diff --git "a/checkpoint-11970/trainer_state.json" "b/checkpoint-11970/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-11970/trainer_state.json"
@@ -0,0 +1,8562 @@
+{
+  "best_metric": 0.9356,
+  "best_model_checkpoint": "checkpoint/vit-large/checkpoint-11970",
+  "epoch": 18.0,
+  "eval_steps": 500,
+  "global_step": 11970,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "grad_norm": 2.7020020484924316,
+      "learning_rate": 9.998496240601504e-06,
+      "loss": 4.6119,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 2.892003059387207,
+      "learning_rate": 9.996992481203008e-06,
+      "loss": 4.5669,
+      "step": 20
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 2.7313311100006104,
+      "learning_rate": 9.995488721804511e-06,
+      "loss": 4.5263,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 2.659421443939209,
+      "learning_rate": 9.993984962406017e-06,
+      "loss": 4.4729,
+      "step": 40
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 2.846480131149292,
+      "learning_rate": 9.992481203007518e-06,
+      "loss": 4.3985,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 3.0880706310272217,
+      "learning_rate": 9.990977443609024e-06,
+      "loss": 4.3247,
+      "step": 60
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 3.202064037322998,
+      "learning_rate": 9.989473684210527e-06,
+      "loss": 4.233,
+      "step": 70
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 4.243076324462891,
+      "learning_rate": 9.98796992481203e-06,
+      "loss": 4.1598,
+      "step": 80
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 3.47965669631958,
+      "learning_rate": 9.986466165413534e-06,
+      "loss": 4.0929,
+      "step": 90
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 3.431342363357544,
+      "learning_rate": 9.984962406015038e-06,
+      "loss": 3.9735,
+      "step": 100
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 3.915376901626587,
+      "learning_rate": 9.983458646616541e-06,
+      "loss": 3.9051,
+      "step": 110
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 3.6235947608947754,
+      "learning_rate": 9.981954887218046e-06,
+      "loss": 3.7795,
+      "step": 120
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 4.0705485343933105,
+      "learning_rate": 9.98045112781955e-06,
+      "loss": 3.6517,
+      "step": 130
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 3.916447162628174,
+      "learning_rate": 9.978947368421053e-06,
+      "loss": 3.5197,
+      "step": 140
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 4.508981227874756,
+      "learning_rate": 9.977443609022557e-06,
+      "loss": 3.4313,
+      "step": 150
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 4.508138179779053,
+      "learning_rate": 9.97593984962406e-06,
+      "loss": 3.3823,
+      "step": 160
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 5.037680625915527,
+      "learning_rate": 9.974436090225564e-06,
+      "loss": 3.2748,
+      "step": 170
+    },
+    {
+      "epoch": 0.27,
+      "grad_norm": 10.304414749145508,
+      "learning_rate": 9.97293233082707e-06,
+      "loss": 3.1054,
+      "step": 180
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 5.1311421394348145,
+      "learning_rate": 9.971428571428571e-06,
+      "loss": 3.0703,
+      "step": 190
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 4.8540120124816895,
+      "learning_rate": 9.969924812030076e-06,
+      "loss": 2.9789,
+      "step": 200
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 4.647185802459717,
+      "learning_rate": 9.96842105263158e-06,
+      "loss": 2.8568,
+      "step": 210
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 6.943347454071045,
+      "learning_rate": 9.966917293233083e-06,
+      "loss": 2.8175,
+      "step": 220
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 17.15534210205078,
+      "learning_rate": 9.965413533834587e-06,
+      "loss": 2.7049,
+      "step": 230
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 5.327253341674805,
+      "learning_rate": 9.963909774436092e-06,
+      "loss": 2.6185,
+      "step": 240
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 5.56367301940918,
+      "learning_rate": 9.962406015037594e-06,
+      "loss": 2.4704,
+      "step": 250
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 6.038745880126953,
+      "learning_rate": 9.960902255639099e-06,
+      "loss": 2.4862,
+      "step": 260
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 5.661726474761963,
+      "learning_rate": 9.959398496240603e-06,
+      "loss": 2.4405,
+      "step": 270
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 5.565981864929199,
+      "learning_rate": 9.957894736842106e-06,
+      "loss": 2.3698,
+      "step": 280
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 7.349733829498291,
+      "learning_rate": 9.95639097744361e-06,
+      "loss": 2.3338,
+      "step": 290
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 6.526618957519531,
+      "learning_rate": 9.954887218045113e-06,
+      "loss": 2.2445,
+      "step": 300
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 5.558746337890625,
+      "learning_rate": 9.953383458646617e-06,
+      "loss": 2.202,
+      "step": 310
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 6.4157633781433105,
+      "learning_rate": 9.951879699248122e-06,
+      "loss": 2.1578,
+      "step": 320
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 5.635522842407227,
+      "learning_rate": 9.950375939849625e-06,
+      "loss": 2.099,
+      "step": 330
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 8.216004371643066,
+      "learning_rate": 9.948872180451129e-06,
+      "loss": 2.0472,
+      "step": 340
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 7.348927021026611,
+      "learning_rate": 9.947368421052632e-06,
+      "loss": 2.0272,
+      "step": 350
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 6.218992710113525,
+      "learning_rate": 9.945864661654136e-06,
+      "loss": 1.9022,
+      "step": 360
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 12.379638671875,
+      "learning_rate": 9.94436090225564e-06,
+      "loss": 1.9707,
+      "step": 370
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 7.454248905181885,
+      "learning_rate": 9.942857142857145e-06,
+      "loss": 1.9612,
+      "step": 380
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 8.975961685180664,
+      "learning_rate": 9.941353383458647e-06,
+      "loss": 1.8519,
+      "step": 390
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 11.839798927307129,
+      "learning_rate": 9.939849624060152e-06,
+      "loss": 1.8246,
+      "step": 400
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 21.542709350585938,
+      "learning_rate": 9.938345864661655e-06,
+      "loss": 1.7591,
+      "step": 410
+    },
+    {
+      "epoch": 0.63,
+      "grad_norm": 6.640402793884277,
+      "learning_rate": 9.936842105263159e-06,
+      "loss": 1.7803,
+      "step": 420
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 7.312070369720459,
+      "learning_rate": 9.935338345864662e-06,
+      "loss": 1.6758,
+      "step": 430
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 6.633362770080566,
+      "learning_rate": 9.933834586466168e-06,
+      "loss": 1.7095,
+      "step": 440
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 9.548731803894043,
+      "learning_rate": 9.93233082706767e-06,
+      "loss": 1.6934,
+      "step": 450
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 9.280405044555664,
+      "learning_rate": 9.930827067669175e-06,
+      "loss": 1.74,
+      "step": 460
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 4.683008670806885,
+      "learning_rate": 9.929323308270678e-06,
+      "loss": 1.5803,
+      "step": 470
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 5.887816905975342,
+      "learning_rate": 9.927819548872182e-06,
+      "loss": 1.6039,
+      "step": 480
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 7.674352645874023,
+      "learning_rate": 9.926315789473685e-06,
+      "loss": 1.6009,
+      "step": 490
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 5.968412399291992,
+      "learning_rate": 9.924812030075189e-06,
+      "loss": 1.5391,
+      "step": 500
+    },
+    {
+      "epoch": 0.77,
+      "grad_norm": 5.869007587432861,
+      "learning_rate": 9.923308270676692e-06,
+      "loss": 1.5226,
+      "step": 510
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 7.70728874206543,
+      "learning_rate": 9.921804511278196e-06,
+      "loss": 1.6173,
+      "step": 520
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 7.701499938964844,
