      "loss": 1.8376,
+      "step": 18280
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0008640284900923896,
+      "loss": 1.6313,
+      "step": 18290
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.0008638214608989085,
+      "loss": 1.7558,
+      "step": 18300
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.000863614299058533,
+      "loss": 1.765,
+      "step": 18310
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.0008634070046467921,
+      "loss": 1.7094,
+      "step": 18320
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.0008631995777392645,
+      "loss": 1.6005,
+      "step": 18330
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.0008629920184115766,
+      "loss": 1.6209,
+      "step": 18340
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 0.000862784326739403,
+      "loss": 1.8374,
+      "step": 18350
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0008625765027984669,
+      "loss": 1.54,
+      "step": 18360
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0008623685466645392,
+      "loss": 1.7219,
+      "step": 18370
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0008621604584134399,
+      "loss": 1.6198,
+      "step": 18380
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0008619522381210359,
+      "loss": 1.7157,
+      "step": 18390
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0008617438858632436,
+      "loss": 1.6988,
+      "step": 18400
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0008615354017160264,
+      "loss": 1.7169,
+      "step": 18410
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0008613267857553963,
+      "loss": 1.5654,
+      "step": 18420
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 0.0008611180380574135,
+      "loss": 1.6486,
+      "step": 18430
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0008609091586981859,
+      "loss": 1.748,
+      "step": 18440
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0008607001477538696,
+      "loss": 1.7695,
+      "step": 18450
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0008604910053006684,
+      "loss": 1.6418,
+      "step": 18460
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0008602817314148346,
+      "loss": 1.6655,
+      "step": 18470
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.0008600723261726677,
+      "loss": 1.6663,
+      "step": 18480
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0008598627896505157,
+      "loss": 1.8274,
+      "step": 18490
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.0008596531219247742,
+      "loss": 1.6767,
+      "step": 18500
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0008594433230718866,
+      "loss": 1.8287,
+      "step": 18510
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0008592333931683443,
+      "loss": 1.7962,
+      "step": 18520
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0008590233322906862,
+      "loss": 1.7732,
+      "step": 18530
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.0008588131405154991,
+      "loss": 1.7255,
+      "step": 18540
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0008586028179194176,
+      "loss": 1.7039,
+      "step": 18550
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0008583923645791241,
+      "loss": 1.6842,
+      "step": 18560
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0008581817805713481,
+      "loss": 1.7112,
+      "step": 18570
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0008579710659728674,
+      "loss": 1.613,
+      "step": 18580
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.0008577602208605071,
+      "loss": 1.6483,
+      "step": 18590
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.0008575492453111399,
+      "loss": 1.7063,
+      "step": 18600
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0008573381394016862,
+      "loss": 1.7714,
+      "step": 18610
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0008571269032091136,
+      "loss": 1.7529,
+      "step": 18620
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0008569155368104375,
+      "loss": 1.6819,
+      "step": 18630
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.0008567040402827209,
+      "loss": 1.7647,
+      "step": 18640
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0008564924137030739,
+      "loss": 1.6556,
+      "step": 18650
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.0008562806571486539,
+      "loss": 1.5576,
+      "step": 18660
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0008560687706966663,
+      "loss": 1.654,
+      "step": 18670
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0008558567544243633,
+      "loss": 1.6531,
+      "step": 18680
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0008556446084090446,
+      "loss": 1.8011,
+      "step": 18690
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0008554323327280575,
+      "loss": 1.6978,
+      "step": 18700
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.0008552199274587959,
+      "loss": 1.7141,
+      "step": 18710
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.0008550073926787019,
+      "loss": 1.7312,
+      "step": 18720
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0008547947284652637,
+      "loss": 1.5014,
+      "step": 18730
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0008545819348960176,
+      "loss": 1.8483,
+      "step": 18740
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.0008543690120485464,
+      "loss": 1.6066,
+      "step": 18750
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.0008541559600004809,
+      "loss": 1.7953,
+      "step": 18760
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.000853942778829498,
+      "loss": 1.5955,
+      "step": 18770
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0008537294686133223,
+      "loss": 1.6985,
+      "step": 18780
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 0.0008535160294297253,
+      "loss": 1.8277,
+      "step": 18790
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0008533024613565256,
+      "loss": 1.6323,
+      "step": 18800
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.0008530887644715885,
+      "loss": 1.7315,
+      "step": 18810
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0008528749388528266,
+      "loss": 1.6013,
+      "step": 18820
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0008526609845781994,
+      "loss": 1.7087,
+      "step": 18830
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.000852446901725713,
+      "loss": 1.7171,
+      "step": 18840
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.0008522326903734205,
+      "loss": 1.733,
+      "step": 18850
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.0008520183505994226,
+      "loss": 1.7025,
+      "step": 18860
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 0.0008518038824818653,
+      "loss": 1.6708,
+      "step": 18870
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0008515892860989427,
+      "loss": 1.7206,
+      "step": 18880
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0008513745615288952,
+      "loss": 1.6166,
+      "step": 18890
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0008511597088500099,
+      "loss": 1.6784,
+      "step": 18900
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0008509447281406209,
+      "loss": 1.6183,
+      "step": 18910
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0008507296194791082,
+      "loss": 1.6644,
+      "step": 18920
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.0008505143829438995,
+      "loss": 1.9086,
+      "step": 18930
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0008502990186134684,
+      "loss": 1.8484,
+      "step": 18940
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.0008500835265663354,
+      "loss": 1.6922,
+      "step": 18950
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0008498679068810673,
+      "loss": 1.5593,
+      "step": 18960
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.0008496521596362777,
+      "loss": 1.7798,
+      "step": 18970
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0008494362849106267,
+      "loss": 1.7348,
+      "step": 18980
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0008492202827828208,
+      "loss": 1.8277,
+      "step": 18990
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.0008490041533316128,
+      "loss": 1.6788,
+      "step": 19000
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.0008487878966358022,
+      "loss": 1.7564,
+      "step": 19010
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.0008485715127742348,
+      "loss": 1.7309,
+      "step": 19020
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0008483550018258024,
+      "loss": 1.6923,
+      "step": 19030
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.0008481383638694439,
+      "loss": 1.7461,
+      "step": 19040
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.0008479215989841438,
+      "loss": 1.7594,
+      "step": 19050
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.000847704707248933,
+      "loss": 1.7689,
+      "step": 19060
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0008474876887428892,
+      "loss": 1.7072,
+      "step": 19070
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.0008472705435451356,
+      "loss": 1.739,
+      "step": 19080
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.000847053271734842,
+      "loss": 1.8056,
+      "step": 19090
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.0008468358733912242,
+      "loss": 1.6447,
+      "step": 19100
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0008466183485935444,
+      "loss": 1.6804,
+      "step": 19110
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0008464006974211105,
+      "loss": 1.6939,
+      "step": 19120
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0008461829199532768,
+      "loss": 1.6983,
+      "step": 19130
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.0008459650162694433,
+      "loss": 1.7643,
+      "step": 19140
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0008457469864490564,
+      "loss": 1.5992,
+      "step": 19150
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0008455288305716085,
+      "loss": 1.795,
+      "step": 19160
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0008453105487166374,
+      "loss": 1.6571,
+      "step": 19170
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0008450921409637275,
+      "loss": 1.6727,
+      "step": 19180
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0008448736073925087,
+      "loss": 1.8219,
+      "step": 19190
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.000844654948082657,
+      "loss": 1.6273,
+      "step": 19200
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.0008444361631138941,
+      "loss": 1.8513,
+      "step": 19210
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.0008442172525659875,
+      "loss": 1.7595,
+      "step": 19220
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0008439982165187505,
+      "loss": 1.7008,
+      "step": 19230
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.0008437790550520424,
+      "loss": 1.6767,
+      "step": 19240
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0008435597682457676,
+      "loss": 1.6715,
+      "step": 19250
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0008433403561798772,
+      "loss": 1.8977,
+      "step": 19260
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0008431208189343669,
+      "loss": 1.9688,
+      "step": 19270
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": {
+        "accuracy": 0.3761348897535668
+      },
+      "eval_f1": {
+        "f1": 0.3427169207054117
+      },
+      "eval_loss": 1.5583436489105225,
+      "eval_precision": {
+        "precision": 0.3765571735514816
+      },
+      "eval_recall": {
+        "recall": 0.3992290557525892
+      },
+      "eval_runtime": 189.3787,
+      "eval_samples_per_second": 20.356,
+      "eval_steps_per_second": 10.181,
+      "step": 19270
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0008429011565892786,
+      "loss": 1.7518,
+      "step": 19280
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.0008426813692246999,
+      "loss": 1.6519,
+      "step": 19290
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.0008424614569207638,
+      "loss": 1.6477,
+      "step": 19300
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.000842241419757649,
+      "loss": 1.7954,
+      "step": 19310
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0008420212578155793,
+      "loss": 1.6735,
+      "step": 19320
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0008418009711748242,
+      "loss": 1.5975,
+      "step": 19330
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0008415805599156992,
+      "loss": 1.7858,
+      "step": 19340
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.0008413600241185644,
+      "loss": 1.6696,
+      "step": 19350
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.000841139363863826,
+      "loss": 1.6722,
+      "step": 19360
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.000840918579231935,
+      "loss": 1.7499,
+      "step": 19370
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0008406976703033879,
+      "loss": 1.7552,
+      "step": 19380
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.000840476637158727,
+      "loss": 1.6308,
+      "step": 19390
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.0008402554798785395,
+      "loss": 1.573,
+      "step": 19400
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.0008400341985434575,
+      "loss": 1.5892,
+      "step": 19410
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.000839812793234159,
+      "loss": 1.5308,
+      "step": 19420
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.000839591264031367,
+      "loss": 1.7723,
+      "step": 19430
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0008393696110158495,
+      "loss": 1.5531,
+      "step": 19440
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0008391478342684195,
+      "loss": 1.6957,
+      "step": 19450
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.0008389259338699355,
+      "loss": 1.636,
+      "step": 19460
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.000838703909901301,
+      "loss": 1.8127,
+      "step": 19470
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.0008384817624434642,
+      "loss": 1.7549,
+      "step": 19480
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.0008382594915774189,
+      "loss": 1.6516,
+      "step": 19490
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.0008380370973842035,
+      "loss": 1.7396,
+      "step": 19500
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 0.0008378145799449013,
+      "loss": 1.6137,
+      "step": 19510
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0008375919393406408,
+      "loss": 1.7072,
+      "step": 19520
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.0008373691756525954,
+      "loss": 1.7045,
+      "step": 19530
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.000837146288961983,
+      "loss": 1.7213,
+      "step": 19540
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.0008369232793500668,
+      "loss": 1.5688,
+      "step": 19550
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.0008367001468981546,
+      "loss": 1.6494,
+      "step": 19560
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0008364768916875989,
+      "loss": 1.7834,
+      "step": 19570
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.0008362535137997973,
+      "loss": 1.5781,
+      "step": 19580
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0008360300133161918,
+      "loss": 1.5922,
+      "step": 19590
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.0008358063903182689,
+      "loss": 1.6863,
+      "step": 19600
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.0008355826448875605,
+      "loss": 1.7007,
+      "step": 19610
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.0008353587771056427,
+      "loss": 1.7944,
+      "step": 19620
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0008351347870541358,
+      "loss": 1.8503,
+      "step": 19630
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.0008349106748147054,
+      "loss": 1.6262,
+      "step": 19640
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0008346864404690612,
+      "loss": 1.7324,
+      "step": 19650
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0008344620840989577,
+      "loss": 1.6073,
+      "step": 19660
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.0008342376057861938,
+      "loss": 1.6837,
+      "step": 19670
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.0008340130056126125,
+      "loss": 1.6396,
+      "step": 19680
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.0008337882836601017,
+      "loss": 1.6971,
+      "step": 19690
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.0008335634400105936,
+      "loss": 1.6755,
+      "step": 19700
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0008333384747460647,
+      "loss": 1.7068,
+      "step": 19710
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.0008331133879485357,
+      "loss": 1.7389,
+      "step": 19720
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0008328881797000719,
+      "loss": 1.6065,
+      "step": 19730
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0008326628500827827,
+      "loss": 1.5788,
+      "step": 19740
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.0008324373991788217,
+      "loss": 1.7423,
+      "step": 19750
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.000832211827070387,
+      "loss": 1.5595,
+      "step": 19760
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0008319861338397205,
+      "loss": 1.6295,
+      "step": 19770
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.0008317603195691086,
+      "loss": 1.9106,
+      "step": 19780
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.0008315343843408816,
+      "loss": 1.6018,
+      "step": 19790
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.000831308328237414,
+      "loss": 1.7538,
+      "step": 19800
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.0008310821513411243,
+      "loss": 1.699,
+      "step": 19810
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0008308558537344753,
+      "loss": 1.7823,
+      "step": 19820
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0008306294354999733,
+      "loss": 1.7676,
+      "step": 19830
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0008304028967201691,
+      "loss": 1.5569,
+      "step": 19840
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.0008301762374776574,
+      "loss": 1.7127,
+      "step": 19850
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.0008299494578550763,
+      "loss": 1.6094,
+      "step": 19860
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.0008297225579351082,
+      "loss": 1.8603,
+      "step": 19870
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0008294955378004796,
+      "loss": 1.8029,
+      "step": 19880
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0008292683975339604,
+      "loss": 1.7637,
+      "step": 19890
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0008290411372183641,
+      "loss": 1.7702,
+      "step": 19900
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0008288137569365488,
+      "loss": 1.7252,
+      "step": 19910
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0008285862567714155,
+      "loss": 1.6501,
+      "step": 19920
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0008283586368059097,
+      "loss": 1.7447,
+      "step": 19930
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.0008281308971230196,
+      "loss": 1.6704,
+      "step": 19940
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.0008279030378057779,
+      "loss": 1.6243,
+      "step": 19950
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.0008276750589372604,
+      "loss": 1.7619,
+      "step": 19960
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.0008274469606005867,
+      "loss": 1.6063,
+      "step": 19970
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0008272187428789203,
+      "loss": 1.6336,
+      "step": 19980
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0008269904058554674,
+      "loss": 1.7225,
+      "step": 19990
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0008267619496134784,
+      "loss": 1.7391,
+      "step": 20000
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.000826533374236247,
+      "loss": 1.6353,
+      "step": 20010
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0008263046798071101,
+      "loss": 1.7558,
+      "step": 20020
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0008260758664094483,
+      "loss": 1.6647,
+      "step": 20030
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0008258469341266853,
+      "loss": 1.7037,
+      "step": 20040
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0008256178830422884,
+      "loss": 1.799,
+      "step": 20050
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.0008253887132397682,
+      "loss": 1.6546,
+      "step": 20060
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0008251594248026784,
+      "loss": 1.7493,
+      "step": 20070
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0008249300178146159,
+      "loss": 1.6069,
+      "step": 20080
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.000824700492359221,
+      "loss": 1.7261,
+      "step": 20090
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0008244708485201775,
+      "loss": 1.7366,
+      "step": 20100
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.0008242410863812117,
+      "loss": 1.7636,
+      "step": 20110
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 0.0008240112060260934,
+      "loss": 1.6129,
+      "step": 20120
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0008237812075386355,
+      "loss": 1.6077,
+      "step": 20130
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.0008235510910026939,
+      "loss": 1.6412,
+      "step": 20140
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.0008233208565021677,
+      "loss": 1.6965,
+      "step": 20150
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.0008230905041209986,
+      "loss": 1.6617,
+      "step": 20160
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0008228600339431718,
+      "loss": 1.7644,
+      "step": 20170
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.000822629446052715,
+      "loss": 1.5249,
+      "step": 20180
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.0008223987405336994,
+      "loss": 1.6829,
+      "step": 20190
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.0008221679174702382,
+      "loss": 1.7577,
+      "step": 20200
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.0008219369769464883,
+      "loss": 1.5796,
+      "step": 20210
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.0008217059190466489,
+      "loss": 1.6066,
+      "step": 20220
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0008214747438549622,
+      "loss": 1.6334,
+      "step": 20230
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0008212434514557133,
+      "loss": 1.7428,
+      "step": 20240
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0008210120419332298,
+      "loss": 1.6272,
+      "step": 20250
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.0008207805153718819,
+      "loss": 1.8874,
+      "step": 20260
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0008205488718560829,
+      "loss": 1.7643,
+      "step": 20270
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.0008203171114702884,
+      "loss": 1.7322,
+      "step": 20280
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.0008200852342989969,
+      "loss": 1.7636,
+      "step": 20290
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.0008198532404267488,
+      "loss": 1.747,
+      "step": 20300
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0008196211299381278,
+      "loss": 1.8149,
+      "step": 20310
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00081938890291776,
+      "loss": 1.507,
+      "step": 20320
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0008191565594503137,
+      "loss": 1.7385,
+      "step": 20330
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0008189240996204997,
+      "loss": 1.6561,
+      "step": 20340
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.0008186915235130714,
+      "loss": 1.6747,
+      "step": 20350
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0008184588312128245,
+      "loss": 1.7173,
+      "step": 20360
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.0008182260228045971,
+      "loss": 1.6156,
+      "step": 20370
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0008179930983732696,
+      "loss": 1.6668,
+      "step": 20380
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.0008177600580037647,
+      "loss": 1.5413,
+      "step": 20390
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0008175269017810472,
+      "loss": 1.7355,
+      "step": 20400
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0008172936297901246,
+      "loss": 1.7727,
+      "step": 20410
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0008170602421160463,
+      "loss": 1.6936,
+      "step": 20420
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.0008168267388439037,
+      "loss": 1.6081,
+      "step": 20430
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0008165931200588306,
+      "loss": 1.5196,
+      "step": 20440
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.0008163593858460027,
+      "loss": 1.7085,
+      "step": 20450
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0008161255362906384,
+      "loss": 1.6289,
+      "step": 20460
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0008158915714779975,
+      "loss": 1.6257,
+      "step": 20470
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.0008156574914933818,
+      "loss": 1.7933,
+      "step": 20480
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.0008154232964221356,
+      "loss": 1.6946,
+      "step": 20490
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.0008151889863496446,
+      "loss": 1.7884,
+      "step": 20500
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.0008149545613613372,
+      "loss": 1.5876,
+      "step": 20510
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0008147200215426826,
+      "loss": 1.7212,
+      "step": 20520
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0008144853669791926,
+      "loss": 1.6056,
+      "step": 20530
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0008142505977564209,
+      "loss": 1.5416,
+      "step": 20540
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0008140157139599627,
+      "loss": 1.5642,
+      "step": 20550
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0008137807156754553,
+      "loss": 1.5649,
+      "step": 20560
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 0.0008135456029885772,
+      "loss": 1.6559,
+      "step": 20570
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0008133103759850489,
+      "loss": 1.6801,
+      "step": 20580
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0008130750347506328,
+      "loss": 1.6896,
+      "step": 20590
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.0008128395793711329,
+      "loss": 1.6411,
+      "step": 20600
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0008126040099323943,
+      "loss": 1.6248,
+      "step": 20610
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0008123683265203042,
+      "loss": 1.6338,
+      "step": 20620
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0008121325292207912,
+      "loss": 1.5133,
+      "step": 20630
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0008118966181198257,
+      "loss": 1.5557,
+      "step": 20640
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.000811660593303419,
+      "loss": 1.7146,
+      "step": 20650
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0008114244548576243,
+      "loss": 1.5742,
+      "step": 20660
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.000811188202868536,
+      "loss": 1.6909,
+      "step": 20670
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.0008109518374222902,
+      "loss": 1.4557,
+      "step": 20680
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0008107153586050641,
+      "loss": 1.6167,
+      "step": 20690
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.0008104787665030762,
+      "loss": 1.5466,
+      "step": 20700
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0008102420612025866,
+      "loss": 1.7289,
+      "step": 20710
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0008100052427898963,
+      "loss": 1.6483,
+      "step": 20720
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.000809768311351348,
+      "loss": 1.5907,
+      "step": 20730
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.000809531266973325,
+      "loss": 1.5286,
+      "step": 20740
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0008092941097422523,
+      "loss": 1.6909,
+      "step": 20750
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.0008090568397445957,
+      "loss": 1.5898,
+      "step": 20760
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.0008088194570668626,
+      "loss": 1.5395,
+      "step": 20770
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.0008085819617956009,
+      "loss": 1.9351,
+      "step": 20780
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0008083443540173998,
+      "loss": 1.5448,
+      "step": 20790
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.0008081066338188895,
+      "loss": 1.5762,
+      "step": 20800
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.0008078688012867413,
+      "loss": 1.629,
+      "step": 20810
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.0008076308565076674,
+      "loss": 1.4561,
+      "step": 20820
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0008073927995684207,
+      "loss": 1.6926,
+      "step": 20830
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.0008071546305557955,
+      "loss": 1.6087,
+      "step": 20840
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0008069163495566263,
+      "loss": 1.6436,
+      "step": 20850
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.000806677956657789,
+      "loss": 1.5916,
+      "step": 20860
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0008064394519462,
+      "loss": 1.5723,
+      "step": 20870
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0008062008355088166,
+      "loss": 1.5607,
+      "step": 20880
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.0008059621074326367,
+      "loss": 1.7147,
+      "step": 20890
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.0008057232678046988,
+      "loss": 1.6282,
+      "step": 20900
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.0008054843167120826,
+      "loss": 1.6724,
+      "step": 20910
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.000805245254241908,
+      "loss": 1.4804,
+      "step": 20920
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0008050060804813355,
+      "loss": 1.6362,
+      "step": 20930
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.0008047667955175662,
+      "loss": 1.6929,
+      "step": 20940
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.0008045273994378419,
+      "loss": 1.4288,
+      "step": 20950
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0008042878923294448,
+      "loss": 1.4788,
+      "step": 20960
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.0008040482742796976,
+      "loss": 1.9197,
+      "step": 20970
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0008038085453759636,
+      "loss": 1.6891,
+      "step": 20980
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0008035687057056462,
+      "loss": 1.5828,
+      "step": 20990
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0008033287553561895,
+      "loss": 1.5094,
+      "step": 21000
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0008030886944150775,
+      "loss": 1.8679,
+      "step": 21010
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0008028485229698352,
+      "loss": 1.6638,
+      "step": 21020
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.0008026082411080273,
+      "loss": 1.5102,
+      "step": 21030
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.000802367848917259,
+      "loss": 1.4554,
+      "step": 21040
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.000802127346485176,
+      "loss": 1.7006,
+      "step": 21050
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.0008018867338994634,
+      "loss": 1.5274,
+      "step": 21060
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.0008016460112478471,
+      "loss": 1.4812,
+      "step": 21070
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0008014051786180931,
+      "loss": 1.712,
+      "step": 21080
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.0008011642360980076,
+      "loss": 1.5969,
+      "step": 21090
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0008009231837754363,
+      "loss": 1.5231,
+      "step": 21100
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.0008006820217382652,
+      "loss": 1.4777,
+      "step": 21110
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0008004407500744207,
+      "loss": 1.4811,
+      "step": 21120
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0008001993688718687,
+      "loss": 1.5022,
+      "step": 21130
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.0007999578782186152,
+      "loss": 1.51,
+      "step": 21140
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0007997162782027061,
+      "loss": 1.7038,
+      "step": 21150
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0007994745689122271,
+      "loss": 1.5184,
+      "step": 21160
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.0007992327504353038,
+      "loss": 1.5444,
+      "step": 21170
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.000798990822860102,
+      "loss": 1.627,
+      "step": 21180
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.0007987487862748262,
+      "loss": 1.5298,
+      "step": 21190
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": {
+        "accuracy": 0.46718547341115435
+      },
+      "eval_f1": {
+        "f1": 0.4188714217488438
+      },
+      "eval_loss": 1.3465330600738525,
+      "eval_precision": {
+        "precision": 0.41426640127137376
+      },
+      "eval_recall": {
+        "recall": 0.4948604123766437
+      },
+      "eval_runtime": 194.4976,
+      "eval_samples_per_second": 19.82,
+      "eval_steps_per_second": 9.913,
+      "step": 21197
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.0007985066407677219,
+      "loss": 1.5332,
+      "step": 21200
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0007982643864270735,
+      "loss": 1.5636,
+      "step": 21210
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0007980220233412055,
+      "loss": 1.5377,
+      "step": 21220
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0007977795515984817,
+      "loss": 1.6112,
+      "step": 21230
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0007975369712873057,
+      "loss": 1.6851,
+      "step": 21240
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0007972942824961207,
+      "loss": 1.5181,
+      "step": 21250
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0007970514853134095,
+      "loss": 1.6028,
+      "step": 21260
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0007968085798276943,
+      "loss": 1.6585,
+      "step": 21270
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 0.0007965655661275368,
+      "loss": 1.509,
+      "step": 21280
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0007963224443015382,
+      "loss": 1.6334,
+      "step": 21290
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.0007960792144383391,
+      "loss": 1.5433,
+      "step": 21300
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.0007958358766266197,
+      "loss": 1.5391,
+      "step": 21310
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.000795592430955099,
+      "loss": 1.5199,
+      "step": 21320
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0007953488775125358,
+      "loss": 1.5157,
+      "step": 21330
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0007951052163877281,
+      "loss": 1.5375,
+      "step": 21340
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0007948614476695133,
+      "loss": 1.4858,
+      "step": 21350
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.0007946175714467678,
+      "loss": 1.4889,
+      "step": 21360
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0007943735878084071,
+      "loss": 1.549,
+      "step": 21370
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.0007941294968433858,
+      "loss": 1.5715,
+      "step": 21380
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.0007938852986406985,
+      "loss": 1.6116,
+      "step": 21390
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0007936409932893778,
+      "loss": 1.4787,
+      "step": 21400
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0007933965808784957,
+      "loss": 1.6186,
+      "step": 21410
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0007931520614971638,
+      "loss": 1.5989,
+      "step": 21420
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.0007929074352345318,
+      "loss": 1.6194,
+      "step": 21430
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.0007926627021797891,
+      "loss": 1.7698,
+      "step": 21440
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 0.0007924178624221636,
+      "loss": 1.5963,
+      "step": 21450
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.0007921729160509225,
+      "loss": 1.5013,
+      "step": 21460
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.0007919278631553713,
+      "loss": 1.5865,
+      "step": 21470
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0007916827038248547,
+      "loss": 1.4129,
+      "step": 21480
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0007914374381487564,
+      "loss": 1.5089,
+      "step": 21490
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0007911920662164985,
+      "loss": 1.4169,
+      "step": 21500
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0007909465881175419,
+      "loss": 1.5392,
+      "step": 21510
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0007907010039413865,
+      "loss": 1.3947,
+      "step": 21520
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.0007904553137775702,
+      "loss": 1.5883,
+      "step": 21530
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0007902095177156707,
+      "loss": 1.593,
+      "step": 21540
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.0007899636158453032,
+      "loss": 1.606,
+      "step": 21550
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0007897176082561219,
+      "loss": 1.5666,
+      "step": 21560
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0007894714950378197,
+      "loss": 1.3754,
+      "step": 21570
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0007892252762801276,
+      "loss": 1.5479,
+      "step": 21580
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0007889789520728155,
+      "loss": 1.3955,
+      "step": 21590
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0007887325225056916,
+      "loss": 1.7297,
+      "step": 21600
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.0007884859876686024,
+      "loss": 1.5064,
+      "step": 21610
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0007882393476514329,
+      "loss": 1.4094,
+      "step": 21620
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0007879926025441066,
+      "loss": 1.4921,
+      "step": 21630
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 0.0007877457524365848,
+      "loss": 1.4637,
+      "step": 21640
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0007874987974188675,
+      "loss": 1.4903,
+      "step": 21650
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.0007872517375809929,
+      "loss": 1.4895,
+      "step": 21660
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.0007870045730130375,
+      "loss": 1.5463,
+      "step": 21670
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.0007867573038051156,
+      "loss": 1.6203,
+      "step": 21680
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0007865099300473804,
+      "loss": 1.7123,
+      "step": 21690
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0007862624518300222,
+      "loss": 1.884,
+      "step": 21700
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.0007860148692432703,
+      "loss": 2.0276,
+      "step": 21710
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.0007857671823773914,
+      "loss": 1.6984,
+      "step": 21720
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0007855193913226905,
+      "loss": 1.6476,
+      "step": 21730
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0007852714961695109,
+      "loss": 1.615,
+      "step": 21740
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0007850234970082333,
+      "loss": 1.5339,
+      "step": 21750
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0007847753939292764,
+      "loss": 1.6289,
+      "step": 21760
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.0007845271870230972,
+      "loss": 1.5788,
+      "step": 21770
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.0007842788763801901,
+      "loss": 1.3275,
+      "step": 21780
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0007840304620910877,
+      "loss": 1.5183,
+      "step": 21790
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0007837819442463601,
+      "loss": 1.3272,
+      "step": 21800
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.0007835333229366154,
+      "loss": 1.3686,
+      "step": 21810
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.000783284598252499,
+      "loss": 1.4811,
+      "step": 21820
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.0007830357702846944,
+      "loss": 1.3942,
+      "step": 21830
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.0007827868391239227,
+      "loss": 1.6312,
+      "step": 21840
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0007825378048609427,
+      "loss": 1.618,
+      "step": 21850
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.0007822886675865503,
+      "loss": 1.4124,
+      "step": 21860
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.0007820394273915794,
+      "loss": 1.5759,
+      "step": 21870
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.0007817900843669015,
+      "loss": 1.6723,
+      "step": 21880
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0007815406386034252,
+      "loss": 1.414,
+      "step": 21890
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0007812910901920969,
+      "loss": 1.4628,
+      "step": 21900
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.0007810414392239002,
+      "loss": 1.5615,
+      "step": 21910
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0007807916857898564,
+      "loss": 1.5874,
+      "step": 21920
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.0007805418299810236,
+      "loss": 1.5437,
+      "step": 21930
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.000780291871888498,
+      "loss": 1.4822,
+      "step": 21940
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0007800418116034124,
+      "loss": 1.5061,
+      "step": 21950
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.000779791649216937,
+      "loss": 1.4001,
+      "step": 21960
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.0007795413848202797,
+      "loss": 1.6048,
+      "step": 21970
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.000779291018504685,
+      "loss": 1.4148,
+      "step": 21980
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.0007790405503614349,
+      "loss": 1.5085,
+      "step": 21990
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0007787899804818483,
+      "loss": 1.4652,
+      "step": 22000
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.0007785393089572814,
+      "loss": 1.7414,
+      "step": 22010
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0007782885358791274,
+      "loss": 1.6251,
+      "step": 22020
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.0007780376613388165,
+      "loss": 1.3955,
+      "step": 22030
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0007777866854278157,
+      "loss": 1.3389,
+      "step": 22040
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0007775356082376296,
+      "loss": 1.5705,
+      "step": 22050
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0007772844298597988,
+      "loss": 1.6243,
+      "step": 22060
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0007770331503859018,
+      "loss": 1.5605,
+      "step": 22070
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0007767817699075528,
+      "loss": 1.5934,
+      "step": 22080
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.0007765302885164038,
+      "loss": 1.5003,
+      "step": 22090
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0007762787063041434,
+      "loss": 1.5809,
+      "step": 22100
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.0007760270233624964,
+      "loss": 1.6217,
+      "step": 22110
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.0007757752397832252,
+      "loss": 1.7394,
+      "step": 22120
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.000775523355658128,
+      "loss": 1.6742,
+      "step": 22130
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.0007752713710790404,
+      "loss": 1.5814,
+      "step": 22140
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.0007750192861378341,
+      "loss": 1.5658,
+      "step": 22150
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0007747671009264178,
+      "loss": 1.5133,
+      "step": 22160
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.0007745148155367364,
+      "loss": 1.4386,
+      "step": 22170
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.0007742624300607715,
+      "loss": 1.5617,
+      "step": 22180
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0007740099445905411,
+      "loss": 1.6447,
+      "step": 22190
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0007737573592181,
+      "loss": 1.4193,
+      "step": 22200
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0007735046740355389,
+      "loss": 1.5798,
+      "step": 22210
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.0007732518891349851,
+      "loss": 1.5869,
+      "step": 22220
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.0007729990046086026,
+      "loss": 1.7619,
+      "step": 22230
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.0007727460205485911,
+      "loss": 1.783,
+      "step": 22240
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.0007724929370471872,
+      "loss": 1.7116,
+      "step": 22250
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.0007722397541966632,
+      "loss": 1.6417,
+      "step": 22260
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0007719864720893281,
+      "loss": 1.5286,
+      "step": 22270
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0007717330908175268,
+      "loss": 1.5593,
+      "step": 22280
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0007714796104736404,
+      "loss": 1.9002,
+      "step": 22290
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0007712260311500863,
+      "loss": 1.548,
+      "step": 22300
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0007709723529393175,
+      "loss": 1.5712,
+      "step": 22310
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.0007707185759338237,
+      "loss": 1.5299,
+      "step": 22320
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0007704647002261303,
+      "loss": 1.7557,
+      "step": 22330
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.0007702107259087987,
+      "loss": 1.6734,
+      "step": 22340
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0007699566530744262,
+      "loss": 1.6217,
+      "step": 22350
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 0.0007697024818156459,
+      "loss": 1.6293,
+      "step": 22360
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0007694482122251272,
+      "loss": 1.8722,
+      "step": 22370
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0007691938443955751,
+      "loss": 1.5977,
+      "step": 22380
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.0007689393784197303,
+      "loss": 1.5815,
+      "step": 22390
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.0007686848143903693,
+      "loss": 1.6873,
+      "step": 22400
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.0007684301524003049,
+      "loss": 1.5311,
+      "step": 22410
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.0007681753925423846,
+      "loss": 1.7016,
+      "step": 22420
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0007679205349094927,
+      "loss": 1.4795,
+      "step": 22430
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0007676655795945484,
+      "loss": 1.6081,
+      "step": 22440
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.0007674105266905065,
+      "loss": 1.5185,
+      "step": 22450
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.000767155376290358,
+      "loss": 1.4479,
+      "step": 22460
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.0007669001284871286,
+      "loss": 1.6247,
+      "step": 22470
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0007666447833738803,
+      "loss": 1.534,
+      "step": 22480
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00076638934104371,
+      "loss": 1.4489,
+      "step": 22490
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0007661338015897505,
+      "loss": 1.5347,
+      "step": 22500
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0007658781651051697,
+      "loss": 1.5885,
+      "step": 22510
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0007656224316831711,
+      "loss": 1.4377,
+      "step": 22520
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 0.0007653666014169931,
+      "loss": 1.5669,
+      "step": 22530
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0007651106743999098,
+      "loss": 1.4317,
+      "step": 22540
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0007648546507252308,
+      "loss": 1.6138,
+      "step": 22550
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.0007645985304863003,
+      "loss": 1.5313,
+      "step": 22560
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.0007643423137764982,
+      "loss": 1.5465,
+      "step": 22570
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0007640860006892394,
+      "loss": 1.7053,
+      "step": 22580
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.000763829591317974,
+      "loss": 1.4791,
+      "step": 22590
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0007635730857561871,
+      "loss": 1.4776,
+      "step": 22600
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.0007633164840973989,
+      "loss": 1.4011,
+      "step": 22610
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0007630597864351645,
+      "loss": 1.4728,
+      "step": 22620
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0007628029928630747,
+      "loss": 1.4482,
+      "step": 22630
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0007625461034747541,
+      "loss": 1.4318,
+      "step": 22640
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.0007622891183638633,
+      "loss": 1.6094,
+      "step": 22650
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0007620320376240972,
+      "loss": 1.6027,
+      "step": 22660
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0007617748613491858,
+      "loss": 1.4384,
+      "step": 22670
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.0007615175896328938,
+      "loss": 1.4359,
+      "step": 22680
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.0007612602225690209,
+      "loss": 1.5231,
+      "step": 22690
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.0007610027602514011,
+      "loss": 1.5305,
+      "step": 22700
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.000760745202773904,
+      "loss": 1.5439,
+      "step": 22710
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.0007604875502304327,
+      "loss": 1.5968,
+      "step": 22720
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0007602298027149262,
+      "loss": 1.3715,
+      "step": 22730
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.0007599719603213572,
+      "loss": 1.4726,
+      "step": 22740
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.0007597140231437333,
+      "loss": 1.5915,
+      "step": 22750
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.000759455991276097,
+      "loss": 1.6387,
+      "step": 22760
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0007591978648125246,
+      "loss": 1.7036,
+      "step": 22770
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.0007589396438471276,
+      "loss": 1.5881,
+      "step": 22780
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0007586813284740514,
+      "loss": 1.5138,
+      "step": 22790
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 0.0007584229187874765,
+      "loss": 1.6262,
+      "step": 22800
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.0007581644148816169,
+      "loss": 1.52,
+      "step": 22810
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.0007579058168507217,
+      "loss": 1.8329,
+      "step": 22820
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0007576471247890739,
+      "loss": 1.4998,
+      "step": 22830
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.000757388338790991,
+      "loss": 1.5454,
+      "step": 22840
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.0007571294589508247,
+      "loss": 1.2622,
+      "step": 22850
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.0007568704853629604,
+      "loss": 1.4487,
+      "step": 22860
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.0007566114181218189,
+      "loss": 1.6154,
+      "step": 22870
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.0007563522573218539,
+      "loss": 1.4601,
+      "step": 22880
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.0007560930030575538,
+      "loss": 1.589,
+      "step": 22890
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.000755833655423441,
+      "loss": 1.682,
+      "step": 22900
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0007555742145140722,
+      "loss": 1.9207,
+      "step": 22910
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0007553146804240374,
+      "loss": 1.8066,
+      "step": 22920
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0007550550532479612,
+      "loss": 1.7496,
+      "step": 22930
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0007547953330805021,
+      "loss": 1.6789,
+      "step": 22940
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.0007545355200163522,
+      "loss": 1.6205,
+      "step": 22950
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.0007542756141502376,
+      "loss": 1.4813,
+      "step": 22960
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.0007540156155769182,
+      "loss": 1.3989,
+      "step": 22970
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.000753755524391188,
+      "loss": 1.6348,
+      "step": 22980
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0007534953406878741,
+      "loss": 1.5096,
+      "step": 22990
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0007532350645618383,
+      "loss": 1.682,
+      "step": 23000
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.0007529746961079753,
+      "loss": 1.7338,
+      "step": 23010
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0007527142354212136,
+      "loss": 1.6581,
+      "step": 23020
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.0007524536825965154,
+      "loss": 2.0326,
+      "step": 23030
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0007521930377288768,
+      "loss": 1.8598,
+      "step": 23040
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0007519323009133272,
+      "loss": 1.7569,
+      "step": 23050
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0007516714722449291,
+      "loss": 1.5758,
+      "step": 23060
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.0007514105518187794,
+      "loss": 1.8387,
+      "step": 23070
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.0007511495397300075,
+      "loss": 1.7581,
+      "step": 23080
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.0007508884360737774,
+      "loss": 1.8619,
+      "step": 23090
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.0007506272409452848,
+      "loss": 1.7843,
+      "step": 23100
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.0007503659544397603,
+      "loss": 1.8354,
+      "step": 23110
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0007501045766524672,
+      "loss": 1.7964,
+      "step": 23120
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": {
+        "accuracy": 0.2059662775616083
+      },
+      "eval_f1": {
+        "f1": 0.15046942523030551
+      },
+      "eval_loss": 2.1154468059539795,
+      "eval_precision": {
+        "precision": 0.24607297755729998
+      },
+      "eval_recall": {
+        "recall": 0.1965173812427065
+      },
+      "eval_runtime": 178.6004,
+      "eval_samples_per_second": 21.584,
+      "eval_steps_per_second": 10.795,
+      "step": 23124
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.000749843107678702,
+      "loss": 1.7875,
+      "step": 23130
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.0007495815476137946,
+      "loss": 1.9281,
+      "step": 23140
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.000749319896553108,
+      "loss": 1.7574,
+      "step": 23150
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0007490581545920388,
+      "loss": 1.7594,
+      "step": 23160
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.0007487963218260156,
+      "loss": 1.8301,
+      "step": 23170
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0007485343983505018,
+      "loss": 1.6574,
+      "step": 23180
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.0007482723842609924,
+      "loss": 1.8448,
+      "step": 23190
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0007480102796530161,
+      "loss": 1.9692,
+      "step": 23200
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0007477480846221346,
+      "loss": 1.9614,
+      "step": 23210
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.0007474857992639425,
+      "loss": 1.8287,
+      "step": 23220
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0007472234236740672,
+      "loss": 1.9055,
+      "step": 23230
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 0.0007469609579481689,
+      "loss": 1.9807,
+      "step": 23240
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0007466984021819412,
+      "loss": 1.8013,
+      "step": 23250
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0007464357564711102,
+      "loss": 1.9062,
+      "step": 23260
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0007461730209114348,
+      "loss": 1.9173,
+      "step": 23270
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.0007459101955987062,
+      "loss": 1.9275,
+      "step": 23280
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0007456472806287493,
+      "loss": 1.8274,
+      "step": 23290
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.0007453842760974207,
+      "loss": 1.8803,
+      "step": 23300
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0007451211821006107,
+      "loss": 1.9396,
+      "step": 23310
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.0007448579987342412,
+      "loss": 1.86,
+      "step": 23320
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0007445947260942672,
+      "loss": 1.8808,
+      "step": 23330
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0007443313642766759,
+      "loss": 1.8754,
+      "step": 23340
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.0007440679133774877,
+      "loss": 1.8806,
+      "step": 23350
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.0007438043734927549,
+      "loss": 1.7697,
+      "step": 23360
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.0007435407447185622,
+      "loss": 1.9435,
+      "step": 23370
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.0007432770271510274,
+      "loss": 1.8623,
+      "step": 23380
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0007430132208862996,
+      "loss": 1.8164,
+      "step": 23390
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.0007427493260205613,
+      "loss": 1.8316,
+      "step": 23400
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0007424853426500264,
+      "loss": 1.9059,
+      "step": 23410
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.0007422212708709418,
+      "loss": 1.781,
+      "step": 23420
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.0007419571107795861,
+      "loss": 1.8236,
+      "step": 23430
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.0007416928624722706,
+      "loss": 1.7679,
+      "step": 23440
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0007414285260453383,
+      "loss": 1.7847,
+      "step": 23450
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0007411641015951645,
+      "loss": 1.8191,
+      "step": 23460
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0007408995892181565,
+      "loss": 1.8611,
+      "step": 23470
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0007406349890107539,
+      "loss": 1.8868,
+      "step": 23480
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0007403703010694281,
+      "loss": 1.7893,
+      "step": 23490
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.0007401055254906829,
+      "loss": 1.7112,
+      "step": 23500
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.0007398406623710533,
+      "loss": 1.8494,
+      "step": 23510
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0007395757118071067,
+      "loss": 1.7485,
+      "step": 23520
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.0007393106738954426,
+      "loss": 1.8959,
+      "step": 23530
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0007390455487326919,
+      "loss": 1.9892,
+      "step": 23540
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.0007387803364155175,
+      "loss": 1.8214,
+      "step": 23550
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0007385150370406138,
+      "loss": 1.861,
+      "step": 23560
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0007382496507047075,
+      "loss": 1.8838,
+      "step": 23570
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.0007379841775045567,
+      "loss": 1.8722,
+      "step": 23580
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.000737718617536951,
+      "loss": 1.7656,
+      "step": 23590
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.0007374529708987119,
+      "loss": 1.9958,
+      "step": 23600
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 0.0007371872376866924,
+      "loss": 1.7937,
+      "step": 23610
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0007369214179977771,
+      "loss": 1.9022,
+      "step": 23620
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.000736655511928882,
+      "loss": 1.8255,
+      "step": 23630
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0007363895195769549,
+      "loss": 1.7848,
+      "step": 23640
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0007361234410389745,
+      "loss": 1.7972,
+      "step": 23650
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0007358572764119516,
+      "loss": 1.7616,
+      "step": 23660
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.000735591025792928,
+      "loss": 1.7677,
+      "step": 23670
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.0007353246892789768,
+      "loss": 1.8088,
+      "step": 23680
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 0.0007350582669672025,
+      "loss": 1.8179,
+      "step": 23690
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.000734791758954741,
+      "loss": 1.7793,
+      "step": 23700
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.0007345251653387594,
+      "loss": 1.8058,
+      "step": 23710
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0007342584862164559,
+      "loss": 1.9176,
+      "step": 23720
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.0007339917216850599,
+      "loss": 1.9782,
+      "step": 23730
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.000733724871841832,
+      "loss": 1.8446,
+      "step": 23740
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.000733457936784064,
+      "loss": 1.8099,
+      "step": 23750
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0007331909166090787,
+      "loss": 1.7463,
+      "step": 23760
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.0007329238114142296,
+      "loss": 1.8202,
+      "step": 23770
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.0007326566212969016,
+      "loss": 1.7897,
+      "step": 23780
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.0007323893463545106,
+      "loss": 1.889,
+      "step": 23790
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.0007321219866845031,
+      "loss": 1.6414,
+      "step": 23800
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.0007318545423843568,
+      "loss": 1.8099,
+      "step": 23810
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00073158701355158,
+      "loss": 1.897,
+      "step": 23820
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.000731319400283712,
+      "loss": 1.8179,
+      "step": 23830
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.0007310517026783228,
+      "loss": 1.7562,
+      "step": 23840
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.0007307839208330132,
+      "loss": 1.853,
+      "step": 23850
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.0007305160548454147,
+      "loss": 1.7114,
+      "step": 23860
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.0007302481048131893,
+      "loss": 1.6786,
+      "step": 23870