+      "learning_rate": 9.920300751879701e-06,
+      "loss": 1.5042,
+      "step": 530
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 6.880636692047119,
+      "learning_rate": 9.918796992481203e-06,
+      "loss": 1.456,
+      "step": 540
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 7.021149635314941,
+      "learning_rate": 9.917293233082708e-06,
+      "loss": 1.5698,
+      "step": 550
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 7.1271138191223145,
+      "learning_rate": 9.915789473684211e-06,
+      "loss": 1.3838,
+      "step": 560
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 7.8424482345581055,
+      "learning_rate": 9.914285714285715e-06,
+      "loss": 1.3997,
+      "step": 570
+    },
+    {
+      "epoch": 0.87,
+      "grad_norm": 8.099345207214355,
+      "learning_rate": 9.912781954887218e-06,
+      "loss": 1.4359,
+      "step": 580
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 7.261110782623291,
+      "learning_rate": 9.911278195488722e-06,
+      "loss": 1.454,
+      "step": 590
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 6.30597448348999,
+      "learning_rate": 9.909774436090226e-06,
+      "loss": 1.3907,
+      "step": 600
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 7.466990947723389,
+      "learning_rate": 9.90827067669173e-06,
+      "loss": 1.329,
+      "step": 610
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 7.6469316482543945,
+      "learning_rate": 9.906766917293234e-06,
+      "loss": 1.2848,
+      "step": 620
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 7.2480244636535645,
+      "learning_rate": 9.905263157894738e-06,
+      "loss": 1.4246,
+      "step": 630
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 7.738135814666748,
+      "learning_rate": 9.903759398496241e-06,
+      "loss": 1.3669,
+      "step": 640
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 8.298168182373047,
+      "learning_rate": 9.902255639097745e-06,
+      "loss": 1.2883,
+      "step": 650
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 6.747694969177246,
+      "learning_rate": 9.900751879699248e-06,
+      "loss": 1.2884,
+      "step": 660
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8834,
+      "eval_loss": 0.8751662373542786,
+      "eval_runtime": 85.9745,
+      "eval_samples_per_second": 116.314,
+      "eval_steps_per_second": 0.465,
+      "step": 665
+    },
+    {
+      "epoch": 1.01,
+      "grad_norm": 7.908563137054443,
+      "learning_rate": 9.899248120300754e-06,
+      "loss": 1.2429,
+      "step": 670
+    },
+    {
+      "epoch": 1.02,
+      "grad_norm": 7.439415454864502,
+      "learning_rate": 9.897744360902255e-06,
+      "loss": 1.2716,
+      "step": 680
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 7.2306599617004395,
+      "learning_rate": 9.89624060150376e-06,
+      "loss": 1.1676,
+      "step": 690
+    },
+    {
+      "epoch": 1.05,
+      "grad_norm": 6.287716388702393,
+      "learning_rate": 9.894736842105264e-06,
+      "loss": 1.2212,
+      "step": 700
+    },
+    {
+      "epoch": 1.07,
+      "grad_norm": 10.363208770751953,
+      "learning_rate": 9.893233082706768e-06,
+      "loss": 1.259,
+      "step": 710
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 6.556034564971924,
+      "learning_rate": 9.891729323308271e-06,
+      "loss": 1.1383,
+      "step": 720
+    },
+    {
+      "epoch": 1.1,
+      "grad_norm": 5.640949726104736,
+      "learning_rate": 9.890225563909776e-06,
+      "loss": 1.2454,
+      "step": 730
+    },
+    {
+      "epoch": 1.11,
+      "grad_norm": 7.140594959259033,
+      "learning_rate": 9.888721804511278e-06,
+      "loss": 1.1363,
+      "step": 740
+    },
+    {
+      "epoch": 1.13,
+      "grad_norm": 7.780942440032959,
+      "learning_rate": 9.887218045112783e-06,
+      "loss": 1.1806,
+      "step": 750
+    },
+    {
+      "epoch": 1.14,
+      "grad_norm": 6.301698684692383,
+      "learning_rate": 9.885714285714287e-06,
+      "loss": 1.1416,
+      "step": 760
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 7.256317138671875,
+      "learning_rate": 9.88421052631579e-06,
+      "loss": 1.1947,
+      "step": 770
+    },
+    {
+      "epoch": 1.17,
+      "grad_norm": 9.818164825439453,
+      "learning_rate": 9.882706766917294e-06,
+      "loss": 1.1865,
+      "step": 780
+    },
+    {
+      "epoch": 1.19,
+      "grad_norm": 11.159587860107422,
+      "learning_rate": 9.881203007518797e-06,
+      "loss": 1.0911,
+      "step": 790
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 12.432153701782227,
+      "learning_rate": 9.879699248120301e-06,
+      "loss": 1.1486,
+      "step": 800
+    },
+    {
+      "epoch": 1.22,
+      "grad_norm": 9.031283378601074,
+      "learning_rate": 9.878195488721806e-06,
+      "loss": 1.0838,
+      "step": 810
+    },
+    {
+      "epoch": 1.23,
+      "grad_norm": 7.850508689880371,
+      "learning_rate": 9.87669172932331e-06,
+      "loss": 1.0953,
+      "step": 820
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 6.112914562225342,
+      "learning_rate": 9.875187969924813e-06,
+      "loss": 1.1159,
+      "step": 830
+    },
+    {
+      "epoch": 1.26,
+      "grad_norm": 6.945899486541748,
+      "learning_rate": 9.873684210526317e-06,
+      "loss": 0.9613,
+      "step": 840
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 7.430254936218262,
+      "learning_rate": 9.87218045112782e-06,
+      "loss": 1.086,
+      "step": 850
+    },
+    {
+      "epoch": 1.29,
+      "grad_norm": 6.00394868850708,
+      "learning_rate": 9.870676691729324e-06,
+      "loss": 1.0873,
+      "step": 860
+    },
+    {
+      "epoch": 1.31,
+      "grad_norm": 5.864083290100098,
+      "learning_rate": 9.869172932330829e-06,
+      "loss": 1.055,
+      "step": 870
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 8.37553882598877,
+      "learning_rate": 9.86766917293233e-06,
+      "loss": 1.0894,
+      "step": 880
+    },
+    {
+      "epoch": 1.34,
+      "grad_norm": 8.302824974060059,
+      "learning_rate": 9.866165413533836e-06,
+      "loss": 1.1037,
+      "step": 890
+    },
+    {
+      "epoch": 1.35,
+      "grad_norm": 6.968749046325684,
+      "learning_rate": 9.86466165413534e-06,
+      "loss": 1.0071,
+      "step": 900
+    },
+    {
+      "epoch": 1.37,
+      "grad_norm": 5.924696922302246,
+      "learning_rate": 9.863157894736843e-06,
+      "loss": 1.0498,
+      "step": 910
+    },
+    {
+      "epoch": 1.38,
+      "grad_norm": 7.433680534362793,
+      "learning_rate": 9.861654135338347e-06,
+      "loss": 1.0456,
+      "step": 920
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 7.614802360534668,
+      "learning_rate": 9.86015037593985e-06,
+      "loss": 0.9659,
+      "step": 930
+    },
+    {
+      "epoch": 1.41,
+      "grad_norm": 9.205560684204102,
+      "learning_rate": 9.858646616541354e-06,
+      "loss": 0.9692,
+      "step": 940
+    },
+    {
+      "epoch": 1.43,
+      "grad_norm": 6.741930961608887,
+      "learning_rate": 9.857142857142859e-06,
+      "loss": 1.0157,
+      "step": 950
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 8.176901817321777,
+      "learning_rate": 9.855639097744362e-06,
+      "loss": 1.0014,
+      "step": 960
+    },
+    {
+      "epoch": 1.46,
+      "grad_norm": 5.335792541503906,
+      "learning_rate": 9.854135338345866e-06,
+      "loss": 0.9361,
+      "step": 970
+    },
+    {
+      "epoch": 1.47,
+      "grad_norm": 6.488396644592285,
+      "learning_rate": 9.85263157894737e-06,
+      "loss": 1.0858,
+      "step": 980
+    },
+    {
+      "epoch": 1.49,
+      "grad_norm": 6.725528240203857,
+      "learning_rate": 9.851127819548873e-06,
+      "loss": 1.0027,
+      "step": 990
+    },
+    {
+      "epoch": 1.5,
+      "grad_norm": 10.740559577941895,
+      "learning_rate": 9.849624060150376e-06,
+      "loss": 1.0121,
+      "step": 1000
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 8.048620223999023,
+      "learning_rate": 9.84812030075188e-06,
+      "loss": 0.967,
+      "step": 1010
+    },
+    {
+      "epoch": 1.53,
+      "grad_norm": 7.225861072540283,
+      "learning_rate": 9.846616541353383e-06,
+      "loss": 1.0329,
+      "step": 1020
+    },
+    {
+      "epoch": 1.55,
+      "grad_norm": 7.12366247177124,
+      "learning_rate": 9.845112781954887e-06,
+      "loss": 1.0117,
+      "step": 1030
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 7.5486273765563965,
+      "learning_rate": 9.843609022556392e-06,
+      "loss": 0.969,
+      "step": 1040
+    },
+    {
+      "epoch": 1.58,
+      "grad_norm": 9.591785430908203,
+      "learning_rate": 9.842105263157896e-06,