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.0007299800708340301,
+      "loss": 1.5227,
+      "step": 23880
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.0007297119530056602,
+      "loss": 1.8636,
+      "step": 23890
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.0007294437514258341,
+      "loss": 1.7088,
+      "step": 23900
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.0007291754661923358,
+      "loss": 1.7058,
+      "step": 23910
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.0007289070974029803,
+      "loss": 1.8528,
+      "step": 23920
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0007286386451556134,
+      "loss": 1.6828,
+      "step": 23930
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0007283701095481109,
+      "loss": 1.8285,
+      "step": 23940
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0007281014906783792,
+      "loss": 1.7722,
+      "step": 23950
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0007278327886443546,
+      "loss": 1.8169,
+      "step": 23960
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.0007275640035440044,
+      "loss": 1.8786,
+      "step": 23970
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.0007272951354753258,
+      "loss": 1.8595,
+      "step": 23980
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0007270261845363464,
+      "loss": 1.8211,
+      "step": 23990
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0007267571508251237,
+      "loss": 1.7611,
+      "step": 24000
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0007264880344397458,
+      "loss": 1.8088,
+      "step": 24010
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.0007262188354783308,
+      "loss": 1.8177,
+      "step": 24020
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.0007259495540390268,
+      "loss": 1.7323,
+      "step": 24030
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.0007256801902200119,
+      "loss": 1.7716,
+      "step": 24040
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 0.0007254107441194943,
+      "loss": 1.7237,
+      "step": 24050
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.0007251412158357126,
+      "loss": 1.6832,
+      "step": 24060
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.0007248716054669347,
+      "loss": 1.709,
+      "step": 24070
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.000724601913111459,
+      "loss": 1.58,
+      "step": 24080
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.0007243321388676134,
+      "loss": 1.7091,
+      "step": 24090
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.000724062282833756,
+      "loss": 1.7448,
+      "step": 24100
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.0007237923451082742,
+      "loss": 1.6554,
+      "step": 24110
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.0007235223257895857,
+      "loss": 1.7116,
+      "step": 24120
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 0.0007232522249761378,
+      "loss": 1.7339,
+      "step": 24130
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.0007229820427664074,
+      "loss": 1.6333,
+      "step": 24140
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.0007227117792589011,
+      "loss": 1.7736,
+      "step": 24150
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0007224414345521553,
+      "loss": 1.666,
+      "step": 24160
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0007221710087447357,
+      "loss": 1.8743,
+      "step": 24170
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.0007219005019352379,
+      "loss": 1.7471,
+      "step": 24180
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.0007216299142222868,
+      "loss": 1.8774,
+      "step": 24190
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.0007213592457045371,
+      "loss": 1.6998,
+      "step": 24200
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.0007210884964806727,
+      "loss": 1.6385,
+      "step": 24210
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.0007208176666494066,
+      "loss": 1.6912,
+      "step": 24220
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.000720546756309482,
+      "loss": 1.5092,
+      "step": 24230
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.0007202757655596708,
+      "loss": 1.6444,
+      "step": 24240
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.0007200046944987746,
+      "loss": 1.7133,
+      "step": 24250
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.000719733543225624,
+      "loss": 1.7087,
+      "step": 24260
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.0007194623118390789,
+      "loss": 1.7841,
+      "step": 24270
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.0007191910004380283,
+      "loss": 1.7016,
+      "step": 24280
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.0007189196091213911,
+      "loss": 1.8609,
+      "step": 24290
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.0007186481379881142,
+      "loss": 1.7609,
+      "step": 24300
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0007183765871371746,
+      "loss": 1.7378,
+      "step": 24310
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.0007181049566675777,
+      "loss": 1.8348,
+      "step": 24320
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.0007178332466783582,
+      "loss": 1.6605,
+      "step": 24330
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.0007175614572685799,
+      "loss": 1.7238,
+      "step": 24340
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.0007172895885373352,
+      "loss": 1.6721,
+      "step": 24350
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.0007170176405837459,
+      "loss": 1.5811,
+      "step": 24360
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.000716745613506962,
+      "loss": 1.5636,
+      "step": 24370
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.0007164735074061632,
+      "loss": 1.5694,
+      "step": 24380
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0007162013223805574,
+      "loss": 1.6894,
+      "step": 24390
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.0007159290585293817,
+      "loss": 1.5094,
+      "step": 24400
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.0007156567159519012,
+      "loss": 1.7822,
+      "step": 24410
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.0007153842947474105,
+      "loss": 1.5964,
+      "step": 24420
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.0007151117950152327,
+      "loss": 1.63,
+      "step": 24430
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.0007148392168547191,
+      "loss": 1.6766,
+      "step": 24440
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0007145665603652501,
+      "loss": 1.6444,
+      "step": 24450
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0007142938256462344,
+      "loss": 1.7126,
+      "step": 24460
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.0007140210127971093,
+      "loss": 1.5439,
+      "step": 24470
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.0007137481219173405,
+      "loss": 1.5264,
+      "step": 24480
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.0007134751531064222,
+      "loss": 1.5395,
+      "step": 24490
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.0007132021064638769,
+      "loss": 1.5994,
+      "step": 24500
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.000712928982089256,
+      "loss": 1.5844,
+      "step": 24510
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.0007126557800821385,
+      "loss": 1.6732,
+      "step": 24520
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.0007123825005421321,
+      "loss": 1.7349,
+      "step": 24530
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.0007121091435688726,
+      "loss": 1.5828,
+      "step": 24540
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.0007118357092620245,
+      "loss": 1.4401,
+      "step": 24550
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0007115621977212797,
+      "loss": 1.6509,
+      "step": 24560
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0007112886090463589,
+      "loss": 1.7305,
+      "step": 24570
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.000711014943337011,
+      "loss": 1.578,
+      "step": 24580
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.0007107412006930121,
+      "loss": 1.456,
+      "step": 24590
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0007104673812141675,
+      "loss": 1.5958,
+      "step": 24600
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.0007101934850003098,
+      "loss": 1.5158,
+      "step": 24610
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.0007099195121512996,
+      "loss": 1.7054,
+      "step": 24620
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.0007096454627670256,
+      "loss": 1.7499,
+      "step": 24630
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0007093713369474046,
+      "loss": 1.557,
+      "step": 24640
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0007090971347923809,
+      "loss": 1.5698,
+      "step": 24650
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0007088228564019268,
+      "loss": 1.6833,
+      "step": 24660
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.0007085485018760424,
+      "loss": 1.5941,
+      "step": 24670
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.0007082740713147555,
+      "loss": 1.6177,
+      "step": 24680
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.0007079995648181219,
+      "loss": 1.4892,
+      "step": 24690
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0007077249824862245,
+      "loss": 1.696,
+      "step": 24700
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0007074503244191745,
+      "loss": 1.791,
+      "step": 24710
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0007071755907171101,
+      "loss": 1.6303,
+      "step": 24720
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.0007069007814801977,
+      "loss": 1.488,
+      "step": 24730
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.0007066258968086307,
+      "loss": 1.6009,
+      "step": 24740
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.0007063509368026305,
+      "loss": 1.677,
+      "step": 24750
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.0007060759015624454,
+      "loss": 1.7456,
+      "step": 24760
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 0.0007058007911883515,
+      "loss": 1.623,
+      "step": 24770
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0007055256057806522,
+      "loss": 1.6232,
+      "step": 24780
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0007052503454396786,
+      "loss": 1.7356,
+      "step": 24790
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0007049750102657884,
+      "loss": 1.5552,
+      "step": 24800
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.0007046996003593671,
+      "loss": 1.5545,
+      "step": 24810
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.0007044241158208275,
+      "loss": 1.5866,
+      "step": 24820
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.0007041485567506091,
+      "loss": 1.6402,
+      "step": 24830
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.0007038729232491792,
+      "loss": 1.616,
+      "step": 24840
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.0007035972154170318,
+      "loss": 1.5711,
+      "step": 24850
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.0007033214333546882,
+      "loss": 1.4633,
+      "step": 24860
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.0007030455771626967,
+      "loss": 1.6538,
+      "step": 24870
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.0007027696469416328,
+      "loss": 1.5591,
+      "step": 24880
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.0007024936427920985,
+      "loss": 1.7478,
+      "step": 24890
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.0007022175648147233,
+      "loss": 1.6796,
+      "step": 24900
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0007019414131101634,
+      "loss": 1.5789,
+      "step": 24910
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0007016651877791018,
+      "loss": 1.6087,
+      "step": 24920
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0007013888889222486,
+      "loss": 1.4751,
+      "step": 24930
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.0007011125166403403,
+      "loss": 1.7013,
+      "step": 24940
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.0007008360710341406,
+      "loss": 1.6228,
+      "step": 24950
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.0007005595522044395,
+      "loss": 1.7012,
+      "step": 24960
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0007002829602520544,
+      "loss": 1.5933,
+      "step": 24970
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.0007000062952778283,
+      "loss": 1.7153,
+      "step": 24980
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.0006997295573826318,
+      "loss": 1.5683,
+      "step": 24990
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.0006994527466673614,
+      "loss": 1.6265,
+      "step": 25000
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.000699175863232941,
+      "loss": 1.5592,
+      "step": 25010
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00069889890718032,
+      "loss": 1.5341,
+      "step": 25020
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.0006986218786104747,
+      "loss": 1.6973,
+      "step": 25030
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.0006983447776244082,
+      "loss": 1.678,
+      "step": 25040
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.0006980676043231495,
+      "loss": 1.6734,
+      "step": 25050
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": {
+        "accuracy": 0.26718547341115434
+      },
+      "eval_f1": {
+        "f1": 0.23757487469054722
+      },
+      "eval_loss": 1.8294413089752197,
+      "eval_precision": {
+        "precision": 0.2619241634609933
+      },
+      "eval_recall": {
+        "recall": 0.28087243077801927
+      },
+      "eval_runtime": 175.7605,
+      "eval_samples_per_second": 21.933,
+      "eval_steps_per_second": 10.969,
+      "step": 25051
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.0006977903588077542,
+      "loss": 1.6655,
+      "step": 25060
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0006975130411793041,
+      "loss": 1.3927,
+      "step": 25070
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0006972356515389074,
+      "loss": 1.626,
+      "step": 25080
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0006969581899876986,
+      "loss": 1.4858,
+      "step": 25090
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.0006966806566268383,
+      "loss": 1.6478,
+      "step": 25100
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.0006964030515575133,
+      "loss": 1.6557,
+      "step": 25110
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.0006961253748809366,
+      "loss": 1.6834,
+      "step": 25120
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.0006958476266983473,
+      "loss": 1.5785,
+      "step": 25130
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.0006955698071110105,
+      "loss": 1.7666,
+      "step": 25140
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.0006952919162202173,
+      "loss": 1.764,
+      "step": 25150
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0006950139541272851,
+      "loss": 1.7854,
+      "step": 25160
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0006947359209335569,
+      "loss": 1.5456,
+      "step": 25170
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.0006944578167404017,
+      "loss": 1.6029,
+      "step": 25180
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.0006941796416492146,
+      "loss": 1.5788,
+      "step": 25190
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.0006939013957614163,
+      "loss": 1.4314,
+      "step": 25200
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.0006936230791784535,
+      "loss": 1.5209,
+      "step": 25210
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.0006933446920017982,
+      "loss": 1.5574,
+      "step": 25220
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.0006930662343329492,
+      "loss": 1.8195,
+      "step": 25230
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.0006927877062734298,
+      "loss": 1.6716,
+      "step": 25240
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.0006925091079247896,
+      "loss": 1.7492,
+      "step": 25250
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.0006922304393886038,
+      "loss": 1.7066,
+      "step": 25260
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.000691951700766473,
+      "loss": 1.5651,
+      "step": 25270
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.0006916728921600235,
+      "loss": 1.5665,
+      "step": 25280
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.0006913940136709072,
+      "loss": 1.6429,
+      "step": 25290
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.000691115065400801,
+      "loss": 1.5361,
+      "step": 25300
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.0006908360474514079,
+      "loss": 1.4306,
+      "step": 25310
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0006905569599244562,
+      "loss": 1.616,
+      "step": 25320
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.0006902778029216987,
+      "loss": 1.5992,
+      "step": 25330
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.0006899985765449148,
+      "loss": 1.4444,
+      "step": 25340
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.0006897192808959082,
+      "loss": 1.4674,
+      "step": 25350
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.0006894399160765085,
+      "loss": 1.5238,
+      "step": 25360
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.0006891604821885702,
+      "loss": 1.5557,
+      "step": 25370
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 0.0006888809793339729,
+      "loss": 1.6263,
+      "step": 25380
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.0006886014076146217,
+      "loss": 1.4557,
+      "step": 25390
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.0006883217671324463,
+      "loss": 1.4924,
+      "step": 25400
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0006880420579894023,
+      "loss": 1.6154,
+      "step": 25410
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.0006877622802874692,
+      "loss": 1.625,
+      "step": 25420
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.0006874824341286525,
+      "loss": 1.712,
+      "step": 25430
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.000687202519614982,
+      "loss": 1.7743,
+      "step": 25440
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.0006869225368485129,
+      "loss": 1.5561,
+      "step": 25450
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.0006866424859313249,
+      "loss": 1.6229,
+      "step": 25460
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0006863623669655229,
+      "loss": 1.5099,
+      "step": 25470
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.0006860821800532361,
+      "loss": 1.6965,
+      "step": 25480
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.0006858019252966192,
+      "loss": 1.6217,
+      "step": 25490
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.0006855216027978508,
+      "loss": 1.4581,
+      "step": 25500
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0006852412126591347,
+      "loss": 1.6703,
+      "step": 25510
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0006849607549826997,
+      "loss": 1.5106,
+      "step": 25520
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.0006846802298707985,
+      "loss": 1.4249,
+      "step": 25530
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.0006843996374257088,
+      "loss": 1.6436,
+      "step": 25540
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0006841189777497326,
+      "loss": 1.689,
+      "step": 25550
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.0006838382509451966,
+      "loss": 1.5731,
+      "step": 25560
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.0006835574571144519,
+      "loss": 1.4337,
+      "step": 25570
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.0006832765963598742,
+      "loss": 1.7067,
+      "step": 25580
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0006829956687838634,
+      "loss": 1.4985,
+      "step": 25590
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.000682714674488844,
+      "loss": 1.4956,
+      "step": 25600
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.0006824336135772643,
+      "loss": 1.5027,
+      "step": 25610
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.0006821524861515975,
+      "loss": 1.6026,
+      "step": 25620
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.0006818712923143408,
+      "loss": 1.5826,
+      "step": 25630
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.0006815900321680155,
+      "loss": 1.665,
+      "step": 25640
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.0006813087058151673,
+      "loss": 1.5066,
+      "step": 25650
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.0006810273133583658,
+      "loss": 1.4476,
+      "step": 25660
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.0006807458549002051,
+      "loss": 1.538,
+      "step": 25670
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.0006804643305433027,
+      "loss": 1.558,
+      "step": 25680
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.0006801827403903007,
+      "loss": 1.4717,
+      "step": 25690
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.000679901084543865,
+      "loss": 1.5761,
+      "step": 25700
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.0006796193631066855,
+      "loss": 1.5223,
+      "step": 25710
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0006793375761814759,
+      "loss": 1.6114,
+      "step": 25720
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0006790557238709739,
+      "loss": 1.6759,
+      "step": 25730
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.000678773806277941,
+      "loss": 1.5074,
+      "step": 25740
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.0006784918235051623,
+      "loss": 1.5035,
+      "step": 25750
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.0006782097756554472,
+      "loss": 1.4393,
+      "step": 25760
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.0006779276628316283,
+      "loss": 1.5455,
+      "step": 25770
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0006776454851365619,
+      "loss": 1.5785,
+      "step": 25780
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.0006773632426731283,
+      "loss": 1.5204,
+      "step": 25790
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0006770809355442313,
+      "loss": 1.5305,
+      "step": 25800
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.000676798563852798,
+      "loss": 1.6237,
+      "step": 25810
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.0006765161277017795,
+      "loss": 1.6322,
+      "step": 25820
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 0.0006762336271941498,
+      "loss": 1.5912,
+      "step": 25830
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.0006759510624329071,
+      "loss": 1.5876,
+      "step": 25840
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.0006756684335210724,
+      "loss": 1.835,
+      "step": 25850
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.0006753857405616903,
+      "loss": 1.4525,
+      "step": 25860
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.0006751029836578289,
+      "loss": 1.543,
+      "step": 25870
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.0006748201629125794,
+      "loss": 1.4412,
+      "step": 25880
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.0006745372784290566,
+      "loss": 1.4816,
+      "step": 25890
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.0006742543303103982,
+      "loss": 1.5681,
+      "step": 25900
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.000673971318659765,
+      "loss": 1.5032,
+      "step": 25910
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.0006736882435803413,
+      "loss": 1.5515,
+      "step": 25920
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.0006734051051753344,
+      "loss": 1.6889,
+      "step": 25930
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.0006731219035479749,
+      "loss": 1.4887,
+      "step": 25940
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0006728386388015161,
+      "loss": 1.4695,
+      "step": 25950
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0006725553110392342,
+      "loss": 1.4356,
+      "step": 25960
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.0006722719203644289,
+      "loss": 1.5508,
+      "step": 25970
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.0006719884668804223,
+      "loss": 1.5133,
+      "step": 25980
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0006717049506905601,
+      "loss": 1.6392,
+      "step": 25990
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0006714213718982101,
+      "loss": 1.6192,
+      "step": 26000
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0006711377306067633,
+      "loss": 1.4534,
+      "step": 26010
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.0006708540269196335,
+      "loss": 1.5005,
+      "step": 26020
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0006705702609402571,
+      "loss": 1.4409,
+      "step": 26030
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0006702864327720934,
+      "loss": 1.5415,
+      "step": 26040
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.0006700025425186243,
+      "loss": 1.344,
+      "step": 26050
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.000669718590283354,
+      "loss": 1.4412,
+      "step": 26060
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00066943457616981,
+      "loss": 1.6003,
+      "step": 26070
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.0006691505002815416,
+      "loss": 1.561,
+      "step": 26080
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.0006688663627221214,
+      "loss": 1.5576,
+      "step": 26090
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.0006685821635951436,
+      "loss": 1.5502,
+      "step": 26100
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0006682979030042259,
+      "loss": 1.4275,
+      "step": 26110
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.0006680135810530073,
+      "loss": 1.513,
+      "step": 26120
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.0006677291978451499,
+      "loss": 1.5188,
+      "step": 26130
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.000667444753484338,
+      "loss": 1.4293,
+      "step": 26140
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.000667160248074278,
+      "loss": 1.595,
+      "step": 26150
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.0006668756817186987,
+      "loss": 1.5701,
+      "step": 26160
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.0006665910545213514,
+      "loss": 1.5886,
+      "step": 26170
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.0006663063665860087,
+      "loss": 1.5982,
+      "step": 26180
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.0006660216180164665,
+      "loss": 1.5353,
+      "step": 26190
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.0006657368089165418,
+      "loss": 1.3894,
+      "step": 26200
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 0.0006654519393900747,
+      "loss": 1.464,
+      "step": 26210
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0006651670095409262,
+      "loss": 1.7545,
+      "step": 26220
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00066488201947298,
+      "loss": 1.5658,
+      "step": 26230
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.0006645969692901416,
+      "loss": 1.5761,
+      "step": 26240
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.0006643118590963385,
+      "loss": 1.6905,
+      "step": 26250
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.0006640266889955199,
+      "loss": 1.4606,
+      "step": 26260
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 0.0006637414590916568,
+      "loss": 1.6384,
+      "step": 26270
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.0006634561694887424,
+      "loss": 1.6132,
+      "step": 26280
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.0006631708202907912,
+      "loss": 1.5332,
+      "step": 26290
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0006628854116018395,
+      "loss": 1.5514,
+      "step": 26300
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.000662599943525946,
+      "loss": 1.629,
+      "step": 26310
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0006623144161671899,
+      "loss": 1.6445,
+      "step": 26320
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.0006620288296296727,
+      "loss": 1.6751,
+      "step": 26330
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 0.0006617431840175174,
+      "loss": 1.494,
+      "step": 26340
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 0.0006614574794348687,
+      "loss": 1.4694,
+      "step": 26350
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.0006611717159858923,
+      "loss": 1.4601,
+      "step": 26360
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.0006608858937747758,
+      "loss": 1.3469,
+      "step": 26370
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.000660600012905728,
+      "loss": 1.4919,
+      "step": 26380
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.0006603140734829795,
+      "loss": 1.5227,
+      "step": 26390
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 0.0006600280756107815,
+      "loss": 1.6549,
+      "step": 26400
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.0006597420193934071,
+      "loss": 1.5458,
+      "step": 26410
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.0006594559049351502,
+      "loss": 1.4529,
+      "step": 26420
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.000659169732340327,
+      "loss": 1.5235,
+      "step": 26430
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.0006588835017132734,
+      "loss": 1.3514,
+      "step": 26440
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.0006585972131583476,
+      "loss": 1.5487,
+      "step": 26450
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.0006583108667799283,
+      "loss": 1.5512,
+      "step": 26460
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.0006580244626824154,
+      "loss": 1.6131,
+      "step": 26470
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.0006577380009702302,
+      "loss": 1.5824,
+      "step": 26480
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.0006574514817478147,
+      "loss": 1.6387,
+      "step": 26490
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.0006571649051196318,
+      "loss": 1.4222,
+      "step": 26500
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.0006568782711901654,
+      "loss": 1.5664,
+      "step": 26510
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.0006565915800639206,
+      "loss": 1.435,
+      "step": 26520
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 0.0006563048318454228,
+      "loss": 1.5311,
+      "step": 26530
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 0.0006560180266392186,
+      "loss": 1.4908,
+      "step": 26540
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0006557311645498751,
+      "loss": 1.5186,
+      "step": 26550
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0006554442456819808,
+      "loss": 1.5297,
+      "step": 26560
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.000655157270140144,
+      "loss": 1.5198,
+      "step": 26570
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.0006548702380289942,
+      "loss": 1.4845,
+      "step": 26580
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.0006545831494531814,
+      "loss": 1.575,
+      "step": 26590
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.0006542960045173761,
+      "loss": 1.5071,
+      "step": 26600
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.0006540088033262698,
+      "loss": 1.3746,
+      "step": 26610
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.0006537215459845738,
+      "loss": 1.4694,
+      "step": 26620
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.0006534342325970201,
+      "loss": 1.385,
+      "step": 26630
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.0006531468632683617,
+      "loss": 1.3878,
+      "step": 26640
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0006528594381033714,
+      "loss": 1.4581,
+      "step": 26650
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 0.0006525719572068422,
+      "loss": 1.4837,
+      "step": 26660
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.000652284420683588,
+      "loss": 1.603,
+      "step": 26670
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.0006519968286384429,
+      "loss": 1.5532,
+      "step": 26680
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.0006517091811762607,
+      "loss": 1.2887,
+      "step": 26690
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.000651421478401916,
+      "loss": 1.5437,
+      "step": 26700
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.0006511337204203032,
+      "loss": 1.5092,
+      "step": 26710
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.0006508459073363369,
+      "loss": 1.5945,
+      "step": 26720
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.0006505580392549518,
+      "loss": 1.5614,
+      "step": 26730
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.000650270116281103,
+      "loss": 1.3385,
+      "step": 26740
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.0006499821385197652,
+      "loss": 1.5485,
+      "step": 26750
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.0006496941060759329,
+      "loss": 1.6625,
+      "step": 26760
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.0006494060190546211,
+      "loss": 1.5114,
+      "step": 26770
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 0.0006491178775608645,
+      "loss": 1.4848,
+      "step": 26780
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 0.0006488296816997174,
+      "loss": 1.5127,
+      "step": 26790
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.0006485414315762539,
+      "loss": 1.5657,
+      "step": 26800
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.0006482531272955685,
+      "loss": 1.6433,
+      "step": 26810
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.0006479647689627748,
+      "loss": 1.5471,
+      "step": 26820
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.0006476763566830064,
+      "loss": 1.3888,
+      "step": 26830
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 0.0006473878905614165,
+      "loss": 1.5099,
+      "step": 26840
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 0.0006470993707031782,
+      "loss": 1.5364,
+      "step": 26850
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.0006468107972134832,
+      "loss": 1.5134,
+      "step": 26860
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.0006465221701975443,
+      "loss": 1.5307,
+      "step": 26870
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.0006462334897605925,
+      "loss": 1.417,
+      "step": 26880
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.000645944756007879,
+      "loss": 1.5311,
+      "step": 26890
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 0.0006456559690446739,
+      "loss": 1.4572,
+      "step": 26900
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 0.0006453671289762675,
+      "loss": 1.5692,
+      "step": 26910
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.0006450782359079685,
+      "loss": 1.5839,
+      "step": 26920
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.0006447892899451059,
+      "loss": 1.4544,
+      "step": 26930
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.0006445002911930272,
+      "loss": 1.2894,
+      "step": 26940
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.0006442112397570992,
+      "loss": 1.4388,
+      "step": 26950
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.0006439221357427085,
+      "loss": 1.534,
+      "step": 26960
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.0006436329792552604,
+      "loss": 1.4376,
+      "step": 26970
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": {
+        "accuracy": 0.4254215304798962
+      },
+      "eval_f1": {
+        "f1": 0.41718777684401953
+      },
+      "eval_loss": 1.3829901218414307,
+      "eval_precision": {
+        "precision": 0.4554477836203232
+      },
+      "eval_recall": {
+        "recall": 0.450319904911865
+      },
+      "eval_runtime": 192.2983,
+      "eval_samples_per_second": 20.047,
+      "eval_steps_per_second": 10.026,
+      "step": 26978
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.0006433437704001794,
+      "loss": 1.4387,
+      "step": 26980
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.0006430545092829091,
+      "loss": 1.3655,
+      "step": 26990
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.0006427651960089123,
+      "loss": 1.5138,
+      "step": 27000
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.0006424758306836707,
+      "loss": 1.2805,
+      "step": 27010
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.0006421864134126848,
+      "loss": 1.6224,
+      "step": 27020
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.0006418969443014745,
+      "loss": 1.5197,
+      "step": 27030
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.000641607423455578,
+      "loss": 1.5643,
+      "step": 27040
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.000641317850980553,
+      "loss": 1.5071,
+      "step": 27050
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.0006410282269819756,
+      "loss": 1.5523,
+      "step": 27060
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.0006407385515654407,
+      "loss": 1.5096,
+      "step": 27070
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.000640448824836562,
+      "loss": 1.6588,
+      "step": 27080
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.0006401590469009721,
+      "loss": 1.6652,
+      "step": 27090
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.0006398692178643221,
+      "loss": 1.4023,
+      "step": 27100
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.0006395793378322817,
+      "loss": 1.4075,
+      "step": 27110
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.0006392894069105391,
+      "loss": 1.7573,
+      "step": 27120
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.0006389994252048015,
+      "loss": 1.4026,
+      "step": 27130
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.000638709392820794,
+      "loss": 1.4055,
+      "step": 27140
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.0006384193098642608,
+      "loss": 1.5018,
+      "step": 27150
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.0006381291764409639,
+      "loss": 1.534,
+      "step": 27160
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.0006378389926566841,
+      "loss": 1.5202,
+      "step": 27170
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.0006375487586172208,
+      "loss": 1.4313,
+      "step": 27180
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.0006372584744283911,
+      "loss": 1.6616,
+      "step": 27190
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.0006369681401960308,
+      "loss": 1.357,
+      "step": 27200
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.0006366777560259936,
+      "loss": 1.801,
+      "step": 27210
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 0.0006363873220241521,
+      "loss": 1.5208,
+      "step": 27220
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 0.0006360968382963961,
+      "loss": 1.6261,
+      "step": 27230
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0006358063049486345,
+      "loss": 1.4024,
+      "step": 27240
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0006355157220867936,
+      "loss": 1.2805,
+      "step": 27250
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.000635225089816818,
+      "loss": 1.5817,
+      "step": 27260
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.0006349344082446702,
+      "loss": 1.4958,
+      "step": 27270
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.0006346436774763312,
+      "loss": 1.5176,
+      "step": 27280
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.000634352897617799,
+      "loss": 1.4122,
+      "step": 27290
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0006340620687750902,
+      "loss": 1.3902,
+      "step": 27300
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0006337711910542391,
+      "loss": 1.4241,
+      "step": 27310
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.0006334802645612978,
+      "loss": 1.5132,
+      "step": 27320
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.0006331892894023364,
+      "loss": 1.4824,
+      "step": 27330
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.000632898265683442,
+      "loss": 1.5123,
+      "step": 27340
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.0006326071935107202,
+      "loss": 1.3313,
+      "step": 27350
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.0006323160729902941,
+      "loss": 1.5495,
+      "step": 27360
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.0006320249042283041,
+      "loss": 1.3872,
+      "step": 27370
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0006317336873309087,
+      "loss": 1.4376,
+      "step": 27380
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0006314424224042832,
+      "loss": 1.4452,
+      "step": 27390
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.0006311511095546212,
+      "loss": 1.4109,
+      "step": 27400
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.0006308597488881334,
+      "loss": 1.2785,
+      "step": 27410
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 0.0006305683405110478,
+      "loss": 1.784,
+      "step": 27420
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 0.0006302768845296101,
+      "loss": 1.6483,
+      "step": 27430
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.0006299853810500832,
+      "loss": 1.5428,
+      "step": 27440
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.0006296938301787472,
+      "loss": 1.4571,
+      "step": 27450
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.0006294022320219002,
+      "loss": 1.7306,
+      "step": 27460
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 0.000629110586685856,
+      "loss": 1.6281,
+      "step": 27470
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 0.0006288188942769474,
+      "loss": 1.5098,
+      "step": 27480
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.000628527154901523,
+      "loss": 1.2499,
+      "step": 27490
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.0006282353686659493,
+      "loss": 1.3956,
+      "step": 27500
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.0006279435356766097,
+      "loss": 1.505,
+      "step": 27510
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.0006276516560399042,
+      "loss": 1.7419,
+      "step": 27520
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.0006273597298622505,
+      "loss": 1.5575,
+      "step": 27530
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.0006270677572500828,
+      "loss": 1.6,
+      "step": 27540
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.0006267757383098527,
+      "loss": 1.5227,
+      "step": 27550
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.0006264836731480279,
+      "loss": 1.6869,
+      "step": 27560
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.0006261915618710936,
+      "loss": 1.5666,
+      "step": 27570
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.0006258994045855518,
+      "loss": 1.5934,
+      "step": 27580
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.000625607201397921,
+      "loss": 1.7175,
+      "step": 27590
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.0006253149524147365,
+      "loss": 1.848,
+      "step": 27600
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 0.0006250226577425505,
+      "loss": 1.4144,
+      "step": 27610
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 0.0006247303174879313,
+      "loss": 1.458,
+      "step": 27620
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.0006244379317574646,
+      "loss": 1.6376,
+      "step": 27630
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.0006241455006577524,
+      "loss": 1.5142,
+      "step": 27640
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.0006238530242954127,
+      "loss": 1.3808,
+      "step": 27650
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.0006235605027770809,
+      "loss": 1.4472,
+      "step": 27660
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.000623267936209408,
+      "loss": 1.8435,
+      "step": 27670
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.0006229753246990623,
+      "loss": 1.6603,
+      "step": 27680
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.0006226826683527276,
+      "loss": 1.4969,
+      "step": 27690
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.0006223899672771047,
+      "loss": 1.4423,
+      "step": 27700
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.0006220972215789101,
+      "loss": 1.6183,
+      "step": 27710
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.0006218044313648775,
+      "loss": 1.4625,
+      "step": 27720
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 0.000621511596741756,
+      "loss": 1.4327,
+      "step": 27730
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.000621218717816311,
+      "loss": 1.5237,
+      "step": 27740
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.0006209257946953246,
+      "loss": 1.5643,
+      "step": 27750
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.0006206328274855941,
+      "loss": 1.5683,
+      "step": 27760
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.0006203398162939341,
+      "loss": 1.4535,
+      "step": 27770
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.0006200467612271738,
+      "loss": 1.3992,
+      "step": 27780
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 0.0006197536623921598,
+      "loss": 1.4676,
+      "step": 27790
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.0006194605198957535,
+      "loss": 1.5684,
+      "step": 27800
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.000619167333844833,
+      "loss": 1.5997,
+      "step": 27810
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.0006188741043462919,
+      "loss": 1.628,
+      "step": 27820
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.0006185808315070396,
+      "loss": 1.6247,
+      "step": 27830
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.0006182875154340016,
+      "loss": 1.5918,
+      "step": 27840
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.000617994156234119,
+      "loss": 1.5176,
+      "step": 27850
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.0006177007540143487,
+      "loss": 1.7319,
+      "step": 27860
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.0006174073088816628,
+      "loss": 1.7409,
+      "step": 27870
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0006171138209430501,
+      "loss": 1.6273,
+      "step": 27880
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0006168202903055136,
+      "loss": 1.7265,
+      "step": 27890
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.0006165267170760733,
+      "loss": 1.4754,
+      "step": 27900
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.0006162331013617637,
+      "loss": 1.4705,
+      "step": 27910
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.0006159394432696352,
+      "loss": 1.6364,
+      "step": 27920
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.0006156457429067535,
+      "loss": 1.5088,
+      "step": 27930
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.0006153520003802001,
+      "loss": 1.6722,
+      "step": 27940
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.0006150582157970714,
+      "loss": 1.5023,
+      "step": 27950
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.0006147643892644793,
+      "loss": 1.6198,
+      "step": 27960
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.0006144705208895509,
+      "loss": 1.3587,
+      "step": 27970
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 0.000614176610779429,
+      "loss": 1.3613,
+      "step": 27980
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.0006138826590412709,
+      "loss": 1.5255,
+      "step": 27990
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.0006135886657822496,
+      "loss": 1.3913,
+      "step": 28000
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.0006132946311095533,
+      "loss": 1.4997,
+      "step": 28010
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.0006130005551303846,
+      "loss": 1.4289,
+      "step": 28020
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.0006127064379519622,
+      "loss": 1.3743,
+      "step": 28030
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.0006124122796815192,
+      "loss": 1.5381,
+      "step": 28040
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 0.0006121180804263033,
+      "loss": 1.4481,
+      "step": 28050
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 0.0006118238402935779,
+      "loss": 1.56,
+      "step": 28060
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.0006115295593906213,
+      "loss": 1.4882,
+      "step": 28070
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.0006112352378247259,
+      "loss": 1.5383,
+      "step": 28080
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.0006109408757031994,
+      "loss": 1.493,
+      "step": 28090
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.0006106464731333646,
+      "loss": 1.9129,
+      "step": 28100
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.0006103520302225585,
+      "loss": 1.4833,
+      "step": 28110
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.0006100575470781332,
+      "loss": 1.4138,
+      "step": 28120
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.0006097630238074552,
+      "loss": 1.5182,
+      "step": 28130
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.0006094684605179058,
+      "loss": 1.5054,
+      "step": 28140
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.0006091738573168805,
+      "loss": 1.5881,
+      "step": 28150
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.0006088792143117902,
+      "loss": 1.5685,
+      "step": 28160
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.0006085845316100595,
+      "loss": 1.514,
+      "step": 28170
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.0006082898093191277,
+      "loss": 1.4664,
+      "step": 28180
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.0006079950475464486,
+      "loss": 1.413,
+      "step": 28190
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.0006077002463994907,
+      "loss": 1.5035,
+      "step": 28200
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.0006074054059857361,
+      "loss": 1.3579,
+      "step": 28210
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.0006071105264126821,
+      "loss": 1.3843,
+      "step": 28220
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.0006068156077878396,
+      "loss": 1.3425,
+      "step": 28230
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.0006065206502187338,
+      "loss": 1.5484,
+      "step": 28240
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.0006062256538129047,
+      "loss": 1.4962,
+      "step": 28250
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.0006059306186779057,
+      "loss": 1.4954,
+      "step": 28260
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.0006056355449213049,
+      "loss": 1.3406,
+      "step": 28270
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.0006053404326506839,
+      "loss": 1.4301,
+      "step": 28280
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.0006050452819736389,
+      "loss": 1.6767,
+      "step": 28290
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 0.00060475009299778,
+      "loss": 1.5351,
+      "step": 28300
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 0.000604454865830731,
+      "loss": 1.6479,
+      "step": 28310
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.0006041596005801297,
+      "loss": 1.4983,
+      "step": 28320
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.0006038642973536281,
+      "loss": 1.4981,
+      "step": 28330
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.0006035689562588914,
+      "loss": 1.7159,
+      "step": 28340
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.0006032735774035994,
+      "loss": 1.6014,
+      "step": 28350
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.0006029781608954451,
+      "loss": 1.4095,
+      "step": 28360
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 0.0006026827068421353,
+      "loss": 1.4774,
+      "step": 28370
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0006023872153513908,
+      "loss": 1.6294,
+      "step": 28380
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0006020916865309458,
+      "loss": 1.4025,
+      "step": 28390
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.0006017961204885481,
+      "loss": 1.3732,
+      "step": 28400
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.0006015005173319591,
+      "loss": 1.4687,
+      "step": 28410
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 0.0006012048771689538,
+      "loss": 1.3476,
+      "step": 28420
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 0.0006009092001073205,
+      "loss": 1.4449,
+      "step": 28430
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.0006006134862548613,
+      "loss": 1.323,
+      "step": 28440
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.0006003177357193915,
+      "loss": 1.3105,
+      "step": 28450
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.0006000219486087396,
+      "loss": 1.3982,
+      "step": 28460
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.0005997261250307479,
+      "loss": 1.4592,
+      "step": 28470
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.0005994302650932714,
+      "loss": 1.5194,
+      "step": 28480
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.000599134368904179,
+      "loss": 1.6046,
+      "step": 28490
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 0.0005988384365713521,
+      "loss": 1.5474,
+      "step": 28500
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.000598542468202686,
+      "loss": 1.4111,
+      "step": 28510
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.0005982464639060887,
+      "loss": 1.5506,
+      "step": 28520
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.0005979504237894814,
+      "loss": 1.2668,
+      "step": 28530
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.0005976543479607983,
+      "loss": 1.4706,
+      "step": 28540
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 0.0005973582365279868,
+      "loss": 1.5549,
+      "step": 28550
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 0.0005970620895990071,
+      "loss": 1.5022,
+      "step": 28560
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.0005967659072818324,
+      "loss": 1.3457,
+      "step": 28570
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.0005964696896844491,
+      "loss": 1.4328,
+      "step": 28580
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.0005961734369148556,
+      "loss": 1.5529,
+      "step": 28590
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.0005958771490810642,
+      "loss": 1.3646,
+      "step": 28600
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.0005955808262910991,
+      "loss": 1.6154,
+      "step": 28610
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.000595284468652998,
+      "loss": 1.3585,
+      "step": 28620
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.0005949880762748106,
+      "loss": 1.4099,
+      "step": 28630
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.0005946916492645999,
+      "loss": 1.4053,
+      "step": 28640
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.000594395187730441,
+      "loss": 1.4784,
+      "step": 28650
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.0005940986917804219,
+      "loss": 1.3018,
+      "step": 28660
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.0005938021615226431,
+      "loss": 1.5371,
+      "step": 28670
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.0005935055970652175,
+      "loss": 1.5364,
+      "step": 28680
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0005932089985162704,
+      "loss": 1.5031,
+      "step": 28690
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0005929123659839399,
+      "loss": 1.479,
+      "step": 28700
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.0005926156995763763,
+      "loss": 1.5163,
+      "step": 28710
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.0005923189994017419,
+      "loss": 1.4877,
+      "step": 28720
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.0005920222655682119,
+      "loss": 1.4887,
+      "step": 28730
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.0005917254981839733,
+      "loss": 1.3458,
+      "step": 28740
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.0005914286973572259,
+      "loss": 1.3289,
+      "step": 28750
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.0005911318631961807,
+      "loss": 1.5668,
+      "step": 28760
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.000590834995809062,
+      "loss": 1.2921,
+      "step": 28770
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.0005905380953041053,
+      "loss": 1.5287,
+      "step": 28780
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.0005902411617895589,
+      "loss": 1.3411,
+      "step": 28790
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.0005899441953736825,
+      "loss": 1.363,
+      "step": 28800
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.0005896471961647483,
+      "loss": 1.3387,
+      "step": 28810
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.00058935016427104,
+      "loss": 1.5612,
+      "step": 28820
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.0005890530998008538,
+      "loss": 1.6634,
+      "step": 28830
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.0005887560028624973,
+      "loss": 1.409,
+      "step": 28840
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.00058845887356429,
+      "loss": 1.4984,
+      "step": 28850
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 0.0005881617120145632,
+      "loss": 1.4476,
+      "step": 28860
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 0.0005878645183216602,
+      "loss": 1.3926,
+      "step": 28870
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.0005875672925939358,
+      "loss": 1.4656,
+      "step": 28880
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.0005872700349397566,
+      "loss": 1.3928,
+      "step": 28890
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.0005869727454675007,
+      "loss": 1.3403,
+      "step": 28900
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": {
+        "accuracy": 0.4926070038910506
+      },
+      "eval_f1": {
+        "f1": 0.44282153705015465
+      },
+      "eval_loss": 1.2193456888198853,
+      "eval_precision": {
+        "precision": 0.5137836037175402
+      },
+      "eval_recall": {
+        "recall": 0.5162930270137864
+      },
+      "eval_runtime": 179.9656,
+      "eval_samples_per_second": 21.421,
+      "eval_steps_per_second": 10.713,
+      "step": 28905
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.0005866754242855577,
+      "loss": 1.325,
+      "step": 28910
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.0005863780715023293,
+      "loss": 1.5068,
+      "step": 28920
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.0005860806872262281,
+      "loss": 1.5348,
+      "step": 28930
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 0.0005857832715656783,
+      "loss": 1.3228,
+      "step": 28940
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 0.0005854858246291161,
+      "loss": 1.4757,
+      "step": 28950
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.0005851883465249884,
+      "loss": 1.4747,
+      "step": 28960
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.000584890837361754,
+      "loss": 1.5082,
+      "step": 28970
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.0005845932972478822,
+      "loss": 1.2903,
+      "step": 28980
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.0005842957262918549,
+      "loss": 1.5229,
+      "step": 28990
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.0005839981246021638,
+      "loss": 1.4184,
+      "step": 29000
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.0005837004922873131,
+      "loss": 1.3969,
+      "step": 29010
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.0005834028294558172,
+      "loss": 1.3151,
+      "step": 29020
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.0005831051362162022,
+      "loss": 1.3888,
+      "step": 29030
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.0005828074126770047,
+      "loss": 1.3395,
+      "step": 29040
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.0005825096589467733,
+      "loss": 1.5509,
+      "step": 29050
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.0005822118751340668,
+      "loss": 1.4212,
+      "step": 29060
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.000581914061347455,
+      "loss": 1.6824,
+      "step": 29070
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.000581616217695519,
+      "loss": 1.4122,
+      "step": 29080
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.0005813183442868506,
+      "loss": 1.3746,
+      "step": 29090
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.0005810204412300526,
+      "loss": 1.5857,
+      "step": 29100
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.0005807225086337383,
+      "loss": 1.5656,
+      "step": 29110
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.0005804245466065321,
+      "loss": 1.2708,
+      "step": 29120
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 0.0005801265552570688,
+      "loss": 1.4179,
+      "step": 29130
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 0.0005798285346939942,
+      "loss": 1.4214,
+      "step": 29140
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.0005795304850259647,
+      "loss": 1.4064,
+      "step": 29150
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.000579232406361647,
+      "loss": 1.3242,
+      "step": 29160
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.0005789342988097189,
+      "loss": 1.3365,
+      "step": 29170
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.000578636162478868,
+      "loss": 1.3818,
+      "step": 29180
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.0005783379974777933,
+      "loss": 1.5153,
+      "step": 29190
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.0005780398039152037,
+      "loss": 1.4606,
+      "step": 29200
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.0005777415818998184,
+      "loss": 1.4678,
+      "step": 29210
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.0005774433315403672,
+      "loss": 1.3823,
+      "step": 29220
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.0005771450529455902,
+      "loss": 1.3732,
+      "step": 29230
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.0005768467462242381,
+      "loss": 1.4186,
+      "step": 29240
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.0005765484114850714,
+      "loss": 1.5088,
+      "step": 29250
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.0005762500488368605,
+      "loss": 1.4081,
+      "step": 29260
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.0005759516583883871,
+      "loss": 1.4918,
+      "step": 29270
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.000575653240248442,
+      "loss": 1.339,
+      "step": 29280
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0005753547945258268,
+      "loss": 1.2755,
+      "step": 29290
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0005750563213293525,
+      "loss": 1.3679,
+      "step": 29300
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 0.0005747578207678406,
+      "loss": 1.4541,
+      "step": 29310
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.0005744592929501223,
+      "loss": 1.4933,
+      "step": 29320
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.0005741607379850392,
+      "loss": 1.628,
+      "step": 29330
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.0005738621559814421,
+      "loss": 1.4701,
+      "step": 29340
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.0005735635470481921,
+      "loss": 1.4364,
+      "step": 29350
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.0005732649112941602,
+      "loss": 1.1453,
+      "step": 29360
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.0005729662488282269,
+      "loss": 1.5368,
+      "step": 29370
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 0.0005726675597592827,
+      "loss": 1.3539,
+      "step": 29380
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 0.0005723688441962273,
+      "loss": 1.3602,
+      "step": 29390
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.0005720701022479709,
+      "loss": 1.5268,
+      "step": 29400
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.0005717713340234323,
+      "loss": 1.2503,
+      "step": 29410
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.0005714725396315408,
+      "loss": 1.5365,
+      "step": 29420
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.0005711737191812349,
+      "loss": 1.3535,
+      "step": 29430
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.0005708748727814621,
+      "loss": 1.3793,
+      "step": 29440
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 0.0005705760005411804,