+      "loss": 0.9415,
+      "step": 1050
+    },
+    {
+      "epoch": 1.59,
+      "grad_norm": 7.203570365905762,
+      "learning_rate": 9.8406015037594e-06,
+      "loss": 1.0409,
+      "step": 1060
+    },
+    {
+      "epoch": 1.61,
+      "grad_norm": 9.412242889404297,
+      "learning_rate": 9.839097744360903e-06,
+      "loss": 1.0352,
+      "step": 1070
+    },
+    {
+      "epoch": 1.62,
+      "grad_norm": 9.688934326171875,
+      "learning_rate": 9.837593984962406e-06,
+      "loss": 1.0524,
+      "step": 1080
+    },
+    {
+      "epoch": 1.64,
+      "grad_norm": 6.9523844718933105,
+      "learning_rate": 9.83609022556391e-06,
+      "loss": 0.9546,
+      "step": 1090
+    },
+    {
+      "epoch": 1.65,
+      "grad_norm": 8.400866508483887,
+      "learning_rate": 9.834586466165415e-06,
+      "loss": 0.9361,
+      "step": 1100
+    },
+    {
+      "epoch": 1.67,
+      "grad_norm": 8.09070110321045,
+      "learning_rate": 9.833082706766917e-06,
+      "loss": 0.9797,
+      "step": 1110
+    },
+    {
+      "epoch": 1.68,
+      "grad_norm": 6.927423000335693,
+      "learning_rate": 9.831578947368422e-06,
+      "loss": 0.9439,
+      "step": 1120
+    },
+    {
+      "epoch": 1.7,
+      "grad_norm": 7.291294574737549,
+      "learning_rate": 9.830075187969926e-06,
+      "loss": 0.9753,
+      "step": 1130
+    },
+    {
+      "epoch": 1.71,
+      "grad_norm": 7.919008731842041,
+      "learning_rate": 9.828571428571429e-06,
+      "loss": 0.9697,
+      "step": 1140
+    },
+    {
+      "epoch": 1.73,
+      "grad_norm": 8.660476684570312,
+      "learning_rate": 9.827067669172933e-06,
+      "loss": 0.8882,
+      "step": 1150
+    },
+    {
+      "epoch": 1.74,
+      "grad_norm": 8.102679252624512,
+      "learning_rate": 9.825563909774438e-06,
+      "loss": 0.8959,
+      "step": 1160
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 5.902896404266357,
+      "learning_rate": 9.82406015037594e-06,
+      "loss": 0.887,
+      "step": 1170
+    },
+    {
+      "epoch": 1.77,
+      "grad_norm": 6.6904778480529785,
+      "learning_rate": 9.822556390977445e-06,
+      "loss": 0.9352,
+      "step": 1180
+    },
+    {
+      "epoch": 1.79,
+      "grad_norm": 6.770270824432373,
+      "learning_rate": 9.821052631578948e-06,
+      "loss": 0.8916,
+      "step": 1190
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 8.353099822998047,
+      "learning_rate": 9.819548872180452e-06,
+      "loss": 0.9523,
+      "step": 1200
+    },
+    {
+      "epoch": 1.82,
+      "grad_norm": 6.385773658752441,
+      "learning_rate": 9.818045112781955e-06,
+      "loss": 0.9175,
+      "step": 1210
+    },
+    {
+      "epoch": 1.83,
+      "grad_norm": 13.28996467590332,
+      "learning_rate": 9.816541353383459e-06,
+      "loss": 0.9375,
+      "step": 1220
+    },
+    {
+      "epoch": 1.85,
+      "grad_norm": 9.252169609069824,
+      "learning_rate": 9.815037593984962e-06,
+      "loss": 1.0589,
+      "step": 1230
+    },
+    {
+      "epoch": 1.86,
+      "grad_norm": 6.009567737579346,
+      "learning_rate": 9.813533834586468e-06,
+      "loss": 0.9608,
+      "step": 1240
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 4.863635063171387,
+      "learning_rate": 9.812030075187971e-06,
+      "loss": 0.9724,
+      "step": 1250
+    },
+    {
+      "epoch": 1.89,
+      "grad_norm": 10.548372268676758,
+      "learning_rate": 9.810526315789475e-06,
+      "loss": 0.9651,
+      "step": 1260
+    },
+    {
+      "epoch": 1.91,
+      "grad_norm": 8.277862548828125,
+      "learning_rate": 9.809022556390978e-06,
+      "loss": 0.8868,
+      "step": 1270
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 6.657036304473877,
+      "learning_rate": 9.807518796992482e-06,
+      "loss": 0.9133,
+      "step": 1280
+    },
+    {
+      "epoch": 1.94,
+      "grad_norm": 7.065949440002441,
+      "learning_rate": 9.806015037593985e-06,
+      "loss": 0.9602,
+      "step": 1290
+    },
+    {
+      "epoch": 1.95,
+      "grad_norm": 9.187036514282227,
+      "learning_rate": 9.80451127819549e-06,
+      "loss": 0.9626,
+      "step": 1300
+    },
+    {
+      "epoch": 1.97,
+      "grad_norm": 7.014963150024414,
+      "learning_rate": 9.803007518796992e-06,
+      "loss": 0.8653,
+      "step": 1310
+    },
+    {
+      "epoch": 1.98,
+      "grad_norm": 8.665754318237305,
+      "learning_rate": 9.801503759398498e-06,
+      "loss": 0.8795,
+      "step": 1320
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 8.92686939239502,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 0.7958,
+      "step": 1330
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9142,
+      "eval_loss": 0.4723776876926422,
+      "eval_runtime": 84.6568,
+      "eval_samples_per_second": 118.124,
+      "eval_steps_per_second": 0.472,
+      "step": 1330
+    },
+    {
+      "epoch": 2.02,
+      "grad_norm": 6.881080627441406,
+      "learning_rate": 9.798496240601505e-06,
+      "loss": 0.9416,
+      "step": 1340
+    },
+    {
+      "epoch": 2.03,
+      "grad_norm": 5.449582099914551,
+      "learning_rate": 9.796992481203008e-06,
+      "loss": 0.8897,
+      "step": 1350
+    },
+    {
+      "epoch": 2.05,
+      "grad_norm": 6.535237789154053,
+      "learning_rate": 9.795488721804513e-06,
+      "loss": 0.8241,
+      "step": 1360
+    },
+    {
+      "epoch": 2.06,
+      "grad_norm": 7.858671188354492,
+      "learning_rate": 9.793984962406015e-06,
+      "loss": 0.7834,
+      "step": 1370
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 6.7106852531433105,
+      "learning_rate": 9.79248120300752e-06,
+      "loss": 0.836,
+      "step": 1380
+    },
+    {
+      "epoch": 2.09,
+      "grad_norm": 6.440729141235352,
+      "learning_rate": 9.790977443609024e-06,
+      "loss": 0.775,
+      "step": 1390
+    },
+    {
+      "epoch": 2.11,
+      "grad_norm": 8.935519218444824,
+      "learning_rate": 9.789473684210527e-06,
+      "loss": 0.7826,
+      "step": 1400
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 7.244870662689209,
+      "learning_rate": 9.787969924812031e-06,
+      "loss": 0.8917,
+      "step": 1410
+    },
+    {
+      "epoch": 2.14,
+      "grad_norm": 9.125387191772461,
+      "learning_rate": 9.786466165413534e-06,
+      "loss": 0.8167,
+      "step": 1420
+    },
+    {
+      "epoch": 2.15,
+      "grad_norm": 6.8634114265441895,
+      "learning_rate": 9.784962406015038e-06,
+      "loss": 0.8052,
+      "step": 1430
+    },
+    {
+      "epoch": 2.17,
+      "grad_norm": 7.519056797027588,
+      "learning_rate": 9.783458646616543e-06,
+      "loss": 0.8204,
+      "step": 1440
+    },
+    {
+      "epoch": 2.18,
+      "grad_norm": 7.861953258514404,
+      "learning_rate": 9.781954887218047e-06,
+      "loss": 0.8404,
+      "step": 1450
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 10.940001487731934,
+      "learning_rate": 9.78045112781955e-06,
+      "loss": 0.7934,
+      "step": 1460
+    },
+    {
+      "epoch": 2.21,
+      "grad_norm": 5.963690757751465,
+      "learning_rate": 9.778947368421054e-06,
+      "loss": 0.7549,
+      "step": 1470
+    },
+    {
+      "epoch": 2.23,
+      "grad_norm": 6.734865665435791,
+      "learning_rate": 9.777443609022557e-06,
+      "loss": 0.8112,
+      "step": 1480
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 7.971401214599609,
+      "learning_rate": 9.77593984962406e-06,
+      "loss": 0.8323,
+      "step": 1490
+    },
+    {
+      "epoch": 2.26,
+      "grad_norm": 9.728713989257812,
+      "learning_rate": 9.774436090225564e-06,
+      "loss": 0.7441,
+      "step": 1500
+    },
+    {
+      "epoch": 2.27,
+      "grad_norm": 9.506553649902344,
+      "learning_rate": 9.772932330827068e-06,
+      "loss": 0.8018,
+      "step": 1510
+    },
+    {
+      "epoch": 2.29,
+      "grad_norm": 7.3224663734436035,
+      "learning_rate": 9.771428571428571e-06,
+      "loss": 0.7869,
+      "step": 1520
+    },
+    {
+      "epoch": 2.3,
+      "grad_norm": 7.251104831695557,
+      "learning_rate": 9.769924812030077e-06,
+      "loss": 0.8472,
+      "step": 1530
+    },
+    {
+      "epoch": 2.32,
+      "grad_norm": 7.3015055656433105,
+      "learning_rate": 9.76842105263158e-06,
+      "loss": 0.7224,
+      "step": 1540
+    },
+    {
+      "epoch": 2.33,
+      "grad_norm": 9.096901893615723,
+      "learning_rate": 9.766917293233084e-06,
+      "loss": 0.8195,
+      "step": 1550
+    },
+    {
+      "epoch": 2.35,
+      "grad_norm": 7.366261959075928,
+      "learning_rate": 9.765413533834587e-06,
+      "loss": 0.8322,
+      "step": 1560
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 9.955854415893555,