+      "loss": 1.2828,
+      "step": 29450
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.0005702771025693563,
+      "loss": 1.3234,
+      "step": 29460
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.0005699781789749662,
+      "loss": 1.3435,
+      "step": 29470
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0005696792298669952,
+      "loss": 1.2255,
+      "step": 29480
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0005693802553544388,
+      "loss": 1.3819,
+      "step": 29490
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 0.0005690812555463007,
+      "loss": 1.3077,
+      "step": 29500
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 0.0005687822305515946,
+      "loss": 1.3535,
+      "step": 29510
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.0005684831804793427,
+      "loss": 1.3831,
+      "step": 29520
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.0005681841054385768,
+      "loss": 1.3282,
+      "step": 29530
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.0005678850055383377,
+      "loss": 1.3916,
+      "step": 29540
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.0005675858808876752,
+      "loss": 1.4343,
+      "step": 29550
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 0.0005672867315956483,
+      "loss": 1.2737,
+      "step": 29560
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.0005669875577713245,
+      "loss": 1.1907,
+      "step": 29570
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.000566688359523781,
+      "loss": 1.2218,
+      "step": 29580
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.0005663891369621032,
+      "loss": 1.538,
+      "step": 29590
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.000566089890195386,
+      "loss": 1.3224,
+      "step": 29600
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.0005657906193327325,
+      "loss": 1.5279,
+      "step": 29610
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.0005654913244832549,
+      "loss": 1.269,
+      "step": 29620
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0005651920057560741,
+      "loss": 1.5014,
+      "step": 29630
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0005648926632603199,
+      "loss": 1.5474,
+      "step": 29640
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.0005645932971051306,
+      "loss": 1.2871,
+      "step": 29650
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.0005642939073996529,
+      "loss": 1.5208,
+      "step": 29660
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.0005639944942530425,
+      "loss": 1.5512,
+      "step": 29670
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.0005636950577744632,
+      "loss": 1.0971,
+      "step": 29680
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.0005633955980730881,
+      "loss": 1.5472,
+      "step": 29690
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.0005630961152580975,
+      "loss": 1.5223,
+      "step": 29700
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.0005627966094386811,
+      "loss": 1.4048,
+      "step": 29710
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.0005624970807240371,
+      "loss": 1.4304,
+      "step": 29720
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.0005621975292233715,
+      "loss": 1.2976,
+      "step": 29730
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.0005618979550458985,
+      "loss": 1.268,
+      "step": 29740
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.0005615983583008411,
+      "loss": 1.3611,
+      "step": 29750
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.0005612987390974302,
+      "loss": 1.403,
+      "step": 29760
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.0005609990975449052,
+      "loss": 1.3407,
+      "step": 29770
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.0005606994337525132,
+      "loss": 1.6409,
+      "step": 29780
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.0005603997478295096,
+      "loss": 1.6322,
+      "step": 29790
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.0005601000398851582,
+      "loss": 1.5174,
+      "step": 29800
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.0005598003100287302,
+      "loss": 1.5226,
+      "step": 29810
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.0005595005583695055,
+      "loss": 1.2034,
+      "step": 29820
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 0.0005592007850167709,
+      "loss": 1.6588,
+      "step": 29830
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.0005589009900798226,
+      "loss": 1.5569,
+      "step": 29840
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.0005586011736679632,
+      "loss": 1.409,
+      "step": 29850
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.000558301335890504,
+      "loss": 1.4667,
+      "step": 29860
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.0005580014768567641,
+      "loss": 1.335,
+      "step": 29870
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.0005577015966760696,
+      "loss": 1.4007,
+      "step": 29880
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.0005574016954577555,
+      "loss": 1.3312,
+      "step": 29890
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.0005571017733111633,
+      "loss": 1.7058,
+      "step": 29900
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.0005568018303456429,
+      "loss": 1.4601,
+      "step": 29910
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.0005565018666705511,
+      "loss": 1.2018,
+      "step": 29920
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.0005562018823952532,
+      "loss": 1.594,
+      "step": 29930
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.000555901877629121,
+      "loss": 1.2801,
+      "step": 29940
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.0005556018524815349,
+      "loss": 1.5199,
+      "step": 29950
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.0005553018070618813,
+      "loss": 1.4142,
+      "step": 29960
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.0005550017414795552,
+      "loss": 1.397,
+      "step": 29970
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.0005547016558439585,
+      "loss": 1.4342,
+      "step": 29980
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.0005544015502645005,
+      "loss": 1.3275,
+      "step": 29990
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.0005541014248505974,
+      "loss": 1.3458,
+      "step": 30000
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.0005538012797116732,
+      "loss": 1.5593,
+      "step": 30010
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 0.0005535011149571589,
+      "loss": 1.4796,
+      "step": 30020
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 0.0005532009306964922,
+      "loss": 1.2613,
+      "step": 30030
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.000552900727039119,
+      "loss": 1.5965,
+      "step": 30040
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0005526005040944907,
+      "loss": 1.4189,
+      "step": 30050
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.0005523002619720671,
+      "loss": 1.4849,
+      "step": 30060
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.0005520000007813145,
+      "loss": 1.2994,
+      "step": 30070
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.000551699720631706,
+      "loss": 1.4458,
+      "step": 30080
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 0.0005513994216327218,
+      "loss": 1.3616,
+      "step": 30090
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.000551099103893849,
+      "loss": 1.4759,
+      "step": 30100
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.0005507987675245812,
+      "loss": 1.4463,
+      "step": 30110
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.0005504984126344193,
+      "loss": 1.466,
+      "step": 30120
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 0.0005501980393328707,
+      "loss": 1.4833,
+      "step": 30130
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 0.0005498976477294494,
+      "loss": 1.4955,
+      "step": 30140
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.0005495972379336762,
+      "loss": 1.3792,
+      "step": 30150
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.0005492968100550786,
+      "loss": 1.4593,
+      "step": 30160
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.0005489963642031906,
+      "loss": 1.3182,
+      "step": 30170
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.0005486959004875527,
+      "loss": 1.556,
+      "step": 30180
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.000548395419017712,
+      "loss": 1.2955,
+      "step": 30190
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.0005480949199032221,
+      "loss": 1.3399,
+      "step": 30200
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.000547794403253643,
+      "loss": 1.4116,
+      "step": 30210
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.000547493869178541,
+      "loss": 1.486,
+      "step": 30220
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.0005471933177874889,
+      "loss": 1.3274,
+      "step": 30230
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.0005468927491900657,
+      "loss": 1.2269,
+      "step": 30240
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.0005465921634958568,
+      "loss": 1.2096,
+      "step": 30250
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.0005462915608144537,
+      "loss": 1.5215,
+      "step": 30260
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.000545990941255454,
+      "loss": 1.2572,
+      "step": 30270
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.000545690304928462,
+      "loss": 1.3519,
+      "step": 30280
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.000545389651943087,
+      "loss": 1.4926,
+      "step": 30290
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.000545088982408946,
+      "loss": 1.3986,
+      "step": 30300
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.0005447882964356606,
+      "loss": 1.3787,
+      "step": 30310
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.0005444875941328589,
+      "loss": 1.4762,
+      "step": 30320
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.0005441868756101749,
+      "loss": 1.3161,
+      "step": 30330
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.0005438861409772489,
+      "loss": 1.3621,
+      "step": 30340
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 0.0005435853903437265,
+      "loss": 1.4219,
+      "step": 30350
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.0005432846238192595,
+      "loss": 1.4518,
+      "step": 30360
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.0005429838415135054,
+      "loss": 1.6096,
+      "step": 30370
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 0.0005426830435361273,
+      "loss": 1.4774,
+      "step": 30380
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 0.0005423822299967943,
+      "loss": 1.2652,
+      "step": 30390
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.000542081401005181,
+      "loss": 1.5349,
+      "step": 30400
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.0005417805566709676,
+      "loss": 1.5714,
+      "step": 30410
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.0005414796971038399,
+      "loss": 1.6111,
+      "step": 30420
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.0005411788224134894,
+      "loss": 1.4296,
+      "step": 30430
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.0005408779327096131,
+      "loss": 1.4444,
+      "step": 30440
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.0005405770281019134,
+      "loss": 1.4881,
+      "step": 30450
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.0005402761087000977,
+      "loss": 1.5718,
+      "step": 30460
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.0005399751746138797,
+      "loss": 1.3801,
+      "step": 30470
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0005396742259529777,
+      "loss": 1.4718,
+      "step": 30480
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0005393732628271159,
+      "loss": 1.4403,
+      "step": 30490
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.0005390722853460231,
+      "loss": 1.3083,
+      "step": 30500
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.0005387712936194338,
+      "loss": 1.2577,
+      "step": 30510
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.0005384702877570879,
+      "loss": 1.54,
+      "step": 30520
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.0005381692678687298,
+      "loss": 1.282,
+      "step": 30530
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.0005378682340641095,
+      "loss": 1.3894,
+      "step": 30540
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.0005375671864529816,
+      "loss": 1.3916,
+      "step": 30550
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.0005372661251451066,
+      "loss": 1.3402,
+      "step": 30560
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.0005369650502502491,
+      "loss": 1.714,
+      "step": 30570
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 0.0005366639618781793,
+      "loss": 1.4657,
+      "step": 30580
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 0.0005363628601386717,
+      "loss": 1.3566,
+      "step": 30590
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.0005360617451415061,
+      "loss": 1.4477,
+      "step": 30600
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.0005357606169964672,
+      "loss": 1.3519,
+      "step": 30610
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.0005354594758133442,
+      "loss": 1.698,
+      "step": 30620
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 0.0005351583217019311,
+      "loss": 1.7847,
+      "step": 30630
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 0.0005348571547720268,
+      "loss": 1.7121,
+      "step": 30640
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.0005345559751334348,
+      "loss": 1.583,
+      "step": 30650
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.0005342547828959631,
+      "loss": 1.5192,
+      "step": 30660
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.0005339535781694245,
+      "loss": 1.4977,
+      "step": 30670
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.0005336523610636361,
+      "loss": 1.5195,
+      "step": 30680
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.0005333511316884199,
+      "loss": 1.419,
+      "step": 30690
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.0005330498901536018,
+      "loss": 1.3941,
+      "step": 30700
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.0005327486365690129,
+      "loss": 1.4306,
+      "step": 30710
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 0.0005324473710444878,
+      "loss": 1.2836,
+      "step": 30720
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.0005321460936898661,
+      "loss": 1.4399,
+      "step": 30730
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.0005318448046149918,
+      "loss": 1.5018,
+      "step": 30740
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.0005315435039297124,
+      "loss": 1.3423,
+      "step": 30750
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.0005312421917438807,
+      "loss": 1.3439,
+      "step": 30760
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.0005309408681673525,
+      "loss": 1.2519,
+      "step": 30770
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.000530639533309989,
+      "loss": 1.3356,
+      "step": 30780
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.0005303381872816544,
+      "loss": 1.5291,
+      "step": 30790
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.0005300368301922177,
+      "loss": 1.3812,
+      "step": 30800
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.0005297354621515518,
+      "loss": 1.3437,
+      "step": 30810
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.0005294340832695333,
+      "loss": 1.3829,
+      "step": 30820
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.0005291326936560428,
+      "loss": 1.4806,
+      "step": 30830
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": {
+        "accuracy": 0.5060959792477302
+      },
+      "eval_f1": {
+        "f1": 0.49960050025350783
+      },
+      "eval_loss": 1.2118881940841675,
+      "eval_precision": {
+        "precision": 0.4999152354253829
+      },
+      "eval_recall": {
+        "recall": 0.5224114833278971
+      },
+      "eval_runtime": 180.7814,
+      "eval_samples_per_second": 21.324,
+      "eval_steps_per_second": 10.665,
+      "step": 30832
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.0005288312934209654,
+      "loss": 1.4065,
+      "step": 30840
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.0005285298826741893,
+      "loss": 1.3169,
+      "step": 30850
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.0005282284615256068,
+      "loss": 1.1324,
+      "step": 30860
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.0005279270300851145,
+      "loss": 1.3435,
+      "step": 30870
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.0005276255884626116,
+      "loss": 1.2843,
+      "step": 30880
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.0005273241367680027,
+      "loss": 1.3872,
+      "step": 30890
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.000527022675111194,
+      "loss": 1.3855,
+      "step": 30900
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.0005267212036020972,
+      "loss": 1.3654,
+      "step": 30910
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.0005264197223506264,
+      "loss": 1.2504,
+      "step": 30920
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.0005261182314666999,
+      "loss": 1.2673,
+      "step": 30930
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.0005258167310602393,
+      "loss": 1.5953,
+      "step": 30940
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.0005255152212411693,
+      "loss": 1.1848,
+      "step": 30950
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 0.0005252137021194189,
+      "loss": 1.5049,
+      "step": 30960
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 0.0005249121738049196,
+      "loss": 1.3496,
+      "step": 30970
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.0005246106364076069,
+      "loss": 1.4917,
+      "step": 30980
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.0005243090900374189,
+      "loss": 1.4941,
+      "step": 30990
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.0005240075348042979,
+      "loss": 1.2742,
+      "step": 31000
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.0005237059708181885,
+      "loss": 1.4572,
+      "step": 31010
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 0.0005234043981890394,
+      "loss": 1.4536,
+      "step": 31020
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 0.0005231028170268016,
+      "loss": 1.4797,
+      "step": 31030
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.0005228012274414298,
+      "loss": 1.3718,
+      "step": 31040
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.0005224996295428814,
+      "loss": 1.1734,
+      "step": 31050
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.0005221980234411172,
+      "loss": 1.2494,
+      "step": 31060
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.0005218964092461008,
+      "loss": 1.3399,
+      "step": 31070
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 0.0005215947870677983,
+      "loss": 1.343,
+      "step": 31080
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 0.0005212931570161796,
+      "loss": 1.4225,
+      "step": 31090
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.0005209915192012169,
+      "loss": 1.3906,
+      "step": 31100
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.0005206898737328856,
+      "loss": 1.1817,
+      "step": 31110
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.000520388220721163,
+      "loss": 1.3771,
+      "step": 31120
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.0005200865602760303,
+      "loss": 1.3793,
+      "step": 31130
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.0005197848925074707,
+      "loss": 1.3394,
+      "step": 31140
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.0005194832175254705,
+      "loss": 1.6424,
+      "step": 31150
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 0.0005191815354400182,
+      "loss": 1.356,
+      "step": 31160
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.000518879846361105,
+      "loss": 1.3209,
+      "step": 31170
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.0005185781503987251,
+      "loss": 1.2125,
+      "step": 31180
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.0005182764476628745,
+      "loss": 1.4353,
+      "step": 31190
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.000517974738263552,
+      "loss": 1.3173,
+      "step": 31200
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 0.0005176730223107591,
+      "loss": 1.3404,
+      "step": 31210
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 0.0005173712999144994,
+      "loss": 1.334,
+      "step": 31220
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.0005170695711847786,
+      "loss": 1.5259,
+      "step": 31230
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.0005167678362316053,
+      "loss": 1.4487,
+      "step": 31240
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.0005164660951649901,
+      "loss": 1.3536,
+      "step": 31250
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.0005161643480949456,
+      "loss": 1.3998,
+      "step": 31260
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.0005158625951314868,
+      "loss": 1.4048,
+      "step": 31270
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.0005155608363846309,
+      "loss": 1.1911,
+      "step": 31280
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.0005152590719643972,
+      "loss": 1.1521,
+      "step": 31290
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.0005149573019808068,
+      "loss": 1.4028,
+      "step": 31300
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.0005146555265438833,
+      "loss": 1.3478,
+      "step": 31310
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.0005143537457636517,
+      "loss": 1.2858,
+      "step": 31320
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.0005140519597501397,
+      "loss": 1.4191,
+      "step": 31330
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.0005137501686133762,
+      "loss": 1.5493,
+      "step": 31340
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.0005134483724633923,
+      "loss": 1.396,
+      "step": 31350
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.0005131465714102207,
+      "loss": 1.2926,
+      "step": 31360
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.0005128447655638964,
+      "loss": 1.331,
+      "step": 31370
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.0005125429550344555,
+      "loss": 1.3502,
+      "step": 31380
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.0005122411399319366,
+      "loss": 1.1982,
+      "step": 31390
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.0005119393203663788,
+      "loss": 1.2854,
+      "step": 31400
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 0.000511637496447824,
+      "loss": 1.2116,
+      "step": 31410
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.0005113356682863152,
+      "loss": 1.472,
+      "step": 31420
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.0005110338359918967,
+      "loss": 1.4195,
+      "step": 31430
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.0005107319996746147,
+      "loss": 1.3654,
+      "step": 31440
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.0005104301594445167,
+      "loss": 1.4585,
+      "step": 31450
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 0.0005101283154116516,
+      "loss": 1.3727,
+      "step": 31460
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 0.0005098264676860699,
+      "loss": 1.39,
+      "step": 31470
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.0005095246163778233,
+      "loss": 1.319,
+      "step": 31480
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.0005092227615969642,
+      "loss": 1.3682,
+      "step": 31490
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.0005089209034535477,
+      "loss": 1.2749,
+      "step": 31500
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.0005086190420576287,
+      "loss": 1.2444,
+      "step": 31510
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.0005083171775192641,
+      "loss": 1.2451,
+      "step": 31520
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.0005080153099485117,
+      "loss": 1.4342,
+      "step": 31530
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 0.0005077134394554305,
+      "loss": 1.3564,
+      "step": 31540
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 0.0005074115661500801,
+      "loss": 1.4324,
+      "step": 31550
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.000507109690142522,
+      "loss": 1.2935,
+      "step": 31560
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.0005068078115428178,
+      "loss": 1.3561,
+      "step": 31570
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.0005065059304610307,
+      "loss": 1.512,
+      "step": 31580
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.0005062040470072244,
+      "loss": 1.3915,
+      "step": 31590
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 0.0005059021612914638,
+      "loss": 1.4817,
+      "step": 31600
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 0.0005056002734238142,
+      "loss": 1.2722,
+      "step": 31610
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.0005052983835143419,
+      "loss": 1.3009,
+      "step": 31620
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.0005049964916731141,
+      "loss": 1.1423,
+      "step": 31630
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.0005046945980101985,
+      "loss": 1.3583,
+      "step": 31640
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.0005043927026356637,
+      "loss": 1.314,
+      "step": 31650
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.0005040908056595786,
+      "loss": 1.3959,
+      "step": 31660
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.0005037889071920127,
+      "loss": 1.4114,
+      "step": 31670
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 0.0005034870073430364,
+      "loss": 1.3932,
+      "step": 31680
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.0005031851062227204,
+      "loss": 1.4043,
+      "step": 31690
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.0005028832039411359,
+      "loss": 1.4077,
+      "step": 31700
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.000502581300608354,
+      "loss": 1.2468,
+      "step": 31710
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.0005022793963344471,
+      "loss": 1.336,
+      "step": 31720
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.0005019774912294874,
+      "loss": 1.438,
+      "step": 31730
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.0005016755854035477,
+      "loss": 1.2096,
+      "step": 31740
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.0005013736789667004,
+      "loss": 1.1854,
+      "step": 31750
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.000501071772029019,
+      "loss": 1.5282,
+      "step": 31760
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.0005007698647005765,
+      "loss": 1.3381,
+      "step": 31770
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.0005004679570914466,
+      "loss": 1.4257,
+      "step": 31780
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.0005001660493117025,
+      "loss": 1.3414,
+      "step": 31790
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 0.0004998641414714181,
+      "loss": 1.2905,
+      "step": 31800
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.0004995622336806664,
+      "loss": 1.4541,
+      "step": 31810
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.0004992603260495219,
+      "loss": 1.3861,
+      "step": 31820
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.0004989584186880573,
+      "loss": 1.5102,
+      "step": 31830
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.0004986565117063464,
+      "loss": 1.2027,
+      "step": 31840
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.0004983546052144625,
+      "loss": 1.3218,
+      "step": 31850
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.0004980526993224786,
+      "loss": 1.2546,
+      "step": 31860
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.0004977507941404677,
+      "loss": 1.2268,
+      "step": 31870
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.0004974488897785022,
+      "loss": 1.2829,
+      "step": 31880
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.0004971469863466548,
+      "loss": 1.2112,
+      "step": 31890
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.0004968450839549972,
+      "loss": 1.3136,
+      "step": 31900
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 0.0004965431827136011,
+      "loss": 1.2921,
+      "step": 31910
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 0.0004962412827325377,
+      "loss": 1.3763,
+      "step": 31920
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.0004959393841218776,
+      "loss": 1.3201,
+      "step": 31930
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.0004956374869916913,
+      "loss": 1.41,
+      "step": 31940
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 0.0004953355914520482,
+      "loss": 1.3281,
+      "step": 31950
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.0004950336976130177,
+      "loss": 1.3459,
+      "step": 31960
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.0004947318055846681,
+      "loss": 1.1623,
+      "step": 31970
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.0004944299154770673,
+      "loss": 1.4356,
+      "step": 31980
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.0004941280274002824,
+      "loss": 1.4489,
+      "step": 31990
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.0004938261414643799,
+      "loss": 1.3236,
+      "step": 32000
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.000493524257779425,
+      "loss": 1.3072,
+      "step": 32010
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.0004932223764554832,
+      "loss": 1.222,
+      "step": 32020
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.0004929204976026181,
+      "loss": 1.276,
+      "step": 32030
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 0.0004926186213308923,
+      "loss": 1.2313,
+      "step": 32040
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 0.0004923167477503686,
+      "loss": 1.3459,
+      "step": 32050
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.0004920148769711074,
+      "loss": 1.4082,
+      "step": 32060
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.0004917130091031697,
+      "loss": 1.4137,
+      "step": 32070
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.0004914111442566137,
+      "loss": 1.3133,
+      "step": 32080
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.0004911092825414976,
+      "loss": 1.2841,
+      "step": 32090
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 0.0004908074240678783,
+      "loss": 1.2953,
+      "step": 32100
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 0.0004905055689458113,
+      "loss": 1.3643,
+      "step": 32110
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.0004902037172853507,
+      "loss": 1.4276,
+      "step": 32120
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.0004899018691965502,
+      "loss": 1.2036,
+      "step": 32130
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0004896000247894611,
+      "loss": 1.0171,
+      "step": 32140
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0004892981841741343,
+      "loss": 1.3768,
+      "step": 32150
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 0.0004889963474606187,
+      "loss": 1.1873,
+      "step": 32160
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 0.0004886945147589615,
+      "loss": 1.3283,
+      "step": 32170
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.0004883926861792098,
+      "loss": 1.4417,
+      "step": 32180
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.00048809086183140745,
+      "loss": 1.3419,
+      "step": 32190
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 0.00048778904182559835,
+      "loss": 1.3082,
+      "step": 32200
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.0004874872262718236,
+      "loss": 1.3742,
+      "step": 32210
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.0004871854152801232,
+      "loss": 1.2704,
+      "step": 32220
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.0004868836089605355,
+      "loss": 1.4269,
+      "step": 32230
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.0004865818074230971,
+      "loss": 1.3591,
+      "step": 32240
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.00048628001077784293,
+      "loss": 1.2668,