+      "learning_rate": 9.76390977443609e-06,
+      "loss": 0.8315,
+      "step": 1570
+    },
+    {
+      "epoch": 2.38,
+      "grad_norm": 6.852784633636475,
+      "learning_rate": 9.762406015037594e-06,
+      "loss": 0.7113,
+      "step": 1580
+    },
+    {
+      "epoch": 2.39,
+      "grad_norm": 9.9766263961792,
+      "learning_rate": 9.7609022556391e-06,
+      "loss": 0.8024,
+      "step": 1590
+    },
+    {
+      "epoch": 2.41,
+      "grad_norm": 9.095175743103027,
+      "learning_rate": 9.759398496240601e-06,
+      "loss": 0.8774,
+      "step": 1600
+    },
+    {
+      "epoch": 2.42,
+      "grad_norm": 5.913175106048584,
+      "learning_rate": 9.757894736842106e-06,
+      "loss": 0.7608,
+      "step": 1610
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 9.544361114501953,
+      "learning_rate": 9.75639097744361e-06,
+      "loss": 0.7984,
+      "step": 1620
+    },
+    {
+      "epoch": 2.45,
+      "grad_norm": 6.991225242614746,
+      "learning_rate": 9.754887218045113e-06,
+      "loss": 0.7942,
+      "step": 1630
+    },
+    {
+      "epoch": 2.47,
+      "grad_norm": 7.531531810760498,
+      "learning_rate": 9.753383458646617e-06,
+      "loss": 0.8005,
+      "step": 1640
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 5.945763111114502,
+      "learning_rate": 9.751879699248122e-06,
+      "loss": 0.7673,
+      "step": 1650
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 8.382121086120605,
+      "learning_rate": 9.750375939849624e-06,
+      "loss": 0.7966,
+      "step": 1660
+    },
+    {
+      "epoch": 2.51,
+      "grad_norm": 5.387685775756836,
+      "learning_rate": 9.74887218045113e-06,
+      "loss": 0.7892,
+      "step": 1670
+    },
+    {
+      "epoch": 2.53,
+      "grad_norm": 7.867427349090576,
+      "learning_rate": 9.747368421052633e-06,
+      "loss": 0.8002,
+      "step": 1680
+    },
+    {
+      "epoch": 2.54,
+      "grad_norm": 7.549880027770996,
+      "learning_rate": 9.745864661654136e-06,
+      "loss": 0.8879,
+      "step": 1690
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 7.67978572845459,
+      "learning_rate": 9.74436090225564e-06,
+      "loss": 0.7849,
+      "step": 1700
+    },
+    {
+      "epoch": 2.57,
+      "grad_norm": 8.076873779296875,
+      "learning_rate": 9.742857142857143e-06,
+      "loss": 0.6503,
+      "step": 1710
+    },
+    {
+      "epoch": 2.59,
+      "grad_norm": 10.748533248901367,
+      "learning_rate": 9.741353383458647e-06,
+      "loss": 0.7868,
+      "step": 1720
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 8.428750991821289,
+      "learning_rate": 9.739849624060152e-06,
+      "loss": 0.8195,
+      "step": 1730
+    },
+    {
+      "epoch": 2.62,
+      "grad_norm": 7.678562164306641,
+      "learning_rate": 9.738345864661655e-06,
+      "loss": 0.8428,
+      "step": 1740
+    },
+    {
+      "epoch": 2.63,
+      "grad_norm": 7.171645164489746,
+      "learning_rate": 9.736842105263159e-06,
+      "loss": 0.809,
+      "step": 1750
+    },
+    {
+      "epoch": 2.65,
+      "grad_norm": 7.041049003601074,
+      "learning_rate": 9.735338345864663e-06,
+      "loss": 0.7417,
+      "step": 1760
+    },
+    {
+      "epoch": 2.66,
+      "grad_norm": 9.66743278503418,
+      "learning_rate": 9.733834586466166e-06,
+      "loss": 0.7952,
+      "step": 1770
+    },
+    {
+      "epoch": 2.68,
+      "grad_norm": 9.864920616149902,
+      "learning_rate": 9.73233082706767e-06,
+      "loss": 0.8427,
+      "step": 1780
+    },
+    {
+      "epoch": 2.69,
+      "grad_norm": 10.242929458618164,
+      "learning_rate": 9.730827067669175e-06,
+      "loss": 0.7599,
+      "step": 1790
+    },
+    {
+      "epoch": 2.71,
+      "grad_norm": 8.138999938964844,
+      "learning_rate": 9.729323308270677e-06,
+      "loss": 0.7517,
+      "step": 1800
+    },
+    {
+      "epoch": 2.72,
+      "grad_norm": 7.668764114379883,
+      "learning_rate": 9.727819548872182e-06,
+      "loss": 0.7766,
+      "step": 1810
+    },
+    {
+      "epoch": 2.74,
+      "grad_norm": 6.978646278381348,
+      "learning_rate": 9.726315789473685e-06,
+      "loss": 0.7323,
+      "step": 1820
+    },
+    {
+      "epoch": 2.75,
+      "grad_norm": 8.794787406921387,
+      "learning_rate": 9.724812030075189e-06,
+      "loss": 0.8644,
+      "step": 1830
+    },
+    {
+      "epoch": 2.77,
+      "grad_norm": 10.154306411743164,
+      "learning_rate": 9.723308270676692e-06,
+      "loss": 0.8235,
+      "step": 1840
+    },
+    {
+      "epoch": 2.78,
+      "grad_norm": 9.513362884521484,
+      "learning_rate": 9.721804511278196e-06,
+      "loss": 0.792,
+      "step": 1850
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 4.891651630401611,
+      "learning_rate": 9.7203007518797e-06,
+      "loss": 0.7343,
+      "step": 1860
+    },
+    {
+      "epoch": 2.81,
+      "grad_norm": 6.595260143280029,
+      "learning_rate": 9.718796992481205e-06,
+      "loss": 0.6451,
+      "step": 1870
+    },
+    {
+      "epoch": 2.83,
+      "grad_norm": 11.115670204162598,
+      "learning_rate": 9.717293233082708e-06,
+      "loss": 0.7841,
+      "step": 1880
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 7.82785701751709,
+      "learning_rate": 9.715789473684212e-06,
+      "loss": 0.8396,
+      "step": 1890
+    },
+    {
+      "epoch": 2.86,
+      "grad_norm": 5.41937780380249,
+      "learning_rate": 9.714285714285715e-06,
+      "loss": 0.772,
+      "step": 1900
+    },
+    {
+      "epoch": 2.87,
+      "grad_norm": 8.092954635620117,
+      "learning_rate": 9.712781954887219e-06,
+      "loss": 0.7124,
+      "step": 1910
+    },
+    {
+      "epoch": 2.89,
+      "grad_norm": 4.913546562194824,
+      "learning_rate": 9.711278195488722e-06,
+      "loss": 0.7824,
+      "step": 1920
+    },
+    {
+      "epoch": 2.9,
+      "grad_norm": 6.090660572052002,
+      "learning_rate": 9.709774436090227e-06,
+      "loss": 0.7911,
+      "step": 1930
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 5.547027111053467,
+      "learning_rate": 9.70827067669173e-06,
+      "loss": 0.7818,
+      "step": 1940
+    },
+    {
+      "epoch": 2.93,
+      "grad_norm": 8.583475112915039,
+      "learning_rate": 9.706766917293234e-06,
+      "loss": 0.7272,
+      "step": 1950
+    },
+    {
+      "epoch": 2.95,
+      "grad_norm": 8.129578590393066,
+      "learning_rate": 9.705263157894738e-06,
+      "loss": 0.7352,
+      "step": 1960
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 9.513014793395996,
+      "learning_rate": 9.703759398496242e-06,
+      "loss": 0.8126,
+      "step": 1970
+    },
+    {
+      "epoch": 2.98,
+      "grad_norm": 5.819597244262695,
+      "learning_rate": 9.702255639097745e-06,
+      "loss": 0.7599,
+      "step": 1980
+    },
+    {
+      "epoch": 2.99,
+      "grad_norm": 7.391184329986572,
+      "learning_rate": 9.700751879699249e-06,
+      "loss": 0.743,
+      "step": 1990
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9207,
+      "eval_loss": 0.3750178813934326,
+      "eval_runtime": 84.5377,
+      "eval_samples_per_second": 118.29,
+      "eval_steps_per_second": 0.473,
+      "step": 1995
+    },
+    {
+      "epoch": 3.01,
+      "grad_norm": 6.347775459289551,
+      "learning_rate": 9.699248120300752e-06,
+      "loss": 0.6685,
+      "step": 2000
+    },
+    {
+      "epoch": 3.02,
+      "grad_norm": 6.109332084655762,
+      "learning_rate": 9.697744360902256e-06,
+      "loss": 0.7398,
+      "step": 2010
+    },
+    {
+      "epoch": 3.04,
+      "grad_norm": 4.770040512084961,
+      "learning_rate": 9.69624060150376e-06,
+      "loss": 0.7587,
+      "step": 2020
+    },
+    {
+      "epoch": 3.05,
+      "grad_norm": 7.322962760925293,
+      "learning_rate": 9.694736842105263e-06,
+      "loss": 0.7917,
+      "step": 2030
+    },
+    {
+      "epoch": 3.07,
+      "grad_norm": 9.097600936889648,
+      "learning_rate": 9.693233082706768e-06,
+      "loss": 0.6967,
+      "step": 2040
+    },
+    {
+      "epoch": 3.08,
+      "grad_norm": 10.507075309753418,
+      "learning_rate": 9.691729323308271e-06,
+      "loss": 0.7229,
+      "step": 2050
+    },
+    {
+      "epoch": 3.1,
+      "grad_norm": 6.249164581298828,
+      "learning_rate": 9.690225563909775e-06,
+      "loss": 0.6491,
+      "step": 2060
+    },
+    {
+      "epoch": 3.11,
+      "grad_norm": 7.525278568267822,
+      "learning_rate": 9.688721804511278e-06,
+      "loss": 0.6476,
+      "step": 2070
+    },
+    {
+      "epoch": 3.13,
+      "grad_norm": 9.564391136169434,
+      "learning_rate": 9.687218045112784e-06,
+      "loss": 0.7133,
+      "step": 2080
+    },
+    {
+      "epoch": 3.14,