+      "step": 32250
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.0004859782191348059,
+      "loss": 1.5294,
+      "step": 32260
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.00048567643260401727,
+      "loss": 1.3206,
+      "step": 32270
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.0004853746512955065,
+      "loss": 1.1212,
+      "step": 32280
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.0004850728753193011,
+      "loss": 1.4466,
+      "step": 32290
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.0004847711047854262,
+      "loss": 1.35,
+      "step": 32300
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.00048446933980390595,
+      "loss": 1.3837,
+      "step": 32310
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.0004841675804847612,
+      "loss": 1.1589,
+      "step": 32320
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.0004838658269380121,
+      "loss": 1.4373,
+      "step": 32330
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.0004835640792736754,
+      "loss": 1.4214,
+      "step": 32340
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.0004832623376017665,
+      "loss": 1.3268,
+      "step": 32350
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.0004829606020322986,
+      "loss": 1.355,
+      "step": 32360
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.0004826588726752822,
+      "loss": 1.137,
+      "step": 32370
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.000482357149640726,
+      "loss": 1.1717,
+      "step": 32380
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.0004820554330386363,
+      "loss": 1.049,
+      "step": 32390
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.0004817537229790166,
+      "loss": 1.2653,
+      "step": 32400
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.0004814520195718689,
+      "loss": 1.2835,
+      "step": 32410
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.00048115032292719184,
+      "loss": 1.1798,
+      "step": 32420
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00048084863315498236,
+      "loss": 1.4177,
+      "step": 32430
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.00048054695036523436,
+      "loss": 1.4025,
+      "step": 32440
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00048024527466793924,
+      "loss": 1.2442,
+      "step": 32450
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00047994360617308637,
+      "loss": 1.3924,
+      "step": 32460
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 0.00047964194499066173,
+      "loss": 1.4489,
+      "step": 32470
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.000479340291230649,
+      "loss": 1.2625,
+      "step": 32480
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.00047903864500302933,
+      "loss": 1.4133,
+      "step": 32490
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.00047873700641778077,
+      "loss": 1.2657,
+      "step": 32500
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.00047843537558487894,
+      "loss": 1.3567,
+      "step": 32510
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.00047813375261429635,
+      "loss": 1.3176,
+      "step": 32520
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.0004778321376160026,
+      "loss": 1.2704,
+      "step": 32530
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.00047753053069996476,
+      "loss": 1.531,
+      "step": 32540
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.0004772289319761465,
+      "loss": 1.2007,
+      "step": 32550
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.0004769273415545089,
+      "loss": 1.3299,
+      "step": 32560
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.00047662575954500973,
+      "loss": 1.3264,
+      "step": 32570
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.00047632418605760376,
+      "loss": 1.3375,
+      "step": 32580
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.00047602262120224285,
+      "loss": 1.5111,
+      "step": 32590
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.00047572106508887545,
+      "loss": 1.2075,
+      "step": 32600
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.0004754195178274467,
+      "loss": 1.3277,
+      "step": 32610
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.0004751179795278992,
+      "loss": 1.1681,
+      "step": 32620
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.00047481645030017126,
+      "loss": 1.2808,
+      "step": 32630
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.000474514930254199,
+      "loss": 1.2182,
+      "step": 32640
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.00047421341949991427,
+      "loss": 1.3732,
+      "step": 32650
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.00047391191814724593,
+      "loss": 1.3252,
+      "step": 32660
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.0004736104263061195,
+      "loss": 1.2992,
+      "step": 32670
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 0.00047330894408645677,
+      "loss": 1.3497,
+      "step": 32680
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 0.0004730074715981763,
+      "loss": 1.2938,
+      "step": 32690
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.00047270600895119296,
+      "loss": 1.2908,
+      "step": 32700
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.00047240455625541787,
+      "loss": 1.313,
+      "step": 32710
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.0004721031136207589,
+      "loss": 1.1201,
+      "step": 32720
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.00047180168115711974,
+      "loss": 1.4221,
+      "step": 32730
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 0.000471500258974401,
+      "loss": 1.4025,
+      "step": 32740
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.00047119884718249916,
+      "loss": 1.2526,
+      "step": 32750
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": {
+        "accuracy": 0.532295719844358
+      },
+      "eval_f1": {
+        "f1": 0.5154889606383344
+      },
+      "eval_loss": 1.1265333890914917,
+      "eval_precision": {
+        "precision": 0.5564411969038197
+      },
+      "eval_recall": {
+        "recall": 0.5477056677098556
+      },
+      "eval_runtime": 190.2273,
+      "eval_samples_per_second": 20.265,
+      "eval_steps_per_second": 10.135,
+      "step": 32759
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.0004708974458913066,
+      "loss": 1.3072,
+      "step": 32760
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 0.0004705960552107128,
+      "loss": 1.2928,
+      "step": 32770
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 0.0004702946752506024,
+      "loss": 1.2815,
+      "step": 32780
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 0.00046999330612085657,
+      "loss": 1.1255,
+      "step": 32790
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 0.0004696919479313526,
+      "loss": 1.2747,
+      "step": 32800
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.0004693906007919636,
+      "loss": 1.2441,
+      "step": 32810
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.000469089264812559,
+      "loss": 1.27,
+      "step": 32820
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 0.0004687879401030036,
+      "loss": 1.3701,
+      "step": 32830
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 0.00046848662677315844,
+      "loss": 1.4237,
+      "step": 32840
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.0004681853249328806,
+      "loss": 1.1196,
+      "step": 32850
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.0004678840346920223,
+      "loss": 1.512,
+      "step": 32860
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.00046758275616043247,
+      "loss": 1.2568,
+      "step": 32870
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 0.00046728148944795504,
+      "loss": 1.4645,
+      "step": 32880
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 0.00046698023466442963,
+      "loss": 1.4739,
+      "step": 32890
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 0.0004666789919196922,
+      "loss": 1.3387,
+      "step": 32900
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.00046637776132357327,
+      "loss": 1.4404,
+      "step": 32910
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.0004660765429859002,
+      "loss": 1.1821,
+      "step": 32920
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 0.0004657753370164948,
+      "loss": 1.8769,
+      "step": 32930
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 0.0004654741435251748,
+      "loss": 1.3142,
+      "step": 32940
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 0.00046517296262175333,
+      "loss": 1.3418,
+      "step": 32950
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 0.0004648717944160391,
+      "loss": 1.2576,
+      "step": 32960
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.0004645706390178358,
+      "loss": 1.2876,
+      "step": 32970
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.000464269496536943,
+      "loss": 1.3905,
+      "step": 32980
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 0.00046396836708315493,
+      "loss": 1.3649,
+      "step": 32990
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.0004636672507662617,
+      "loss": 1.3245,
+      "step": 33000
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.00046336614769604837,
+      "loss": 1.4106,
+      "step": 33010
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.0004630650579822945,
+      "loss": 1.276,
+      "step": 33020
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.00046276398173477625,
+      "loss": 1.1453,
+      "step": 33030
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 0.0004624629190632633,
+      "loss": 1.2115,
+      "step": 33040
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 0.0004621618700775217,
+      "loss": 1.1263,
+      "step": 33050
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.00046186083488731155,
+      "loss": 1.414,
+      "step": 33060
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.0004615598136023883,
+      "loss": 1.1799,
+      "step": 33070
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0004612588063325025,
+      "loss": 1.279,
+      "step": 33080
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.00046095781318739927,
+      "loss": 1.2359,
+      "step": 33090
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.0004606568342768187,
+      "loss": 1.3484,
+      "step": 33100
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.00046035586971049586,
+      "loss": 1.4379,
+      "step": 33110
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 0.0004600549195981602,
+      "loss": 1.4189,
+      "step": 33120
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 0.0004597539840495364,
+      "loss": 1.2899,
+      "step": 33130
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.00045945306317434354,
+      "loss": 1.4371,
+      "step": 33140
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.00045915215708229503,
+      "loss": 1.4694,
+      "step": 33150
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 0.00045885126588309975,
+      "loss": 1.2922,
+      "step": 33160
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 0.00045855038968646,
+      "loss": 1.4085,
+      "step": 33170
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 0.00045824952860207404,
+      "loss": 1.347,
+      "step": 33180
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 0.0004579486827396333,
+      "loss": 1.381,
+      "step": 33190
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 0.00045764785220882417,
+      "loss": 1.5498,
+      "step": 33200
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 0.00045734703711932767,
+      "loss": 1.2068,
+      "step": 33210
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00045704623758081886,
+      "loss": 1.2963,
+      "step": 33220
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00045674545370296744,
+      "loss": 1.1968,
+      "step": 33230
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 0.0004564446855954371,
+      "loss": 1.4504,
+      "step": 33240
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 0.0004561439333678858,
+      "loss": 1.4101,
+      "step": 33250
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 0.000455843197129966,
+      "loss": 1.2781,
+      "step": 33260
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.000455542476991324,
+      "loss": 1.3239,
+      "step": 33270
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.00045524177306160036,
+      "loss": 1.1776,
+      "step": 33280
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0004549410854504299,
+      "loss": 1.3882,
+      "step": 33290
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0004546404142674412,
+      "loss": 1.3576,
+      "step": 33300
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 0.000454339759622257,
+      "loss": 1.1007,
+      "step": 33310
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 0.00045403912162449416,
+      "loss": 1.3416,
+      "step": 33320
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.0004537385003837629,
+      "loss": 1.2438,
+      "step": 33330
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.0004534378960096682,
+      "loss": 1.3082,
+      "step": 33340
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.000453137308611808,
+      "loss": 1.1539,
+      "step": 33350
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.0004528367382997749,
+      "loss": 1.1055,
+      "step": 33360
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 0.00045253618518315453,
+      "loss": 1.3305,
+      "step": 33370
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 0.0004522356493715266,
+      "loss": 1.317,
+      "step": 33380
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 0.00045193513097446466,
+      "loss": 1.3217,
+      "step": 33390
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 0.0004516346301015354,
+      "loss": 1.4124,
+      "step": 33400
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 0.0004513341468622999,
+      "loss": 1.4669,
+      "step": 33410
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 0.00045103368136631205,
+      "loss": 1.2534,
+      "step": 33420
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.0004507332337231195,
+      "loss": 1.3457,
+      "step": 33430
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.00045043280404226377,
+      "loss": 1.5724,
+      "step": 33440
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 0.00045013239243327945,
+      "loss": 1.439,
+      "step": 33450
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 0.00044983199900569423,
+      "loss": 1.544,
+      "step": 33460
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 0.00044953162386903025,
+      "loss": 1.3212,
+      "step": 33470
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 0.0004492312671328015,
+      "loss": 1.3035,
+      "step": 33480
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 0.00044893092890651686,
+      "loss": 1.0139,
+      "step": 33490
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 0.0004486306092996773,
+      "loss": 1.2701,
+      "step": 33500
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00044833030842177704,
+      "loss": 1.3412,
+      "step": 33510
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.0004480300263823043,
+      "loss": 1.0816,
+      "step": 33520
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 0.00044772976329073954,
+      "loss": 1.3172,
+      "step": 33530
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.0004474295192565571,
+      "loss": 1.4898,
+      "step": 33540
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.00044712929438922355,
+      "loss": 1.4747,
+      "step": 33550
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 0.0004468290887981991,
+      "loss": 1.3322,
+      "step": 33560
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 0.00044652890259293677,
+      "loss": 1.1698,
+      "step": 33570
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.0004462287358828825,
+      "loss": 1.1561,
+      "step": 33580
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.00044592858877747464,
+      "loss": 1.1885,
+      "step": 33590
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 0.0004456284613861455,
+      "loss": 1.4391,
+      "step": 33600
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 0.00044532835381831887,
+      "loss": 1.7141,
+      "step": 33610
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 0.00044502826618341264,
+      "loss": 1.4475,
+      "step": 33620
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 0.00044472819859083635,
+      "loss": 1.2546,
+      "step": 33630
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.0004444281511499926,
+      "loss": 1.1711,
+      "step": 33640
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.00044412812397027704,
+      "loss": 1.248,
+      "step": 33650
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 0.00044382811716107723,
+      "loss": 1.363,
+      "step": 33660
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 0.000443528130831774,
+      "loss": 1.3491,
+      "step": 33670
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 0.00044322816509174024,
+      "loss": 1.2969,
+      "step": 33680
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 0.0004429282200503413,
+      "loss": 1.2357,
+      "step": 33690
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0004426282958169355,
+      "loss": 1.1101,
+      "step": 33700
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0004423283925008728,
+      "loss": 1.2007,
+      "step": 33710
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.00044202851021149644,
+      "loss": 1.4982,
+      "step": 33720
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.0004417286490581413,
+      "loss": 1.3533,
+      "step": 33730
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00044142880915013443,
+      "loss": 1.1156,
+      "step": 33740
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00044112899059679616,
+      "loss": 1.4041,
+      "step": 33750
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 0.0004408291935074379,
+      "loss": 1.3639,
+      "step": 33760
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 0.0004405294179913636,
+      "loss": 1.1678,
+      "step": 33770
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 0.0004402296641578697,
+      "loss": 1.0913,
+      "step": 33780
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.00043992993211624425,
+      "loss": 1.1865,
+      "step": 33790
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.0004396302219757678,
+      "loss": 1.291,
+      "step": 33800
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 0.0004393305338457126,
+      "loss": 1.314,
+      "step": 33810
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 0.00043903086783534274,
+      "loss": 1.2264,
+      "step": 33820
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 0.00043873122405391473,
+      "loss": 1.3849,
+      "step": 33830
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 0.00043843160261067656,
+      "loss": 1.43,
+      "step": 33840
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.0004381320036148683,
+      "loss": 1.2667,
+      "step": 33850
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.0004378324271757219,
+      "loss": 1.3175,
+      "step": 33860
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 0.00043753287340246046,
+      "loss": 1.2426,
+      "step": 33870
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 0.00043723334240429993,
+      "loss": 1.3312,
+      "step": 33880
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.0004369338342904467,
+      "loss": 1.3243,
+      "step": 33890
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.00043663434917010015,
+      "loss": 1.281,
+      "step": 33900
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 0.00043633488715244995,
+      "loss": 1.1786,
+      "step": 33910
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 0.0004360354483466782,
+      "loss": 1.1661,
+      "step": 33920
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.0004357360328619584,
+      "loss": 1.3953,
+      "step": 33930
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.0004354366408074553,
+      "loss": 1.1139,
+      "step": 33940
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 0.0004351372722923253,
+      "loss": 1.4813,
+      "step": 33950
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 0.0004348379274257163,
+      "loss": 1.2341,
+      "step": 33960
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 0.0004345386063167672,
+      "loss": 1.3576,
+      "step": 33970
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 0.00043423930907460873,
+      "loss": 1.3797,
+      "step": 33980
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 0.0004339400358083627,
+      "loss": 1.3352,
+      "step": 33990
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 0.00043364078662714184,
+      "loss": 1.1802,
+      "step": 34000
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.00043334156164005083,
+      "loss": 1.4299,
+      "step": 34010
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.0004330423609561848,
+      "loss": 1.3838,
+      "step": 34020
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 0.00043274318468463057,
+      "loss": 1.2414,
+      "step": 34030
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 0.00043244403293446585,
+      "loss": 1.4269,
+      "step": 34040
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 0.0004321449058147589,
+      "loss": 1.3718,
+      "step": 34050
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 0.0004318458034345702,
+      "loss": 1.3573,
+      "step": 34060
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 0.00043154672590295005,
+      "loss": 1.4109,
+      "step": 34070
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.00043124767332894013,
+      "loss": 1.2255,
+      "step": 34080
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.0004309486458215731,
+      "loss": 1.2714,
+      "step": 34090
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.00043064964348987246,
+      "loss": 1.2782,
+      "step": 34100
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.0004303506664428526,
+      "loss": 1.3067,
+      "step": 34110
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 0.0004300517147895183,
+      "loss": 1.2769,
+      "step": 34120
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 0.00042975278863886546,
+      "loss": 1.2583,
+      "step": 34130
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.00042945388809988074,
+      "loss": 1.1077,
+      "step": 34140
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.0004291550132815409,
+      "loss": 1.3682,
+      "step": 34150
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.0004288561642928142,
+      "loss": 1.3351,
+      "step": 34160
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.0004285573412426589,
+      "loss": 1.2713,
+      "step": 34170
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 0.0004282585442400234,
+      "loss": 1.2615,
+      "step": 34180
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 0.00042795977339384777,
+      "loss": 1.2001,
+      "step": 34190
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 0.00042766102881306134,
+      "loss": 1.2469,
+      "step": 34200
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 0.0004273623106065849,
+      "loss": 1.2205,
+      "step": 34210
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.0004270636188833289,
+      "loss": 1.5071,
+      "step": 34220
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.0004267649537521942,
+      "loss": 1.451,
+      "step": 34230
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 0.00042646631532207235,
+      "loss": 1.1487,
+      "step": 34240
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 0.000426167703701845,
+      "loss": 1.3217,
+      "step": 34250
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 0.0004258691190003836,
+      "loss": 1.2258,
+      "step": 34260
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 0.00042557056132655057,
+      "loss": 1.2706,
+      "step": 34270
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 0.0004252720307891978,
+      "loss": 1.7372,
+      "step": 34280
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 0.00042497352749716773,
+      "loss": 1.1559,
+      "step": 34290
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.0004246750515592927,
+      "loss": 1.3996,
+      "step": 34300
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.0004243766030843946,
+      "loss": 1.2233,
+      "step": 34310
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 0.0004240781821812865,
+      "loss": 1.2468,
+      "step": 34320
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 0.0004237797889587699,
+      "loss": 1.286,
+      "step": 34330
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 0.0004234814235256378,
+      "loss": 1.3083,
+      "step": 34340
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.00042318308599067166,
+      "loss": 1.3836,
+      "step": 34350
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.0004228847764626436,
+      "loss": 1.5828,
+      "step": 34360
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.0004225864950503152,
+      "loss": 1.259,
+      "step": 34370
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.00042228824186243795,
+      "loss": 1.249,
+      "step": 34380
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 0.0004219900170077531,
+      "loss": 1.369,
+      "step": 34390
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 0.0004216918205949913,
+      "loss": 1.2843,
+      "step": 34400
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.00042139365273287296,
+      "loss": 1.3679,
+      "step": 34410
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.0004210955135301082,
+      "loss": 1.1884,
+      "step": 34420
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 0.00042079740309539684,
+      "loss": 1.3216,
+      "step": 34430
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 0.0004204993215374273,
+      "loss": 1.3334,
+      "step": 34440
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 0.0004202012689648789,
+      "loss": 1.2871,
+      "step": 34450
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 0.0004199032454864189,
+      "loss": 1.2873,
+      "step": 34460
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.0004196052512107054,
+      "loss": 1.3029,
+      "step": 34470
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.0004193072862463847,
+      "loss": 1.2683,
+      "step": 34480
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 0.0004190093507020929,
+      "loss": 1.3812,
+      "step": 34490
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 0.0004187114446864555,
+      "loss": 1.1789,
+      "step": 34500
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 0.0004184135683080868,
+      "loss": 1.3416,
+      "step": 34510
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 0.0004181157216755908,
+      "loss": 1.3188,
+      "step": 34520
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.0004178179048975603,
+      "loss": 1.3811,
+      "step": 34530
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.0004175201180825773,
+      "loss": 1.2802,
+      "step": 34540
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 0.00041722236133921297,
+      "loss": 1.3054,
+      "step": 34550
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 0.00041692463477602747,
+      "loss": 1.2783,
+      "step": 34560
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.0004166269385015697,
+      "loss": 1.3791,
+      "step": 34570
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 0.00041632927262437816,
+      "loss": 1.1963,
+      "step": 34580
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.0004160316372529794,
+      "loss": 1.1996,
+      "step": 34590
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.00041573403249588984,
+      "loss": 1.3441,
+      "step": 34600
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.00041543645846161366,
+      "loss": 1.1905,
+      "step": 34610
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.00041513891525864454,
+      "loss": 1.2577,
+      "step": 34620
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.000414841402995465,
+      "loss": 1.2258,
+      "step": 34630
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 0.00041454392178054574,
+      "loss": 1.2097,
+      "step": 34640
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 0.00041424647172234663,
+      "loss": 1.3121,
+      "step": 34650
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.00041394905292931603,
+      "loss": 1.1876,
+      "step": 34660
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.0004136516655098906,
+      "loss": 1.2824,
+      "step": 34670
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.0004133543095724961,
+      "loss": 1.3673,
+      "step": 34680
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": {
+        "accuracy": 0.5553826199740597
+      },
+      "eval_f1": {
+        "f1": 0.5594361896393714
+      },
+      "eval_loss": 1.0845227241516113,
+      "eval_precision": {
+        "precision": 0.5621230845137732
+      },
+      "eval_recall": {
+        "recall": 0.5721227129539878
+      },
+      "eval_runtime": 179.9876,
+      "eval_samples_per_second": 21.418,
+      "eval_steps_per_second": 10.712,
+      "step": 34686
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.00041305698522554635,
+      "loss": 1.1621,
+      "step": 34690
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 0.00041275969257744397,
+      "loss": 1.1116,
+      "step": 34700
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 0.00041246243173657986,
+      "loss": 1.2467,
+      "step": 34710
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 0.00041216520281133325,
+      "loss": 1.1756,
+      "step": 34720
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 0.000411868005910072,
+      "loss": 1.0914,
+      "step": 34730
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.0004115708411411521,
+      "loss": 1.2702,
+      "step": 34740
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.00041127370861291756,
+      "loss": 1.3949,
+      "step": 34750
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 0.00041097660843370144,
+      "loss": 1.1813,
+      "step": 34760
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 0.00041067954071182395,
+      "loss": 1.3767,
+      "step": 34770
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.00041038250555559466,
+      "loss": 1.0969,
+      "step": 34780
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.00041008550307331003,
+      "loss": 1.3206,
+      "step": 34790
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.0004097885333732554,
+      "loss": 1.4119,
+      "step": 34800
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.0004094915965637041,
+      "loss": 1.4198,
+      "step": 34810
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 0.0004091946927529171,
+      "loss": 1.3672,
+      "step": 34820
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 0.00040889782204914373,
+      "loss": 1.3114,
+      "step": 34830
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 0.000408600984560621,
+      "loss": 1.5056,
+      "step": 34840
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.00040830418039557387,
+      "loss": 1.2495,
+      "step": 34850
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.0004080074096622153,
+      "loss": 1.0847,
+      "step": 34860
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.0004077106724687457,
+      "loss": 1.0688,
+      "step": 34870
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.0004074139689233538,
+      "loss": 1.2333,
+      "step": 34880
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 0.00040711729913421563,
+      "loss": 1.4418,
+      "step": 34890
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 0.0004068206632094947,
+      "loss": 1.3232,
+      "step": 34900
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 0.00040652406125734307,
+      "loss": 1.2772,
+      "step": 34910
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 0.00040622749338589946,
+      "loss": 1.3507,
+      "step": 34920
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 0.0004059309597032906,
+      "loss": 1.132,
+      "step": 34930
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 0.00040563446031763084,
+      "loss": 1.2173,
+      "step": 34940
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.0004053379953370217,
+      "loss": 1.2749,
+      "step": 34950
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.0004050415648695526,
+      "loss": 1.1855,
+      "step": 34960
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 0.00040474516902329994,
+      "loss": 1.3556,
+      "step": 34970
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 0.0004044488079063276,
+      "loss": 1.2579,
+      "step": 34980
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00040415248162668717,
+      "loss": 1.2194,
+      "step": 34990
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00040385619029241693,
+      "loss": 1.3279,
+      "step": 35000
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.000403559934011543,
+      "loss": 1.2924,
+      "step": 35010
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.00040326371289207854,
+      "loss": 1.184,
+      "step": 35020
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 0.0004029675270420232,
+      "loss": 1.2529,
+      "step": 35030
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 0.0004026713765693652,
+      "loss": 1.197,
+      "step": 35040
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.0004023752615820785,
+      "loss": 1.4416,
+      "step": 35050
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.0004020791821881247,
+      "loss": 1.2365,
+      "step": 35060
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 0.0004017831384954526,
+      "loss": 1.2452,
+      "step": 35070
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 0.00040148713061199747,
+      "loss": 1.2745,
+      "step": 35080
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 0.0004011911586456822,
+      "loss": 1.2988,
+      "step": 35090
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 0.000400895222704416,
+      "loss": 1.2941,
+      "step": 35100
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.00040059932289609513,
+      "loss": 1.2082,
+      "step": 35110
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 0.0004003034593286028,
+      "loss": 1.1569,
+      "step": 35120
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 0.00040000763210980874,
+      "loss": 1.2027,
+      "step": 35130
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 0.0003997118413475698,
+      "loss": 1.3531,
+      "step": 35140
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 0.0003994160871497294,
+      "loss": 1.3077,
+      "step": 35150
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.000399120369624117,
+      "loss": 1.5495,
+      "step": 35160
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.00039882468887855006,
+      "loss": 1.2649,
+      "step": 35170
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 0.00039852904502083104,
+      "loss": 1.133,
+      "step": 35180
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 0.00039823343815875045,
+      "loss": 1.1089,
+      "step": 35190
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.00039793786840008416,
+      "loss": 1.3199,
+      "step": 35200
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.0003976423358525949,
+      "loss": 1.2802,
+      "step": 35210
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.0003973468406240321,
+      "loss": 1.4159,
+      "step": 35220
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.0003970513828221313,
+      "loss": 1.1331,
+      "step": 35230
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.00039675596255461435,
+      "loss": 1.3066,
+      "step": 35240
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.0003964605799291897,
+      "loss": 1.1429,
+      "step": 35250
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.0003961652350535517,
+      "loss": 1.1558,
+      "step": 35260
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.00039586992803538137,
+      "loss": 1.3597,
+      "step": 35270
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 0.0003955746589823455,
+      "loss": 1.2347,
+      "step": 35280
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 0.00039527942800209726,
+      "loss": 1.3914,
+      "step": 35290
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.000394984235202276,
+      "loss": 1.2175,
+      "step": 35300
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.0003946890806905069,
+      "loss": 1.2971,
+      "step": 35310
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 0.0003943939645744017,
+      "loss": 1.2145,
+      "step": 35320
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 0.00039409888696155743,
+      "loss": 1.3394,
+      "step": 35330
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 0.00039380384795955736,
+      "loss": 1.2953,
+      "step": 35340
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 0.0003935088476759711,
+      "loss": 0.9967,
+      "step": 35350
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.00039321388621835353,
+      "loss": 1.1092,
+      "step": 35360
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.0003929189636942458,
+      "loss": 1.4514,
+      "step": 35370
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 0.0003926240802111746,
+      "loss": 1.3606,
+      "step": 35380
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 0.00039232923587665254,
+      "loss": 1.325,
+      "step": 35390
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 0.000392034430798178,
+      "loss": 1.0505,
+      "step": 35400
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.00039173966508323485,
+      "loss": 1.4851,
+      "step": 35410
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.0003914449388392926,
+      "loss": 1.2219,
+      "step": 35420
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.0003911502521738068,
+      "loss": 1.4181,
+      "step": 35430
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.00039085560519421794,
+      "loss": 1.1061,
+      "step": 35440
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 0.0003905609980079526,
+      "loss": 1.248,
+      "step": 35450
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 0.00039026643072242263,
+      "loss": 1.1194,
+      "step": 35460
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.00038997190344502487,
+      "loss": 1.37,
+      "step": 35470
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.00038967741628314267,
+      "loss": 1.3905,
+      "step": 35480
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 0.0003893829693441434,
+      "loss": 1.1066,
+      "step": 35490
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 0.0003890885627353812,
+      "loss": 1.4736,
+      "step": 35500
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.0003887941965641943,
+      "loss": 1.2751,
+      "step": 35510
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.00038849987093790664,
+      "loss": 1.2882,
+      "step": 35520
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 0.00038820558596382756,
+      "loss": 1.2649,
+      "step": 35530
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 0.00038791134174925135,
+      "loss": 1.0624,
+      "step": 35540
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 0.00038761713840145743,
+      "loss": 1.2262,
+      "step": 35550
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 0.00038732297602771064,
+      "loss": 1.2645,
+      "step": 35560
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.00038702885473526036,
+      "loss": 1.4366,
+      "step": 35570
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.0003867347746313414,
+      "loss": 1.2932,
+      "step": 35580
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 0.00038644073582317354,
+      "loss": 1.0602,
+      "step": 35590
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 0.00038614673841796086,
+      "loss": 1.2689,
+      "step": 35600
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 0.00038585278252289366,
+      "loss": 1.2157,
+      "step": 35610
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 0.00038555886824514547,
+      "loss": 1.1997,
+      "step": 35620
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 0.0003852649956918763,
+      "loss": 1.3589,
+      "step": 35630
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.0003849711649702296,
+      "loss": 1.2771,
+      "step": 35640
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 0.0003846773761873341,
+      "loss": 1.2148,
+      "step": 35650
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.0003843836294503036,
+      "loss": 1.2696,
+      "step": 35660
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.00038408992486623584,
+      "loss": 1.2066,
+      "step": 35670
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 0.0003837962625422139,
+      "loss": 1.2548,
+      "step": 35680
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 0.0003835026425853049,
+      "loss": 1.2661,
+      "step": 35690
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 0.0003832090651025607,
+      "loss": 1.2982,
+      "step": 35700
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 0.00038291553020101786,
+      "loss": 1.357,
+      "step": 35710
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 0.00038262203798769735,
+      "loss": 1.2948,
+      "step": 35720
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 0.000382328588569604,
+      "loss": 1.1903,
+      "step": 35730
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.0003820351820537282,
+      "loss": 1.1776,
+      "step": 35740
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.00038174181854704325,
+      "loss": 1.3431,
+      "step": 35750
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 0.0003814484981565084,
+      "loss": 1.3489,
+      "step": 35760
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 0.00038115522098906574,
+      "loss": 1.149,
+      "step": 35770
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 0.00038086198715164225,
+      "loss": 1.3054,
+      "step": 35780
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 0.0003805687967511491,
+      "loss": 1.2456,
+      "step": 35790
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 0.0003802756498944816,
+      "loss": 1.5201,
+      "step": 35800
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 0.0003799825466885193,
+      "loss": 1.3203,
+      "step": 35810
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.0003796894872401254,
+      "loss": 1.1953,
+      "step": 35820
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.0003793964716561476,
+      "loss": 1.3314,
+      "step": 35830
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 0.0003791035000434174,
+      "loss": 1.1225,
+      "step": 35840
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 0.0003788105725087503,
+      "loss": 1.3604,
+      "step": 35850
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.0003785176891589458,
+      "loss": 1.2848,
+      "step": 35860
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.0003782248501007873,
+      "loss": 1.1991,
+      "step": 35870
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.00037793205544104166,
+      "loss": 1.0993,
+      "step": 35880
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.00037763930528646026,
+      "loss": 1.3085,
+      "step": 35890
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 0.0003773465997437777,
+      "loss": 1.3069,
+      "step": 35900
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 0.00037705393891971224,
+      "loss": 1.3493,
+      "step": 35910
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 0.0003767613229209665,
+      "loss": 1.1889,
+      "step": 35920
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.000376468751854226,
+      "loss": 1.2469,
+      "step": 35930
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.00037617622582616044,
+      "loss": 1.4553,
+      "step": 35940
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.0003758837449434228,
+      "loss": 1.0797,
+      "step": 35950
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.00037559130931264965,
+      "loss": 1.3007,
+      "step": 35960
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 0.0003752989190404611,
+      "loss": 1.3574,
+      "step": 35970
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 0.00037500657423346063,
+      "loss": 1.1673,
+      "step": 35980
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.00037471427499823534,
+      "loss": 1.2788,
+      "step": 35990
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.0003744220214413556,
+      "loss": 1.1508,
+      "step": 36000
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 0.000374129813669375,
+      "loss": 1.2657,
+      "step": 36010
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 0.00037383765178883066,
+      "loss": 1.2294,
+      "step": 36020
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.00037354553590624297,
+      "loss": 1.2384,
+      "step": 36030
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.0003732534661281151,
+      "loss": 1.3487,
+      "step": 36040
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 0.00037296144256093413,
+      "loss": 1.3554,
+      "step": 36050
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 0.00037266946531116954,
+      "loss": 1.2402,
+      "step": 36060
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.0003723775344852747,
+      "loss": 1.4537,
+      "step": 36070
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.00037208565018968543,
+      "loss": 1.3144,
+      "step": 36080
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 0.0003717938125308207,
+      "loss": 1.0749,
+      "step": 36090
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 0.00037150202161508294,
+      "loss": 1.3439,
+      "step": 36100
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.00037121027754885685,
+      "loss": 1.2742,
+      "step": 36110
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.00037091858043851054,
+      "loss": 1.1754,
+      "step": 36120
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 0.000370626930390395,
+      "loss": 1.2819,
+      "step": 36130
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 0.0003703353275108435,
+      "loss": 1.1174,
+      "step": 36140
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.00037004377190617285,
+      "loss": 1.4172,
+      "step": 36150
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.00036975226368268213,
+      "loss": 1.1371,
+      "step": 36160
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 0.0003694608029466534,
+      "loss": 1.258,
+      "step": 36170
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.0003691693898043514,
+      "loss": 1.1931,
+      "step": 36180
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.00036887802436202303,
+      "loss": 1.1473,
+      "step": 36190
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 0.00036858670672589884,
+      "loss": 1.2873,
+      "step": 36200
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 0.0003682954370021908,
+      "loss": 1.4725,
+      "step": 36210
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 0.000368004215297094,
+      "loss": 1.2486,
+      "step": 36220
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 0.0003677130417167861,
+      "loss": 1.3322,
+      "step": 36230
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.0003674219163674268,
+      "loss": 1.2421,
+      "step": 36240
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.0003671308393551589,
+      "loss": 1.2459,
+      "step": 36250
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 0.00036683981078610686,
+      "loss": 1.3441,
+      "step": 36260
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 0.00036654883076637763,
+      "loss": 1.3031,
+      "step": 36270
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.00036625789940206095,
+      "loss": 1.2798,
+      "step": 36280
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 0.00036596701679922816,
+      "loss": 1.1078,
+      "step": 36290
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.0003656761830639334,
+      "loss": 1.2459,
+      "step": 36300
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.00036538539830221274,
+      "loss": 1.2681,
+      "step": 36310
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.00036509466262008385,
+      "loss": 1.2906,
+      "step": 36320
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.0003648039761235477,
+      "loss": 1.3193,
+      "step": 36330
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.00036451333891858613,
+      "loss": 1.3212,
+      "step": 36340
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.0003642227511111641,
+      "loss": 1.0846,
+      "step": 36350
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 0.00036393221280722754,
+      "loss": 1.4399,
+      "step": 36360
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 0.0003636417241127048,
+      "loss": 1.1885,
+      "step": 36370
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 0.00036335128513350647,
+      "loss": 1.2013,
+      "step": 36380
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 0.00036306089597552436,
+      "loss": 1.3481,
+      "step": 36390
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.00036277055674463254,
+      "loss": 1.2334,
+      "step": 36400
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.0003624802675466869,
+      "loss": 1.3493,
+      "step": 36410
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 0.0003621900284875247,
+      "loss": 1.169,
+      "step": 36420
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 0.0003618998396729655,
+      "loss": 1.0292,
+      "step": 36430
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 0.00036160970120880997,
+      "loss": 1.3127,
+      "step": 36440
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00036131961320084056,
+      "loss": 1.1866,
+      "step": 36450
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00036102957575482187,
+      "loss": 1.098,
+      "step": 36460
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 0.00036073958897649914,
+      "loss": 1.1919,
+      "step": 36470
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 0.0003604496529716002,
+      "loss": 1.1552,
+      "step": 36480
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.0003601597678458333,
+      "loss": 1.2796,
+      "step": 36490
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.00035986993370488865,
+      "loss": 1.0696,
+      "step": 36500
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.00035958015065443817,
+      "loss": 1.3401,
+      "step": 36510
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.00035929041880013466,
+      "loss": 1.2204,
+      "step": 36520
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 0.00035900073824761235,
+      "loss": 1.3608,
+      "step": 36530
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 0.00035871110910248703,
+      "loss": 1.3219,
+      "step": 36540
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.0003584215314703555,
+      "loss": 1.1538,
+      "step": 36550
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.0003581320054567959,
+      "loss": 1.2279,
+      "step": 36560
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 0.00035784253116736756,
+      "loss": 1.2057,
+      "step": 36570
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 0.0003575531087076104,
+      "loss": 1.2543,
+      "step": 36580
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 0.00035726373818304656,
+      "loss": 1.1545,
+      "step": 36590
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 0.00035697441969917816,
+      "loss": 1.2896,
+      "step": 36600
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.0003566851533614893,
+      "loss": 1.241,
+      "step": 36610
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": {
+        "accuracy": 0.5317769130998703
+      },
+      "eval_f1": {
+        "f1": 0.5281353810835646
+      },
+      "eval_loss": 1.1498997211456299,
+      "eval_precision": {
+        "precision": 0.5775200417239811
+      },
+      "eval_recall": {
+        "recall": 0.5464467844676986
+      },
+      "eval_runtime": 181.4138,
+      "eval_samples_per_second": 21.25,
+      "eval_steps_per_second": 10.628,
+      "step": 36613
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.0003563959392754441,
+      "loss": 1.1548,
+      "step": 36620
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 0.00035610677754648816,
+      "loss": 1.3273,
+      "step": 36630
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 0.00035581766828004824,
+      "loss": 1.4879,
+      "step": 36640
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 0.00035552861158153127,
+      "loss": 1.2715,
+      "step": 36650
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 0.0003552396075563257,
+      "loss": 1.1791,
+      "step": 36660
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 0.00035495065630980043,
+      "loss": 1.2419,
+      "step": 36670
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 0.00035466175794730506,
+      "loss": 1.2273,
+      "step": 36680
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 0.00035437291257417014,
+      "loss": 1.1562,
+      "step": 36690
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.00035408412029570667,
+      "loss": 1.222,
+      "step": 36700
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.00035379538121720647,
+      "loss": 1.1811,
+      "step": 36710
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.00035350669544394187,
+      "loss": 1.1933,
+      "step": 36720
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 0.00035321806308116576,
+      "loss": 1.303,
+      "step": 36730
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.0003529294842341118,
+      "loss": 1.1722,
+      "step": 36740
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.00035264095900799377,
+      "loss": 1.3004,
+      "step": 36750
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.00035235248750800576,
+      "loss": 1.1781,
+      "step": 36760
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.0003520640698393231,
+      "loss": 1.1024,
+      "step": 36770
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 0.0003517757061071006,
+      "loss": 1.2749,
+      "step": 36780
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 0.00035148739641647386,
+      "loss": 1.2207,
+      "step": 36790
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 0.00035119914087255885,
+      "loss": 1.1902,
+      "step": 36800
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 0.00035091093958045127,
+      "loss": 1.076,
+      "step": 36810
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.0003506227926452278,
+      "loss": 1.148,
+      "step": 36820
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.00035033470017194456,
+      "loss": 1.2094,
+      "step": 36830
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 0.00035004666226563845,
+      "loss": 1.2694,
+      "step": 36840
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 0.0003497586790313261,
+      "loss": 1.0114,
+      "step": 36850
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 0.00034947075057400404,
+      "loss": 1.2797,
+      "step": 36860
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 0.0003491828769986495,
+      "loss": 1.3843,
+      "step": 36870
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.00034889505841021904,
+      "loss": 1.4654,
+      "step": 36880
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.0003486072949136492,
+      "loss": 1.3881,
+      "step": 36890
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.00034831958661385714,
+      "loss": 1.1556,
+      "step": 36900
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.00034803193361573887,
+      "loss": 1.241,
+      "step": 36910
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 0.0003477443360241713,
+      "loss": 1.2936,
+      "step": 36920
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 0.0003474567939440103,
+      "loss": 1.174,
+      "step": 36930
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00034716930748009166,
+      "loss": 1.2874,
+      "step": 36940
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00034688187673723137,
+      "loss": 1.2312,
+      "step": 36950
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 0.00034659450182022457,
+      "loss": 1.2579,
+      "step": 36960
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 0.00034630718283384633,
+      "loss": 1.2673,
+      "step": 36970
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 0.0003460199198828513,
+      "loss": 1.2528,
+      "step": 36980
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 0.00034573271307197345,
+      "loss": 1.2022,
+      "step": 36990
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 0.00034544556250592676,
+      "loss": 1.2006,
+      "step": 37000
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 0.00034515846828940436,
+      "loss": 1.1486,
+      "step": 37010
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 0.00034487143052707846,
+      "loss": 1.1352,
+      "step": 37020
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.00034458444932360186,
+      "loss": 1.2711,
+      "step": 37030
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.0003442975247836053,
+      "loss": 1.1487,
+      "step": 37040
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 0.0003440106570117002,
+      "loss": 1.1447,
+      "step": 37050
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 0.0003437238461124763,
+      "loss": 1.1887,
+      "step": 37060
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.0003434370921905028,
+      "loss": 1.2543,
+      "step": 37070
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.00034315039535032867,
+      "loss": 1.1668,
+      "step": 37080
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 0.0003428637556964813,
+      "loss": 1.1835,
+      "step": 37090
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 0.000342577173333468,
+      "loss": 1.2849,
+      "step": 37100
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 0.00034229064836577457,
+      "loss": 1.145,
+      "step": 37110
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 0.000342004180897866,
+      "loss": 1.221,
+      "step": 37120
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.00034171777103418676,
+      "loss": 1.2846,
+      "step": 37130
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.0003414314188791596,
+      "loss": 1.2273,
+      "step": 37140
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.00034114512453718704,
+      "loss": 1.0219,
+      "step": 37150
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.00034085888811264997,
+      "loss": 1.2793,
+      "step": 37160
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 0.00034057270970990784,
+      "loss": 1.1606,
+      "step": 37170
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 0.00034028658943330016,
+      "loss": 1.1581,
+      "step": 37180
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.0003400005273871439,
+      "loss": 1.3327,
+      "step": 37190
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.00033971452367573555,
+      "loss": 1.1686,
+      "step": 37200
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 0.0003394285784033503,
+      "loss": 1.1321,
+      "step": 37210
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 0.0003391426916742417,
+      "loss": 1.2609,
+      "step": 37220
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 0.0003388568635926424,
+      "loss": 1.2408,
+      "step": 37230
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 0.0003385710942627632,
+      "loss": 1.1465,
+      "step": 37240
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 0.0003382853837887938,
+      "loss": 0.931,
+      "step": 37250
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 0.00033799973227490246,
+      "loss": 1.27,
+      "step": 37260
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 0.0003377141398252355,
+      "loss": 1.2039,
+      "step": 37270
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.0003374286065439184,
+      "loss": 1.1148,
+      "step": 37280
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.0003371431325350547,
+      "loss": 1.1849,
+      "step": 37290
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 0.0003368577179027258,
+      "loss": 1.2413,
+      "step": 37300
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 0.00033657236275099275,
+      "loss": 1.0847,
+      "step": 37310
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.0003362870671838934,
+      "loss": 1.0531,
+      "step": 37320
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.0003360018313054454,
+      "loss": 1.316,
+      "step": 37330
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.00033571665521964325,
+      "loss": 1.1227,
+      "step": 37340
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.00033543153903046037,
+      "loss": 1.3868,
+      "step": 37350
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.00033514648284184846,
+      "loss": 1.4158,
+      "step": 37360
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 0.00033486148675773697,
+      "loss": 1.2597,
+      "step": 37370
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 0.00033457655088203346,
+      "loss": 0.9719,
+      "step": 37380
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 0.00033429167531862395,
+      "loss": 1.0494,
+      "step": 37390
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00033400686017137193,
+      "loss": 1.1833,
+      "step": 37400
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00033372210554411934,
+      "loss": 1.3747,
+      "step": 37410
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 0.00033343741154068577,
+      "loss": 1.5163,
+      "step": 37420
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 0.00033315277826486854,
+      "loss": 1.2932,
+      "step": 37430
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.0003328682058204433,
+      "loss": 1.3332,
+      "step": 37440
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.0003325836943111632,
+      "loss": 1.2144,
+      "step": 37450
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 0.00033229924384075925,
+      "loss": 1.1852,
+      "step": 37460
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 0.00033201485451294044,
+      "loss": 1.3218,
+      "step": 37470
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 0.0003317305264313927,
+      "loss": 1.2416,
+      "step": 37480
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 0.0003314462596997807,
+      "loss": 1.227,
+      "step": 37490
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 0.00033116205442174594,
+      "loss": 1.2896,
+      "step": 37500
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 0.0003308779107009077,
+      "loss": 1.2222,
+      "step": 37510
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 0.0003305938286408632,
+      "loss": 1.2723,
+      "step": 37520
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.00033030980834518645,
+      "loss": 1.3063,
+      "step": 37530
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.0003300258499174299,
+      "loss": 1.1504,
+      "step": 37540
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 0.0003297419534611225,
+      "loss": 1.1132,
+      "step": 37550
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 0.0003294581190797711,
+      "loss": 1.3024,
+      "step": 37560
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.00032917434687686,
+      "loss": 1.1257,
+      "step": 37570
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.0003288906369558504,
+      "loss": 1.0688,
+      "step": 37580
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.0003286069894201814,
+      "loss": 1.1599,
+      "step": 37590
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.0003283234043732689,
+      "loss": 1.1923,
+      "step": 37600
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 0.00032803988191850575,
+      "loss": 1.2157,
+      "step": 37610
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 0.00032775642215926296,
+      "loss": 1.1677,
+      "step": 37620
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 0.00032747302519888746,
+      "loss": 1.1471,
+      "step": 37630
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 0.00032718969114070454,
+      "loss": 1.3661,
+      "step": 37640
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.0003269064200880155,
+      "loss": 1.1045,
+      "step": 37650
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.0003266232121440991,
+      "loss": 1.3216,
+      "step": 37660
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 0.00032634006741221133,
+      "loss": 1.2846,
+      "step": 37670
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 0.00032605698599558463,
+      "loss": 1.1084,
+      "step": 37680
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 0.00032577396799742875,
+      "loss": 1.344,
+      "step": 37690
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 0.00032549101352093016,
+      "loss": 1.1252,
+      "step": 37700
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.00032520812266925204,
+      "loss": 1.2917,
+      "step": 37710
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.0003249252955455349,
+      "loss": 1.2326,
+      "step": 37720
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 0.0003246425322528953,
+      "loss": 1.1449,
+      "step": 37730
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 0.00032435983289442674,
+      "loss": 1.4343,
+      "step": 37740
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 0.00032407719757319997,
+      "loss": 1.3219,
+      "step": 37750
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.0003237946263922615,