+      "grad_norm": 6.4053955078125,
+      "learning_rate": 9.685714285714285e-06,
+      "loss": 0.8012,
+      "step": 2090
+    },
+    {
+      "epoch": 3.16,
+      "grad_norm": 6.6741251945495605,
+      "learning_rate": 9.68421052631579e-06,
+      "loss": 0.7507,
+      "step": 2100
+    },
+    {
+      "epoch": 3.17,
+      "grad_norm": 12.398811340332031,
+      "learning_rate": 9.682706766917294e-06,
+      "loss": 0.7138,
+      "step": 2110
+    },
+    {
+      "epoch": 3.19,
+      "grad_norm": 8.505881309509277,
+      "learning_rate": 9.681203007518798e-06,
+      "loss": 0.7136,
+      "step": 2120
+    },
+    {
+      "epoch": 3.2,
+      "grad_norm": 5.51025915145874,
+      "learning_rate": 9.679699248120301e-06,
+      "loss": 0.6037,
+      "step": 2130
+    },
+    {
+      "epoch": 3.22,
+      "grad_norm": 6.39398193359375,
+      "learning_rate": 9.678195488721805e-06,
+      "loss": 0.6799,
+      "step": 2140
+    },
+    {
+      "epoch": 3.23,
+      "grad_norm": 6.508944511413574,
+      "learning_rate": 9.676691729323308e-06,
+      "loss": 0.6587,
+      "step": 2150
+    },
+    {
+      "epoch": 3.25,
+      "grad_norm": 7.027959823608398,
+      "learning_rate": 9.675187969924813e-06,
+      "loss": 0.7096,
+      "step": 2160
+    },
+    {
+      "epoch": 3.26,
+      "grad_norm": 8.690469741821289,
+      "learning_rate": 9.673684210526317e-06,
+      "loss": 0.6169,
+      "step": 2170
+    },
+    {
+      "epoch": 3.28,
+      "grad_norm": 9.489269256591797,
+      "learning_rate": 9.67218045112782e-06,
+      "loss": 0.6643,
+      "step": 2180
+    },
+    {
+      "epoch": 3.29,
+      "grad_norm": 12.934528350830078,
+      "learning_rate": 9.670676691729324e-06,
+      "loss": 0.7485,
+      "step": 2190
+    },
+    {
+      "epoch": 3.31,
+      "grad_norm": 9.072815895080566,
+      "learning_rate": 9.669172932330828e-06,
+      "loss": 0.704,
+      "step": 2200
+    },
+    {
+      "epoch": 3.32,
+      "grad_norm": 7.934593200683594,
+      "learning_rate": 9.667669172932331e-06,
+      "loss": 0.7134,
+      "step": 2210
+    },
+    {
+      "epoch": 3.34,
+      "grad_norm": 6.846796989440918,
+      "learning_rate": 9.666165413533836e-06,
+      "loss": 0.7372,
+      "step": 2220
+    },
+    {
+      "epoch": 3.35,
+      "grad_norm": 6.8362250328063965,
+      "learning_rate": 9.664661654135338e-06,
+      "loss": 0.6854,
+      "step": 2230
+    },
+    {
+      "epoch": 3.37,
+      "grad_norm": 8.184903144836426,
+      "learning_rate": 9.663157894736843e-06,
+      "loss": 0.6662,
+      "step": 2240
+    },
+    {
+      "epoch": 3.38,
+      "grad_norm": 8.972626686096191,
+      "learning_rate": 9.661654135338347e-06,
+      "loss": 0.7895,
+      "step": 2250
+    },
+    {
+      "epoch": 3.4,
+      "grad_norm": 6.524502754211426,
+      "learning_rate": 9.66015037593985e-06,
+      "loss": 0.8453,
+      "step": 2260
+    },
+    {
+      "epoch": 3.41,
+      "grad_norm": 6.215096950531006,
+      "learning_rate": 9.658646616541354e-06,
+      "loss": 0.716,
+      "step": 2270
+    },
+    {
+      "epoch": 3.43,
+      "grad_norm": 6.9516401290893555,
+      "learning_rate": 9.657142857142859e-06,
+      "loss": 0.7453,
+      "step": 2280
+    },
+    {
+      "epoch": 3.44,
+      "grad_norm": 9.131119728088379,
+      "learning_rate": 9.655639097744361e-06,
+      "loss": 0.6474,
+      "step": 2290
+    },
+    {
+      "epoch": 3.46,
+      "grad_norm": 7.063914775848389,
+      "learning_rate": 9.654135338345866e-06,
+      "loss": 0.6535,
+      "step": 2300
+    },
+    {
+      "epoch": 3.47,
+      "grad_norm": 9.410021781921387,
+      "learning_rate": 9.65263157894737e-06,
+      "loss": 0.708,
+      "step": 2310
+    },
+    {
+      "epoch": 3.49,
+      "grad_norm": 7.179042816162109,
+      "learning_rate": 9.651127819548873e-06,
+      "loss": 0.7098,
+      "step": 2320
+    },
+    {
+      "epoch": 3.5,
+      "grad_norm": 8.097248077392578,
+      "learning_rate": 9.649624060150377e-06,
+      "loss": 0.7052,
+      "step": 2330
+    },
+    {
+      "epoch": 3.52,
+      "grad_norm": 4.258429050445557,
+      "learning_rate": 9.64812030075188e-06,
+      "loss": 0.6803,
+      "step": 2340
+    },
+    {
+      "epoch": 3.53,
+      "grad_norm": 7.451633930206299,
+      "learning_rate": 9.646616541353384e-06,
+      "loss": 0.6569,
+      "step": 2350
+    },
+    {
+      "epoch": 3.55,
+      "grad_norm": 9.551535606384277,
+      "learning_rate": 9.645112781954889e-06,
+      "loss": 0.6411,
+      "step": 2360
+    },
+    {
+      "epoch": 3.56,
+      "grad_norm": 6.456385612487793,
+      "learning_rate": 9.643609022556392e-06,
+      "loss": 0.5938,
+      "step": 2370
+    },
+    {
+      "epoch": 3.58,
+      "grad_norm": 5.947078704833984,
+      "learning_rate": 9.642105263157896e-06,
+      "loss": 0.6175,
+      "step": 2380
+    },
+    {
+      "epoch": 3.59,
+      "grad_norm": 5.279054164886475,
+      "learning_rate": 9.6406015037594e-06,
+      "loss": 0.6788,
+      "step": 2390
+    },
+    {
+      "epoch": 3.61,
+      "grad_norm": 7.593316555023193,
+      "learning_rate": 9.639097744360903e-06,
+      "loss": 0.6255,
+      "step": 2400
+    },
+    {
+      "epoch": 3.62,
+      "grad_norm": 7.478080749511719,
+      "learning_rate": 9.637593984962407e-06,
+      "loss": 0.7908,
+      "step": 2410
+    },
+    {
+      "epoch": 3.64,
+      "grad_norm": 9.64027214050293,
+      "learning_rate": 9.636090225563912e-06,
+      "loss": 0.7241,
+      "step": 2420
+    },
+    {
+      "epoch": 3.65,
+      "grad_norm": 14.006696701049805,
+      "learning_rate": 9.634586466165414e-06,
+      "loss": 0.686,
+      "step": 2430
+    },
+    {
+      "epoch": 3.67,
+      "grad_norm": 5.0339789390563965,
+      "learning_rate": 9.633082706766919e-06,
+      "loss": 0.7082,
+      "step": 2440
+    },
+    {
+      "epoch": 3.68,
+      "grad_norm": 8.148447036743164,
+      "learning_rate": 9.631578947368422e-06,
+      "loss": 0.6859,
+      "step": 2450
+    },
+    {
+      "epoch": 3.7,
+      "grad_norm": 7.614720344543457,
+      "learning_rate": 9.630075187969926e-06,
+      "loss": 0.7117,
+      "step": 2460
+    },
+    {
+      "epoch": 3.71,
+      "grad_norm": 9.017003059387207,
+      "learning_rate": 9.62857142857143e-06,
+      "loss": 0.6505,
+      "step": 2470
+    },
+    {
+      "epoch": 3.73,
+      "grad_norm": 6.4466986656188965,
+      "learning_rate": 9.627067669172933e-06,
+      "loss": 0.71,
+      "step": 2480
+    },
+    {
+      "epoch": 3.74,
+      "grad_norm": 7.973327159881592,
+      "learning_rate": 9.625563909774436e-06,
+      "loss": 0.6466,
+      "step": 2490
+    },
+    {
+      "epoch": 3.76,
+      "grad_norm": 6.712606906890869,
+      "learning_rate": 9.62406015037594e-06,
+      "loss": 0.6767,
+      "step": 2500
+    },
+    {
+      "epoch": 3.77,
+      "grad_norm": 8.149372100830078,
+      "learning_rate": 9.622556390977445e-06,
+      "loss": 0.783,
+      "step": 2510
+    },
+    {
+      "epoch": 3.79,
+      "grad_norm": 8.645270347595215,
+      "learning_rate": 9.621052631578947e-06,
+      "loss": 0.6978,
+      "step": 2520
+    },
+    {
+      "epoch": 3.8,
+      "grad_norm": 9.347142219543457,
+      "learning_rate": 9.619548872180452e-06,
+      "loss": 0.583,
+      "step": 2530
+    },
+    {
+      "epoch": 3.82,
+      "grad_norm": 7.905392169952393,
+      "learning_rate": 9.618045112781956e-06,
+      "loss": 0.6884,
+      "step": 2540
+    },
+    {
+      "epoch": 3.83,
+      "grad_norm": 8.783331871032715,
+      "learning_rate": 9.61654135338346e-06,
+      "loss": 0.691,
+      "step": 2550
+    },
+    {
+      "epoch": 3.85,
+      "grad_norm": 8.456209182739258,
+      "learning_rate": 9.615037593984963e-06,
+      "loss": 0.7281,
+      "step": 2560
+    },
+    {
+      "epoch": 3.86,
+      "grad_norm": 6.667693138122559,
+      "learning_rate": 9.613533834586468e-06,
+      "loss": 0.6912,
+      "step": 2570
+    },
+    {
+      "epoch": 3.88,
+      "grad_norm": 8.541569709777832,
+      "learning_rate": 9.61203007518797e-06,
+      "loss": 0.6853,
+      "step": 2580
+    },
+    {
+      "epoch": 3.89,
+      "grad_norm": 4.732927322387695,
+      "learning_rate": 9.610526315789475e-06,
+      "loss": 0.6647,
+      "step": 2590
+    },
+    {
+      "epoch": 3.91,
+      "grad_norm": 7.604156017303467,
+      "learning_rate": 9.609022556390978e-06,
+      "loss": 0.7526,
+      "step": 2600
+    },
+    {
+      "epoch": 3.92,
+      "grad_norm": 8.218050956726074,
+      "learning_rate": 9.607518796992482e-06,
+      "loss": 0.6828,
+      "step": 2610
+    },
+    {
+      "epoch": 3.94,
+      "grad_norm": 5.613206386566162,
+      "learning_rate": 9.606015037593985e-06,
+      "loss": 0.6964,
+      "step": 2620
+    },
+    {
+      "epoch": 3.95,
+      "grad_norm": 9.644120216369629,
+      "learning_rate": 9.604511278195489e-06,
+      "loss": 0.6912,
+      "step": 2630
+    },
+    {
+      "epoch": 3.97,
+      "grad_norm": 8.14504337310791,
+      "learning_rate": 9.603007518796993e-06,
+      "loss": 0.7527,
+      "step": 2640
+    },
+    {
+      "epoch": 3.98,
+      "grad_norm": 6.1560468673706055,
+      "learning_rate": 9.601503759398498e-06,
+      "loss": 0.6145,
+      "step": 2650
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 20.564706802368164,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.6935,
+      "step": 2660
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.9236,
+      "eval_loss": 0.319810152053833,
+      "eval_runtime": 84.5508,
+      "eval_samples_per_second": 118.272,
+      "eval_steps_per_second": 0.473,
+      "step": 2660
+    },
+    {
+      "epoch": 4.02,
+      "grad_norm": 9.448854446411133,
+      "learning_rate": 9.598496240601505e-06,
+      "loss": 0.631,
+      "step": 2670
+    },
+    {
+      "epoch": 4.03,
+      "grad_norm": 10.194000244140625,
+      "learning_rate": 9.596992481203008e-06,
+      "loss": 0.6046,
+      "step": 2680
+    },
+    {
+      "epoch": 4.05,
+      "grad_norm": 8.277205467224121,
+      "learning_rate": 9.595488721804512e-06,
+      "loss": 0.5853,
+      "step": 2690
+    },
+    {
+      "epoch": 4.06,
+      "grad_norm": 7.616865158081055,
+      "learning_rate": 9.593984962406015e-06,
+      "loss": 0.6168,
+      "step": 2700
+    },
+    {
+      "epoch": 4.08,
+      "grad_norm": 5.158208847045898,
+      "learning_rate": 9.59248120300752e-06,
+      "loss": 0.5893,
+      "step": 2710
+    },
+    {
+      "epoch": 4.09,
+      "grad_norm": 10.609253883361816,
+      "learning_rate": 9.590977443609022e-06,
+      "loss": 0.5819,
+      "step": 2720
+    },
+    {
+      "epoch": 4.11,
+      "grad_norm": 7.288332462310791,
+      "learning_rate": 9.589473684210528e-06,
+      "loss": 0.6384,
+      "step": 2730
+    },
+    {
+      "epoch": 4.12,
+      "grad_norm": 6.625866889953613,
+      "learning_rate": 9.587969924812031e-06,
+      "loss": 0.6841,
+      "step": 2740
+    },
+    {
+      "epoch": 4.14,
+      "grad_norm": 8.38702392578125,
+      "learning_rate": 9.586466165413535e-06,
+      "loss": 0.5815,
+      "step": 2750
+    },
+    {
+      "epoch": 4.15,
+      "grad_norm": 6.58852481842041,
+      "learning_rate": 9.584962406015038e-06,
+      "loss": 0.7091,
+      "step": 2760
+    },
+    {
+      "epoch": 4.17,
+      "grad_norm": 5.776881217956543,
+      "learning_rate": 9.583458646616542e-06,
+      "loss": 0.6426,
+      "step": 2770
+    },
+    {
+      "epoch": 4.18,
+      "grad_norm": 7.806540489196777,
+      "learning_rate": 9.581954887218045e-06,
+      "loss": 0.6459,
+      "step": 2780
+    },
+    {
+      "epoch": 4.2,
+      "grad_norm": 7.378940582275391,
+      "learning_rate": 9.58045112781955e-06,
+      "loss": 0.6553,
+      "step": 2790
+    },
+    {
+      "epoch": 4.21,
+      "grad_norm": 8.37366008758545,
+      "learning_rate": 9.578947368421054e-06,
+      "loss": 0.6024,
+      "step": 2800
+    },
+    {
+      "epoch": 4.23,
+      "grad_norm": 5.783264636993408,
+      "learning_rate": 9.577443609022557e-06,
+      "loss": 0.5959,
+      "step": 2810
+    },
+    {
+      "epoch": 4.24,
+      "grad_norm": 6.4687676429748535,
+      "learning_rate": 9.575939849624061e-06,
+      "loss": 0.5669,
+      "step": 2820
+    },
+    {
+      "epoch": 4.26,
+      "grad_norm": 10.575803756713867,
+      "learning_rate": 9.574436090225564e-06,
+      "loss": 0.6461,
+      "step": 2830
+    },
+    {
+      "epoch": 4.27,
+      "grad_norm": 9.703124046325684,
+      "learning_rate": 9.572932330827068e-06,
+      "loss": 0.5784,
+      "step": 2840
+    },
+    {
+      "epoch": 4.29,
+      "grad_norm": 9.454757690429688,
+      "learning_rate": 9.571428571428573e-06,
+      "loss": 0.7381,
+      "step": 2850
+    },
+    {
+      "epoch": 4.3,
+      "grad_norm": 7.034806728363037,
+      "learning_rate": 9.569924812030075e-06,
+      "loss": 0.6277,
+      "step": 2860
+    },
+    {
+      "epoch": 4.32,
+      "grad_norm": 10.1060791015625,
+      "learning_rate": 9.56842105263158e-06,
+      "loss": 0.6973,
+      "step": 2870
+    },
+    {
+      "epoch": 4.33,
+      "grad_norm": 7.225138187408447,
+      "learning_rate": 9.566917293233084e-06,
+      "loss": 0.587,
+      "step": 2880
+    },
+    {
+      "epoch": 4.35,
+      "grad_norm": 6.221525192260742,
+      "learning_rate": 9.565413533834587e-06,
+      "loss": 0.6006,
+      "step": 2890
+    },
+    {
+      "epoch": 4.36,
+      "grad_norm": 6.329552173614502,
+      "learning_rate": 9.56390977443609e-06,
+      "loss": 0.6474,
+      "step": 2900
+    },
+    {
+      "epoch": 4.38,
+      "grad_norm": 8.411649703979492,
+      "learning_rate": 9.562406015037596e-06,
+      "loss": 0.5376,
+      "step": 2910
+    },
+    {
+      "epoch": 4.39,
+      "grad_norm": 8.27790355682373,
+      "learning_rate": 9.560902255639098e-06,
+      "loss": 0.6324,
+      "step": 2920
+    },
+    {
+      "epoch": 4.41,
+      "grad_norm": 6.995235443115234,
+      "learning_rate": 9.559398496240603e-06,
+      "loss": 0.6178,
+      "step": 2930
+    },
+    {
+      "epoch": 4.42,
+      "grad_norm": 8.169748306274414,
+      "learning_rate": 9.557894736842107e-06,
+      "loss": 0.5546,
+      "step": 2940
+    },
+    {
+      "epoch": 4.44,
+      "grad_norm": 7.832982063293457,
+      "learning_rate": 9.55639097744361e-06,
+      "loss": 0.5873,
+      "step": 2950
+    },
+    {
+      "epoch": 4.45,
+      "grad_norm": 7.024545192718506,
+      "learning_rate": 9.554887218045114e-06,
+      "loss": 0.5919,
+      "step": 2960
+    },
+    {
+      "epoch": 4.47,
+      "grad_norm": 8.610920906066895,
+      "learning_rate": 9.553383458646617e-06,
+      "loss": 0.6324,
+      "step": 2970
+    },
+    {
+      "epoch": 4.48,
+      "grad_norm": 8.49885368347168,
+      "learning_rate": 9.55187969924812e-06,
+      "loss": 0.6392,
+      "step": 2980
+    },
+    {
+      "epoch": 4.5,
+      "grad_norm": 6.013737678527832,
+      "learning_rate": 9.550375939849624e-06,
+      "loss": 0.6267,
+      "step": 2990
+    },
+    {
+      "epoch": 4.51,
+      "grad_norm": 9.457529067993164,
+      "learning_rate": 9.54887218045113e-06,
+      "loss": 0.633,
+      "step": 3000
+    },
+    {
+      "epoch": 4.53,
+      "grad_norm": 7.126248359680176,
+      "learning_rate": 9.547368421052631e-06,
+      "loss": 0.5527,
+      "step": 3010
+    },
+    {
+      "epoch": 4.54,
+      "grad_norm": 8.481447219848633,
+      "learning_rate": 9.545864661654136e-06,
+      "loss": 0.7163,
+      "step": 3020
+    },
+    {
+      "epoch": 4.56,
+      "grad_norm": 10.300518035888672,
+      "learning_rate": 9.54436090225564e-06,
+      "loss": 0.5921,
+      "step": 3030
+    },
+    {
+      "epoch": 4.57,
+      "grad_norm": 8.265804290771484,
+      "learning_rate": 9.542857142857143e-06,
+      "loss": 0.5952,
+      "step": 3040
+    },
+    {
+      "epoch": 4.59,
+      "grad_norm": 5.028606414794922,
+      "learning_rate": 9.541353383458647e-06,
+      "loss": 0.6269,
+      "step": 3050
+    },
+    {
+      "epoch": 4.6,
+      "grad_norm": 8.997878074645996,
+      "learning_rate": 9.53984962406015e-06,
+      "loss": 0.5761,
+      "step": 3060
+    },
+    {
+      "epoch": 4.62,
+      "grad_norm": 9.585675239562988,
+      "learning_rate": 9.538345864661654e-06,
+      "loss": 0.5851,
+      "step": 3070
+    },
+    {
+      "epoch": 4.63,
+      "grad_norm": 5.83755350112915,
+      "learning_rate": 9.53684210526316e-06,
+      "loss": 0.6105,
+      "step": 3080
+    },
+    {
+      "epoch": 4.65,
+      "grad_norm": 5.208207607269287,
+      "learning_rate": 9.535338345864663e-06,
+      "loss": 0.5786,
+      "step": 3090
+    },
+    {
+      "epoch": 4.66,
+      "grad_norm": 9.895461082458496,
+      "learning_rate": 9.533834586466166e-06,
+      "loss": 0.623,
+      "step": 3100
+    },
+    {
+      "epoch": 4.68,
+      "grad_norm": 8.958138465881348,
+      "learning_rate": 9.53233082706767e-06,
+      "loss": 0.5708,
+      "step": 3110
+    },
+    {
+      "epoch": 4.69,
+      "grad_norm": 10.452126502990723,
+      "learning_rate": 9.530827067669173e-06,
+      "loss": 0.5694,
+      "step": 3120
+    },
+    {
+      "epoch": 4.71,
+      "grad_norm": 7.20021915435791,
+      "learning_rate": 9.529323308270677e-06,
+      "loss": 0.6267,