+      "loss": 1.1324,
+      "step": 37760
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.00032351211945463545,
+      "loss": 1.2337,
+      "step": 37770
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.0003232296768633215,
+      "loss": 1.1533,
+      "step": 37780
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.0003229472987212965,
+      "loss": 1.2927,
+      "step": 37790
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 0.00032266498513151365,
+      "loss": 1.0759,
+      "step": 37800
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 0.0003223827361969026,
+      "loss": 1.307,
+      "step": 37810
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.0003221005520203696,
+      "loss": 1.1211,
+      "step": 37820
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.00032181843270479693,
+      "loss": 1.0802,
+      "step": 37830
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 0.0003215363783530435,
+      "loss": 1.2235,
+      "step": 37840
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 0.00032125438906794446,
+      "loss": 1.4051,
+      "step": 37850
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 0.0003209724649523114,
+      "loss": 1.1972,
+      "step": 37860
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 0.00032069060610893154,
+      "loss": 1.3673,
+      "step": 37870
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 0.0003204088126405692,
+      "loss": 0.944,
+      "step": 37880
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 0.00032012708464996403,
+      "loss": 1.3239,
+      "step": 37890
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.0003198454222398327,
+      "loss": 1.1424,
+      "step": 37900
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.000319563825512867,
+      "loss": 1.2759,
+      "step": 37910
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.00031928229457173533,
+      "loss": 1.1436,
+      "step": 37920
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.00031900082951908217,
+      "loss": 1.0582,
+      "step": 37930
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 0.0003187194304575276,
+      "loss": 1.2655,
+      "step": 37940
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 0.0003184380974896681,
+      "loss": 1.1242,
+      "step": 37950
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.00031815683071807554,
+      "loss": 1.2732,
+      "step": 37960
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.0003178756302452981,
+      "loss": 1.4371,
+      "step": 37970
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.00031759449617385963,
+      "loss": 1.2628,
+      "step": 37980
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.00031731342860625964,
+      "loss": 1.2014,
+      "step": 37990
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.00031703242764497344,
+      "loss": 1.2273,
+      "step": 38000
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 0.00031675149339245237,
+      "loss": 1.2852,
+      "step": 38010
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 0.0003164706259511228,
+      "loss": 1.3147,
+      "step": 38020
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.0003161898254233877,
+      "loss": 1.1807,
+      "step": 38030
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.0003159090919116247,
+      "loss": 1.2854,
+      "step": 38040
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 0.0003156284255181873,
+      "loss": 1.3105,
+      "step": 38050
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 0.00031534782634540486,
+      "loss": 1.2812,
+      "step": 38060
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.0003150672944955818,
+      "loss": 1.3502,
+      "step": 38070
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.0003147868300709984,
+      "loss": 1.1581,
+      "step": 38080
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 0.0003145064331739099,
+      "loss": 1.1369,
+      "step": 38090
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 0.0003142261039065472,
+      "loss": 1.1917,
+      "step": 38100
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.0003139458423711166,
+      "loss": 1.3765,
+      "step": 38110
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.00031366564866979943,
+      "loss": 1.3137,
+      "step": 38120
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 0.00031338552290475266,
+      "loss": 1.3079,
+      "step": 38130
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 0.00031310546517810816,
+      "loss": 1.0707,
+      "step": 38140
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.00031282547559197306,
+      "loss": 1.179,
+      "step": 38150
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.0003125455542484299,
+      "loss": 1.2167,
+      "step": 38160
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.0003122657012495362,
+      "loss": 1.1553,
+      "step": 38170
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.000311985916697324,
+      "loss": 1.2933,
+      "step": 38180
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.0003117062006938015,
+      "loss": 1.0811,
+      "step": 38190
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.0003114265533409508,
+      "loss": 1.1643,
+      "step": 38200
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.00031114697474073007,
+      "loss": 1.2889,
+      "step": 38210
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.00031086746499507135,
+      "loss": 1.2492,
+      "step": 38220
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.000310588024205882,
+      "loss": 1.4177,
+      "step": 38230
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.00031030865247504456,
+      "loss": 1.3321,
+      "step": 38240
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 0.0003100293499044159,
+      "loss": 1.1153,
+      "step": 38250
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 0.00030975011659582807,
+      "loss": 1.1505,
+      "step": 38260
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.0003094709526510877,
+      "loss": 1.3138,
+      "step": 38270
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.0003091918581719758,
+      "loss": 1.0964,
+      "step": 38280
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.0003089128332602487,
+      "loss": 1.2586,
+      "step": 38290
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 0.0003086338780176369,
+      "loss": 1.3142,
+      "step": 38300
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 0.00030835499254584564,
+      "loss": 1.3334,
+      "step": 38310
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.00030807617694655486,
+      "loss": 1.1618,
+      "step": 38320
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.0003077974313214184,
+      "loss": 1.3603,
+      "step": 38330
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.00030751875577206576,
+      "loss": 1.1558,
+      "step": 38340
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.0003072401504000998,
+      "loss": 1.2848,
+      "step": 38350
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 0.0003069616153070981,
+      "loss": 1.3002,
+      "step": 38360
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 0.00030668315059461293,
+      "loss": 1.3967,
+      "step": 38370
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 0.00030640475636417054,
+      "loss": 1.1685,
+      "step": 38380
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 0.0003061264327172717,
+      "loss": 1.3942,
+      "step": 38390
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.00030584817975539146,
+      "loss": 1.1858,
+      "step": 38400
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.00030556999757997886,
+      "loss": 1.3163,
+      "step": 38410
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.0003052918862924573,
+      "loss": 1.1412,
+      "step": 38420
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.00030501384599422425,
+      "loss": 1.1051,
+      "step": 38430
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 0.0003047358767866516,
+      "loss": 1.3188,
+      "step": 38440
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 0.000304457978771085,
+      "loss": 1.2313,
+      "step": 38450
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 0.00030418015204884387,
+      "loss": 1.1833,
+      "step": 38460
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 0.00030390239672122254,
+      "loss": 1.2472,
+      "step": 38470
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.00030362471288948844,
+      "loss": 1.2204,
+      "step": 38480
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.00030334710065488317,
+      "loss": 1.3357,
+      "step": 38490
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 0.00030306956011862265,
+      "loss": 1.0947,
+      "step": 38500
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 0.0003027920913818959,
+      "loss": 1.1956,
+      "step": 38510
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 0.0003025146945458666,
+      "loss": 1.3021,
+      "step": 38520
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 0.0003022373697116716,
+      "loss": 1.2457,
+      "step": 38530
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0003019601169804216,
+      "loss": 1.2881,
+      "step": 38540
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": {
+        "accuracy": 0.5732814526588845
+      },
+      "eval_f1": {
+        "f1": 0.5822889712524011
+      },
+      "eval_loss": 1.036847710609436,
+      "eval_precision": {
+        "precision": 0.5902075134144437
+      },
+      "eval_recall": {
+        "recall": 0.5939011675173421
+      },
+      "eval_runtime": 193.0544,
+      "eval_samples_per_second": 19.968,
+      "eval_steps_per_second": 9.987,
+      "step": 38540
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.00030168293645320134,
+      "loss": 1.1887,
+      "step": 38550
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.0003014058282310688,
+      "loss": 1.2063,
+      "step": 38560
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 0.0003011287924150559,
+      "loss": 1.1669,
+      "step": 38570
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 0.00030085182910616814,
+      "loss": 1.1266,
+      "step": 38580
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.0003005749384053842,
+      "loss": 1.4623,
+      "step": 38590
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.00030029812041365713,
+      "loss": 1.1109,
+      "step": 38600
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.00030002137523191225,
+      "loss": 1.2336,
+      "step": 38610
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.00029974470296104973,
+      "loss": 1.1438,
+      "step": 38620
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 0.000299468103701942,
+      "loss": 1.168,
+      "step": 38630
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 0.00029919157755543526,
+      "loss": 1.2298,
+      "step": 38640
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.0002989151246223494,
+      "loss": 1.3831,
+      "step": 38650
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.00029863874500347725,
+      "loss": 1.3132,
+      "step": 38660
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.00029836243879958487,
+      "loss": 1.1118,
+      "step": 38670
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.00029808620611141177,
+      "loss": 1.2783,
+      "step": 38680
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.00029781004703967047,
+      "loss": 1.4042,
+      "step": 38690
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.00029753396168504686,
+      "loss": 1.0744,
+      "step": 38700
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 0.0002972579501481998,
+      "loss": 1.049,
+      "step": 38710
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 0.00029698201252976113,
+      "loss": 1.1417,
+      "step": 38720
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 0.0002967061489303362,
+      "loss": 1.3331,
+      "step": 38730
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 0.00029643035945050255,
+      "loss": 1.2294,
+      "step": 38740
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.00029615464419081183,
+      "loss": 1.1556,
+      "step": 38750
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.0002958790032517875,
+      "loss": 1.2407,
+      "step": 38760
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 0.00029560343673392644,
+      "loss": 1.2292,
+      "step": 38770
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 0.0002953279447376986,
+      "loss": 1.2538,
+      "step": 38780
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.00029505252736354646,
+      "loss": 1.2908,
+      "step": 38790
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.0002947771847118854,
+      "loss": 1.235,
+      "step": 38800
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 0.0002945019168831035,
+      "loss": 1.3215,
+      "step": 38810
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 0.00029422672397756147,
+      "loss": 1.3122,
+      "step": 38820
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 0.000293951606095593,
+      "loss": 1.1951,
+      "step": 38830
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.0002936765633375042,
+      "loss": 1.0952,
+      "step": 38840
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.0002934015958035737,
+      "loss": 1.1631,
+      "step": 38850
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.000293126703594053,
+      "loss": 1.256,
+      "step": 38860
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.00029285188680916593,
+      "loss": 1.3125,
+      "step": 38870
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 0.0002925771455491089,
+      "loss": 1.1347,
+      "step": 38880
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 0.00029230247991405086,
+      "loss": 1.0993,
+      "step": 38890
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.00029202789000413266,
+      "loss": 1.0758,
+      "step": 38900
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.00029175337591946836,
+      "loss": 1.1601,
+      "step": 38910
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.00029147893776014386,
+      "loss": 1.2035,
+      "step": 38920
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.00029120457562621763,
+      "loss": 1.1634,
+      "step": 38930
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 0.00029093028961772017,
+      "loss": 1.0486,
+      "step": 38940
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 0.000290656079834654,
+      "loss": 1.1006,
+      "step": 38950
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.0002903819463769948,
+      "loss": 1.1973,
+      "step": 38960
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.00029010788934468964,
+      "loss": 1.2473,
+      "step": 38970
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.00028983390883765756,
+      "loss": 1.3851,
+      "step": 38980
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.0002895600049557903,
+      "loss": 1.3719,
+      "step": 38990
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.00028928617779895127,
+      "loss": 1.0738,
+      "step": 39000
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.00028901242746697637,
+      "loss": 1.2905,
+      "step": 39010
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 0.00028873875405967275,
+      "loss": 1.3089,
+      "step": 39020
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 0.0002884651576768201,
+      "loss": 1.2547,
+      "step": 39030
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 0.00028819163841817003,
+      "loss": 1.3677,
+      "step": 39040
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 0.0002879181963834455,
+      "loss": 1.2863,
+      "step": 39050
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 0.00028764483167234177,
+      "loss": 1.3187,
+      "step": 39060
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00028737154438452614,
+      "loss": 1.2603,
+      "step": 39070
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 0.00028709833461963686,
+      "loss": 1.2859,
+      "step": 39080
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 0.0002868252024772847,
+      "loss": 1.2315,
+      "step": 39090
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 0.00028655214805705146,
+      "loss": 1.1796,
+      "step": 39100
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 0.00028627917145849157,
+      "loss": 1.2675,
+      "step": 39110
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 0.0002860062727811302,
+      "loss": 1.3544,
+      "step": 39120
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 0.0002857334521244642,
+      "loss": 1.1339,
+      "step": 39130
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 0.00028546070958796236,
+      "loss": 1.138,
+      "step": 39140
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 0.00028518804527106497,
+      "loss": 1.1483,
+      "step": 39150
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 0.00028491545927318324,
+      "loss": 1.2368,
+      "step": 39160
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 0.00028464295169370045,
+      "loss": 1.2943,
+      "step": 39170
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 0.00028437052263197106,
+      "loss": 1.1498,
+      "step": 39180
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 0.00028409817218732103,
+      "loss": 1.2677,
+      "step": 39190
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 0.00028382590045904714,
+      "loss": 1.0287,
+      "step": 39200
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 0.0002835537075464182,
+      "loss": 1.1375,
+      "step": 39210
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 0.000283281593548674,
+      "loss": 1.1159,
+      "step": 39220
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 0.00028300955856502515,
+      "loss": 1.2206,
+      "step": 39230
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 0.000282737602694654,
+      "loss": 1.1379,
+      "step": 39240
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 0.00028246572603671407,
+      "loss": 1.2603,
+      "step": 39250
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 0.00028219392869032936,
+      "loss": 1.3507,
+      "step": 39260
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.00028192221075459573,
+      "loss": 1.2479,
+      "step": 39270
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.00028165057232857927,
+      "loss": 1.1846,
+      "step": 39280
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.0002813790135113182,
+      "loss": 1.1452,
+      "step": 39290
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 0.0002811075344018209,
+      "loss": 1.2895,
+      "step": 39300
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 0.00028083613509906645,
+      "loss": 1.0824,
+      "step": 39310
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 0.00028056481570200547,
+      "loss": 1.1082,
+      "step": 39320
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 0.0002802935763095594,
+      "loss": 1.1545,
+      "step": 39330
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 0.00028002241702062007,
+      "loss": 1.182,
+      "step": 39340
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 0.00027975133793405044,
+      "loss": 1.2907,
+      "step": 39350
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 0.00027948033914868415,
+      "loss": 1.1596,
+      "step": 39360
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 0.0002792094207633258,
+      "loss": 1.0701,
+      "step": 39370
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 0.0002789385828767501,
+      "loss": 1.1884,
+      "step": 39380
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 0.00027866782558770266,
+      "loss": 1.1663,
+      "step": 39390
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 0.00027839714899490024,
+      "loss": 1.1378,
+      "step": 39400
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 0.0002781265531970294,
+      "loss": 1.3202,
+      "step": 39410
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.0002778560382927478,
+      "loss": 1.1055,
+      "step": 39420
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.0002775856043806831,
+      "loss": 1.1296,
+      "step": 39430
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 0.0002773152515594338,
+      "loss": 1.0761,
+      "step": 39440
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 0.00027704497992756905,
+      "loss": 1.1237,
+      "step": 39450
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 0.0002767747895836277,
+      "loss": 1.0235,
+      "step": 39460
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 0.0002765046806261195,
+      "loss": 1.2635,
+      "step": 39470
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.0002762346531535246,
+      "loss": 1.1013,
+      "step": 39480
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.0002759647072642928,
+      "loss": 1.1812,
+      "step": 39490
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.00027569484305684487,
+      "loss": 1.1605,
+      "step": 39500
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 0.0002754250606295716,
+      "loss": 1.1981,
+      "step": 39510
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 0.00027515536008083366,
+      "loss": 0.9425,
+      "step": 39520
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 0.0002748857415089623,
+      "loss": 1.1863,
+      "step": 39530
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 0.0002746162050122581,
+      "loss": 1.164,
+      "step": 39540
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 0.0002743467506889931,
+      "loss": 1.3006,
+      "step": 39550
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 0.00027407737863740817,
+      "loss": 1.289,
+      "step": 39560
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 0.0002738080889557144,
+      "loss": 1.3579,
+      "step": 39570
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 0.00027353888174209317,
+      "loss": 1.0017,
+      "step": 39580
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 0.0002732697570946957,
+      "loss": 1.3477,
+      "step": 39590
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 0.00027300071511164326,
+      "loss": 1.386,
+      "step": 39600
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 0.0002727317558910263,
+      "loss": 1.1881,
+      "step": 39610
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 0.00027246287953090585,
+      "loss": 1.0385,
+      "step": 39620
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.00027219408612931263,
+      "loss": 1.2821,
+      "step": 39630
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.00027192537578424667,
+      "loss": 1.0839,
+      "step": 39640
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 0.0002716567485936776,
+      "loss": 1.1401,
+      "step": 39650
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 0.00027138820465554605,
+      "loss": 1.2137,
+      "step": 39660
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 0.0002711197440677606,
+      "loss": 1.2702,
+      "step": 39670
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 0.0002708513669282006,
+      "loss": 1.2721,
+      "step": 39680
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 0.0002705830733347143,
+      "loss": 1.1733,
+      "step": 39690
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 0.0002703148633851199,
+      "loss": 1.3803,
+      "step": 39700
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.0002700467371772052,
+      "loss": 1.1441,
+      "step": 39710
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.0002697786948087268,
+      "loss": 1.242,
+      "step": 39720
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 0.00026951073637741135,
+      "loss": 1.3418,
+      "step": 39730
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 0.000269242861980955,
+      "loss": 1.0691,
+      "step": 39740
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 0.00026897507171702253,
+      "loss": 1.322,
+      "step": 39750
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 0.00026870736568324874,
+      "loss": 1.2821,
+      "step": 39760
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 0.0002684397439772374,
+      "loss": 1.1297,
+      "step": 39770
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 0.0002681722066965618,
+      "loss": 1.2664,
+      "step": 39780
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 0.0002679047539387639,
+      "loss": 1.0509,
+      "step": 39790
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 0.00026763738580135553,
+      "loss": 0.9593,
+      "step": 39800
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 0.00026737010238181734,
+      "loss": 1.204,
+      "step": 39810
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 0.00026710290377759906,
+      "loss": 1.1319,
+      "step": 39820
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 0.00026683579008611913,
+      "loss": 1.1717,
+      "step": 39830
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 0.0002665687614047658,
+      "loss": 1.3699,
+      "step": 39840
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 0.00026630181783089603,
+      "loss": 1.2318,
+      "step": 39850
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 0.00026603495946183567,
+      "loss": 1.1743,
+      "step": 39860
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 0.00026576818639487923,
+      "loss": 1.1741,
+      "step": 39870
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 0.0002655014987272905,
+      "loss": 1.1627,
+      "step": 39880
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 0.0002652348965563024,
+      "loss": 1.2481,
+      "step": 39890
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 0.0002649683799791158,
+      "loss": 1.0844,
+      "step": 39900
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 0.0002647019490929011,
+      "loss": 1.1144,
+      "step": 39910
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.00026443560399479747,
+      "loss": 1.154,
+      "step": 39920
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.00026416934478191225,
+      "loss": 1.2489,
+      "step": 39930
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 0.0002639031715513217,
+      "loss": 1.5209,
+      "step": 39940
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 0.00026363708440007135,
+      "loss": 1.2268,
+      "step": 39950
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 0.0002633710834251742,
+      "loss": 1.3809,
+      "step": 39960
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 0.0002631051687236129,
+      "loss": 1.217,
+      "step": 39970
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 0.0002628393403923376,
+      "loss": 1.1967,
+      "step": 39980
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 0.00026257359852826834,
+      "loss": 1.0959,
+      "step": 39990
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.00026230794322829244,
+      "loss": 0.9996,
+      "step": 40000
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.00026204237458926584,
+      "loss": 1.023,
+      "step": 40010
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 0.00026177689270801344,
+      "loss": 1.2478,
+      "step": 40020
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 0.000261511497681328,
+      "loss": 0.9404,
+      "step": 40030
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 0.00026124618960597103,
+      "loss": 1.3133,
+      "step": 40040
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 0.00026098096857867176,
+      "loss": 1.2357,
+      "step": 40050
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 0.0002607158346961282,
+      "loss": 1.1756,
+      "step": 40060
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 0.0002604507880550066,
+      "loss": 1.4344,
+      "step": 40070
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 0.00026018582875194073,
+      "loss": 1.3091,
+      "step": 40080
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 0.0002599209568835334,
+      "loss": 1.2607,
+      "step": 40090
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 0.0002596561725463551,
+      "loss": 1.1957,
+      "step": 40100
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 0.0002593914758369442,
+      "loss": 1.1149,
+      "step": 40110
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 0.00025912686685180777,
+      "loss": 1.2671,
+      "step": 40120
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.00025886234568741994,
+      "loss": 1.2207,
+      "step": 40130
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.0002585979124402238,
+      "loss": 1.1914,
+      "step": 40140
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 0.00025833356720663,
+      "loss": 1.0492,
+      "step": 40150
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 0.0002580693100830167,
+      "loss": 1.3062,
+      "step": 40160
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 0.0002578051411657305,
+      "loss": 1.1622,
+      "step": 40170
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 0.00025754106055108576,
+      "loss": 1.1898,
+      "step": 40180
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 0.0002572770683353643,
+      "loss": 1.156,
+      "step": 40190
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 0.000257013164614816,
+      "loss": 1.2677,
+      "step": 40200
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 0.0002567493494856584,
+      "loss": 1.2109,
+      "step": 40210
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 0.0002564856230440771,
+      "loss": 1.1145,
+      "step": 40220
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 0.00025622198538622455,
+      "loss": 1.2414,
+      "step": 40230
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 0.0002559584366082212,
+      "loss": 1.1956,
+      "step": 40240
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 0.0002556949768061557,
+      "loss": 1.0952,
+      "step": 40250
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 0.0002554316060760834,
+      "loss": 1.1434,
+      "step": 40260
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 0.0002551683245140278,
+      "loss": 1.0144,
+      "step": 40270
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 0.00025490513221597915,
+      "loss": 1.3246,
+      "step": 40280
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 0.00025464202927789594,
+      "loss": 1.3807,
+      "step": 40290
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 0.0002543790157957039,
+      "loss": 1.3442,
+      "step": 40300
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 0.0002541160918652956,
+      "loss": 1.0797,
+      "step": 40310
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 0.00025385325758253165,
+      "loss": 1.1703,
+      "step": 40320
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 0.0002535905130432397,
+      "loss": 1.2565,
+      "step": 40330
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 0.00025332785834321445,
+      "loss": 1.099,
+      "step": 40340
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.0002530652935782182,
+      "loss": 1.1452,
+      "step": 40350
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.00025280281884398046,
+      "loss": 1.0487,
+      "step": 40360
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 0.00025254043423619755,
+      "loss": 0.9842,
+      "step": 40370
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 0.0002522781398505334,
+      "loss": 1.2213,
+      "step": 40380
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 0.0002520159357826183,
+      "loss": 1.2949,
+      "step": 40390
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 0.00025175382212805076,
+      "loss": 1.2475,
+      "step": 40400
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 0.0002514917989823955,
+      "loss": 1.1433,
+      "step": 40410
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 0.00025122986644118405,
+      "loss": 1.1607,
+      "step": 40420
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 0.00025096802459991565,
+      "loss": 1.2003,
+      "step": 40430
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 0.0002507062735540561,
+      "loss": 1.1264,
+      "step": 40440
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 0.0002504446133990379,
+      "loss": 1.1918,
+      "step": 40450
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 0.00025018304423026064,
+      "loss": 1.1817,
+      "step": 40460
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": {
+        "accuracy": 0.562905317769131
+      },
+      "eval_f1": {
+        "f1": 0.5595449409010723
+      },
+      "eval_loss": 1.092657208442688,
+      "eval_precision": {
+        "precision": 0.6241213136239965
+      },
+      "eval_recall": {
+        "recall": 0.5901142810563812
+      },
+      "eval_runtime": 181.1563,
+      "eval_samples_per_second": 21.28,
+      "eval_steps_per_second": 10.643,
+      "step": 40467
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 57810,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 30,
+  "save_steps": 500,
+  "total_flos": 1.866769462896604e+19,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}