+      "step": 3130
+    },
+    {
+      "epoch": 4.72,
+      "grad_norm": 7.995909690856934,
+      "learning_rate": 9.527819548872182e-06,
+      "loss": 0.6989,
+      "step": 3140
+    },
+    {
+      "epoch": 4.74,
+      "grad_norm": 6.9314985275268555,
+      "learning_rate": 9.526315789473684e-06,
+      "loss": 0.6093,
+      "step": 3150
+    },
+    {
+      "epoch": 4.75,
+      "grad_norm": 10.158616065979004,
+      "learning_rate": 9.524812030075189e-06,
+      "loss": 0.5696,
+      "step": 3160
+    },
+    {
+      "epoch": 4.77,
+      "grad_norm": 7.637181758880615,
+      "learning_rate": 9.523308270676693e-06,
+      "loss": 0.7494,
+      "step": 3170
+    },
+    {
+      "epoch": 4.78,
+      "grad_norm": 7.443474769592285,
+      "learning_rate": 9.521804511278196e-06,
+      "loss": 0.7636,
+      "step": 3180
+    },
+    {
+      "epoch": 4.8,
+      "grad_norm": 6.130582809448242,
+      "learning_rate": 9.5203007518797e-06,
+      "loss": 0.7397,
+      "step": 3190
+    },
+    {
+      "epoch": 4.81,
+      "grad_norm": 8.699774742126465,
+      "learning_rate": 9.518796992481205e-06,
+      "loss": 0.5461,
+      "step": 3200
+    },
+    {
+      "epoch": 4.83,
+      "grad_norm": 8.060851097106934,
+      "learning_rate": 9.517293233082707e-06,
+      "loss": 0.5424,
+      "step": 3210
+    },
+    {
+      "epoch": 4.84,
+      "grad_norm": 6.084632396697998,
+      "learning_rate": 9.515789473684212e-06,
+      "loss": 0.6181,
+      "step": 3220
+    },
+    {
+      "epoch": 4.86,
+      "grad_norm": 8.804571151733398,
+      "learning_rate": 9.514285714285715e-06,
+      "loss": 0.6696,
+      "step": 3230
+    },
+    {
+      "epoch": 4.87,
+      "grad_norm": 8.552626609802246,
+      "learning_rate": 9.512781954887219e-06,
+      "loss": 0.6237,
+      "step": 3240
+    },
+    {
+      "epoch": 4.89,
+      "grad_norm": 8.930567741394043,
+      "learning_rate": 9.511278195488722e-06,
+      "loss": 0.6427,
+      "step": 3250
+    },
+    {
+      "epoch": 4.9,
+      "grad_norm": 8.916244506835938,
+      "learning_rate": 9.509774436090226e-06,
+      "loss": 0.608,
+      "step": 3260
+    },
+    {
+      "epoch": 4.92,
+      "grad_norm": 12.679169654846191,
+      "learning_rate": 9.50827067669173e-06,
+      "loss": 0.622,
+      "step": 3270
+    },
+    {
+      "epoch": 4.93,
+      "grad_norm": 9.21071720123291,
+      "learning_rate": 9.506766917293235e-06,
+      "loss": 0.6153,
+      "step": 3280
+    },
+    {
+      "epoch": 4.95,
+      "grad_norm": 8.040297508239746,
+      "learning_rate": 9.505263157894738e-06,
+      "loss": 0.654,
+      "step": 3290
+    },
+    {
+      "epoch": 4.96,
+      "grad_norm": 6.395382404327393,
+      "learning_rate": 9.503759398496242e-06,
+      "loss": 0.6729,
+      "step": 3300
+    },
+    {
+      "epoch": 4.98,
+      "grad_norm": 8.437057495117188,
+      "learning_rate": 9.502255639097745e-06,
+      "loss": 0.6457,
+      "step": 3310
+    },
+    {
+      "epoch": 4.99,
+      "grad_norm": 7.987279415130615,
+      "learning_rate": 9.500751879699249e-06,
+      "loss": 0.6159,
+      "step": 3320
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9289,
+      "eval_loss": 0.2945062816143036,
+      "eval_runtime": 84.9367,
+      "eval_samples_per_second": 117.735,
+      "eval_steps_per_second": 0.471,
+      "step": 3325
+    },
+    {
+      "epoch": 5.01,
+      "grad_norm": 6.12150239944458,
+      "learning_rate": 9.499248120300752e-06,
+      "loss": 0.4871,
+      "step": 3330
+    },
+    {
+      "epoch": 5.02,
+      "grad_norm": 8.20666217803955,
+      "learning_rate": 9.497744360902257e-06,
+      "loss": 0.6011,
+      "step": 3340
+    },
+    {
+      "epoch": 5.04,
+      "grad_norm": 8.818642616271973,
+      "learning_rate": 9.49624060150376e-06,
+      "loss": 0.5681,
+      "step": 3350
+    },
+    {
+      "epoch": 5.05,
+      "grad_norm": 5.606151103973389,
+      "learning_rate": 9.494736842105265e-06,
+      "loss": 0.5494,
+      "step": 3360
+    },
+    {
+      "epoch": 5.07,
+      "grad_norm": 6.230663299560547,
+      "learning_rate": 9.493233082706768e-06,
+      "loss": 0.6138,
+      "step": 3370
+    },
+    {
+      "epoch": 5.08,
+      "grad_norm": 6.923035621643066,
+      "learning_rate": 9.491729323308272e-06,
+      "loss": 0.6398,
+      "step": 3380
+    },
+    {
+      "epoch": 5.1,
+      "grad_norm": 8.464035034179688,
+      "learning_rate": 9.490225563909775e-06,
+      "loss": 0.533,
+      "step": 3390
+    },
+    {
+      "epoch": 5.11,
+      "grad_norm": 5.440852165222168,
+      "learning_rate": 9.488721804511279e-06,
+      "loss": 0.5901,
+      "step": 3400
+    },
+    {
+      "epoch": 5.13,
+      "grad_norm": 6.880829334259033,
+      "learning_rate": 9.487218045112782e-06,
+      "loss": 0.5699,
+      "step": 3410
+    },
+    {
+      "epoch": 5.14,
+      "grad_norm": 6.773617267608643,
+      "learning_rate": 9.485714285714287e-06,
+      "loss": 0.5164,
+      "step": 3420
+    },
+    {
+      "epoch": 5.16,
+      "grad_norm": 6.794729232788086,
+      "learning_rate": 9.484210526315791e-06,
+      "loss": 0.4939,
+      "step": 3430
+    },
+    {
+      "epoch": 5.17,
+      "grad_norm": 8.347722053527832,
+      "learning_rate": 9.482706766917294e-06,
+      "loss": 0.6138,
+      "step": 3440
+    },
+    {
+      "epoch": 5.19,
+      "grad_norm": 6.279055595397949,
+      "learning_rate": 9.481203007518798e-06,
+      "loss": 0.4792,
+      "step": 3450
+    },
+    {
+      "epoch": 5.2,
+      "grad_norm": 7.93798828125,
+      "learning_rate": 9.479699248120301e-06,
+      "loss": 0.5945,
+      "step": 3460
+    },
+    {
+      "epoch": 5.22,
+      "grad_norm": 6.767178535461426,
+      "learning_rate": 9.478195488721805e-06,
+      "loss": 0.5878,
+      "step": 3470
+    },
+    {
+      "epoch": 5.23,
+      "grad_norm": 6.87293004989624,
+      "learning_rate": 9.476691729323308e-06,
+      "loss": 0.566,
+      "step": 3480
+    },
+    {
+      "epoch": 5.25,
+      "grad_norm": 2.407437562942505,
+      "learning_rate": 9.475187969924814e-06,
+      "loss": 0.5014,
+      "step": 3490
+    },
+    {
+      "epoch": 5.26,
+      "grad_norm": 8.233712196350098,
+      "learning_rate": 9.473684210526315e-06,
+      "loss": 0.5879,
+      "step": 3500
+    },
+    {
+      "epoch": 5.28,
+      "grad_norm": 7.905375003814697,
+      "learning_rate": 9.47218045112782e-06,
+      "loss": 0.6127,
+      "step": 3510
+    },
+    {
+      "epoch": 5.29,
+      "grad_norm": 5.8037238121032715,
+      "learning_rate": 9.470676691729324e-06,
+      "loss": 0.6048,
+      "step": 3520
+    },
+    {
+      "epoch": 5.31,
+      "grad_norm": 9.2665433883667,
+      "learning_rate": 9.469172932330828e-06,
+      "loss": 0.6233,
+      "step": 3530
+    },
+    {
+      "epoch": 5.32,
+      "grad_norm": 5.650614261627197,
+      "learning_rate": 9.467669172932331e-06,
+      "loss": 0.5703,
+      "step": 3540
+    },
+    {
+      "epoch": 5.34,
+      "grad_norm": 5.246155738830566,
+      "learning_rate": 9.466165413533835e-06,
+      "loss": 0.5108,
+      "step": 3550
+    },
+    {
+      "epoch": 5.35,
+      "grad_norm": 8.701322555541992,
+      "learning_rate": 9.464661654135338e-06,
+      "loss": 0.5783,
+      "step": 3560
+    },
+    {
+      "epoch": 5.37,
+      "grad_norm": 5.870892524719238,
+      "learning_rate": 9.463157894736844e-06,
+      "loss": 0.5546,
+      "step": 3570
+    },
+    {
+      "epoch": 5.38,
+      "grad_norm": 8.061163902282715,
+      "learning_rate": 9.461654135338347e-06,
+      "loss": 0.5973,
+      "step": 3580
+    },
+    {
+      "epoch": 5.4,
+      "grad_norm": 4.166900157928467,
+      "learning_rate": 9.46015037593985e-06,
+      "loss": 0.6042,
+      "step": 3590
+    },
+    {
+      "epoch": 5.41,
+      "grad_norm": 5.8524346351623535,
+      "learning_rate": 9.458646616541354e-06,
+      "loss": 0.5307,
+      "step": 3600
+    },
+    {
+      "epoch": 5.43,
+      "grad_norm": 7.229081153869629,
+      "learning_rate": 9.457142857142858e-06,
+      "loss": 0.6533,
+      "step": 3610
+    },
+    {
+      "epoch": 5.44,
+      "grad_norm": 5.403026580810547,
+      "learning_rate": 9.455639097744361e-06,
+      "loss": 0.5059,
+      "step": 3620
+    },
+    {
+      "epoch": 5.46,
+      "grad_norm": 7.53814697265625,
+      "learning_rate": 9.454135338345866e-06,
+      "loss": 0.6576,
+      "step": 3630
+    },
+    {
+      "epoch": 5.47,
+      "grad_norm": 8.08530330657959,
+      "learning_rate": 9.452631578947368e-06,
+      "loss": 0.5881,
+      "step": 3640
+    },
+    {
+      "epoch": 5.49,
+      "grad_norm": 7.80808687210083,
+      "learning_rate": 9.451127819548873e-06,
+      "loss": 0.5725,
+      "step": 3650
+    },
+    {
+      "epoch": 5.5,
+      "grad_norm": 4.368475914001465,
+      "learning_rate": 9.449624060150377e-06,
+      "loss": 0.5538,
+      "step": 3660
+    },
+    {
+      "epoch": 5.52,
+      "grad_norm": 7.806415557861328,
+      "learning_rate": 9.44812030075188e-06,
+      "loss": 0.6739,
+      "step": 3670
+    },
+    {
+      "epoch": 5.53,
+      "grad_norm": 8.047362327575684,
+      "learning_rate": 9.446616541353384e-06,
+      "loss": 0.5229,
+      "step": 3680
+    },
+    {
+      "epoch": 5.55,
+      "grad_norm": 5.89243221282959,
+      "learning_rate": 9.445112781954887e-06,
+      "loss": 0.599,
+      "step": 3690
+    },
+    {
+      "epoch": 5.56,
+      "grad_norm": 6.812804222106934,
+      "learning_rate": 9.443609022556391e-06,
+      "loss": 0.6213,
+      "step": 3700
+    },
+    {
+      "epoch": 5.58,
+      "grad_norm": 7.177376747131348,
+      "learning_rate": 9.442105263157896e-06,
+      "loss": 0.5568,
+      "step": 3710
+    },
+    {
+      "epoch": 5.59,
+      "grad_norm": 4.684938907623291,
+      "learning_rate": 9.4406015037594e-06,
+      "loss": 0.5266,
+      "step": 3720
+    },
+    {
+      "epoch": 5.61,
+      "grad_norm": 9.440715789794922,
+      "learning_rate": 9.439097744360903e-06,
+      "loss": 0.6431,
+      "step": 3730
+    },
+    {
+      "epoch": 5.62,
+      "grad_norm": 6.564364910125732,
+      "learning_rate": 9.437593984962407e-06,
+      "loss": 0.4702,
+      "step": 3740
+    },
+    {
+      "epoch": 5.64,
+      "grad_norm": 7.499642372131348,
+      "learning_rate": 9.43609022556391e-06,
+      "loss": 0.5249,
+      "step": 3750
+    },
+    {
+      "epoch": 5.65,
+      "grad_norm": 5.27653169631958,
+      "learning_rate": 9.434586466165414e-06,
+      "loss": 0.4517,
+      "step": 3760
+    },
+    {
+      "epoch": 5.67,
+      "grad_norm": 4.881272792816162,
+      "learning_rate": 9.433082706766919e-06,
+      "loss": 0.4682,
+      "step": 3770
+    },
+    {
+      "epoch": 5.68,
+      "grad_norm": 9.005290985107422,
+      "learning_rate": 9.43157894736842e-06,
+      "loss": 0.5642,
+      "step": 3780
+    },
+    {
+      "epoch": 5.7,
+      "grad_norm": 6.421182155609131,
+      "learning_rate": 9.430075187969926e-06,
+      "loss": 0.4855,
+      "step": 3790
+    },
+    {
+      "epoch": 5.71,
+      "grad_norm": 7.035130023956299,
+      "learning_rate": 9.42857142857143e-06,
+      "loss": 0.6362,
+      "step": 3800
+    },
+    {
+      "epoch": 5.73,
+      "grad_norm": 5.77438497543335,
+      "learning_rate": 9.427067669172933e-06,
+      "loss": 0.5876,
+      "step": 3810
+    },
+    {
+      "epoch": 5.74,
+      "grad_norm": 8.328043937683105,
+      "learning_rate": 9.425563909774437e-06,
+      "loss": 0.4768,
+      "step": 3820
+    },
+    {
+      "epoch": 5.76,
+      "grad_norm": 5.7907586097717285,
+      "learning_rate": 9.424060150375942e-06,
+      "loss": 0.5941,
+      "step": 3830
+    },
+    {
+      "epoch": 5.77,
+      "grad_norm": 8.730267524719238,
+      "learning_rate": 9.422556390977444e-06,
+      "loss": 0.6426,
+      "step": 3840
+    },
+    {
+      "epoch": 5.79,
+      "grad_norm": 8.710532188415527,
+      "learning_rate": 9.421052631578949e-06,
+      "loss": 0.5909,
+      "step": 3850
+    },
+    {
+      "epoch": 5.8,
+      "grad_norm": 8.74202823638916,
+      "learning_rate": 9.419548872180452e-06,
+      "loss": 0.5735,
+      "step": 3860
+    },
+    {
+      "epoch": 5.82,
+      "grad_norm": 7.489967346191406,
+      "learning_rate": 9.418045112781956e-06,
+      "loss": 0.6072,
+      "step": 3870
+    },
+    {
+      "epoch": 5.83,
+      "grad_norm": 6.942547798156738,
+      "learning_rate": 9.41654135338346e-06,
+      "loss": 0.5513,
+      "step": 3880
+    },
+    {
+      "epoch": 5.85,
+      "grad_norm": 5.517817497253418,
+      "learning_rate": 9.415037593984963e-06,
+      "loss": 0.5317,
+      "step": 3890
+    },
+    {
+      "epoch": 5.86,
+      "grad_norm": 6.26224946975708,
+      "learning_rate": 9.413533834586466e-06,
+      "loss": 0.6014,
+      "step": 3900
+    },
+    {
+      "epoch": 5.88,
+      "grad_norm": 4.848892688751221,
+      "learning_rate": 9.412030075187972e-06,
+      "loss": 0.5628,
+      "step": 3910
+    },
+    {
+      "epoch": 5.89,
+      "grad_norm": 7.279343128204346,
+      "learning_rate": 9.410526315789475e-06,
+      "loss": 0.631,
+      "step": 3920
+    },
+    {
+      "epoch": 5.91,
+      "grad_norm": 5.791496753692627,
+      "learning_rate": 9.409022556390979e-06,
+      "loss": 0.5312,
+      "step": 3930
+    },
+    {
+      "epoch": 5.92,
+      "grad_norm": 4.935235977172852,
+      "learning_rate": 9.407518796992482e-06,
+      "loss": 0.4694,
+      "step": 3940
+    },
+    {
+      "epoch": 5.94,
+      "grad_norm": 5.741876125335693,
+      "learning_rate": 9.406015037593986e-06,
+      "loss": 0.6073,
+      "step": 3950
+    },
+    {
+      "epoch": 5.95,
+      "grad_norm": 5.398350715637207,
+      "learning_rate": 9.40451127819549e-06,
+      "loss": 0.6009,
+      "step": 3960
+    },
+    {
+      "epoch": 5.97,
+      "grad_norm": 6.093377590179443,
+      "learning_rate": 9.403007518796994e-06,
+      "loss": 0.5845,
+      "step": 3970
+    },
+    {
+      "epoch": 5.98,
+      "grad_norm": 8.6488676071167,
+      "learning_rate": 9.401503759398496e-06,
+      "loss": 0.5932,
+      "step": 3980
+    },
+    {
+      "epoch": 6.0,
+      "grad_norm": 3.0173494815826416,
+      "learning_rate": 9.4e-06,
+      "loss": 0.4423,
+      "step": 3990
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.925,
+      "eval_loss": 0.2876322865486145,
+      "eval_runtime": 85.2695,
+      "eval_samples_per_second": 117.275,
+      "eval_steps_per_second": 0.469,
+      "step": 3990
+    },
+    {
+      "epoch": 6.02,
+      "grad_norm": 4.103921890258789,
+      "learning_rate": 9.398496240601505e-06,
+      "loss": 0.5346,
+      "step": 4000
+    },
+    {
+      "epoch": 6.03,
+      "grad_norm": 4.625704288482666,
+      "learning_rate": 9.396992481203009e-06,
+      "loss": 0.5522,
+      "step": 4010
+    },
+    {
+      "epoch": 6.05,
+      "grad_norm": 7.251491546630859,
+      "learning_rate": 9.395488721804512e-06,
+      "loss": 0.6501,
+      "step": 4020
+    },
+    {
+      "epoch": 6.06,
+      "grad_norm": 8.962389945983887,
+      "learning_rate": 9.393984962406016e-06,
+      "loss": 0.5402,
+      "step": 4030
+    },
+    {
+      "epoch": 6.08,
+      "grad_norm": 6.207771301269531,
+      "learning_rate": 9.392481203007519e-06,
+      "loss": 0.5545,
+      "step": 4040
+    },
+    {
+      "epoch": 6.09,
+      "grad_norm": 5.253688812255859,
+      "learning_rate": 9.390977443609023e-06,
+      "loss": 0.5084,
+      "step": 4050
+    },
+    {
+      "epoch": 6.11,
+      "grad_norm": 7.277046203613281,
+      "learning_rate": 9.389473684210528e-06,
+      "loss": 0.5115,
+      "step": 4060
+    },
+    {
+      "epoch": 6.12,
+      "grad_norm": 5.671750068664551,
+      "learning_rate": 9.38796992481203e-06,
+      "loss": 0.5508,
+      "step": 4070
+    },
+    {
+      "epoch": 6.14,
+      "grad_norm": 3.9672038555145264,
+      "learning_rate": 9.386466165413535e-06,
+      "loss": 0.494,
+      "step": 4080
+    },
+    {
+      "epoch": 6.15,
+      "grad_norm": 6.129919052124023,
+      "learning_rate": 9.384962406015038e-06,
+      "loss": 0.5898,
+      "step": 4090
+    },
+    {
+      "epoch": 6.17,
+      "grad_norm": 6.198451519012451,
+      "learning_rate": 9.383458646616542e-06,
+      "loss": 0.5378,
+      "step": 4100
+    },
+    {
+      "epoch": 6.18,
+      "grad_norm": 9.286908149719238,
+      "learning_rate": 9.381954887218045e-06,
+      "loss": 0.5656,
+      "step": 4110
+    },
+    {
+      "epoch": 6.2,
+      "grad_norm": 6.862420082092285,
+      "learning_rate": 9.38045112781955e-06,
+      "loss": 0.5457,
+      "step": 4120
+    },
+    {
+      "epoch": 6.21,
+      "grad_norm": 5.948605537414551,
+      "learning_rate": 9.378947368421052e-06,
+      "loss": 0.5564,
+      "step": 4130
+    },
+    {
+      "epoch": 6.23,
+      "grad_norm": 7.1652116775512695,
+      "learning_rate": 9.377443609022558e-06,
+      "loss": 0.5624,
+      "step": 4140
+    },
+    {
+      "epoch": 6.24,
+      "grad_norm": 7.091752052307129,
+      "learning_rate": 9.375939849624061e-06,
+      "loss": 0.532,
+      "step": 4150
+    },
+    {
+      "epoch": 6.26,
+      "grad_norm": 4.2119903564453125,
+      "learning_rate": 9.374436090225565e-06,
+      "loss": 0.385,
+      "step": 4160
+    },
+    {
+      "epoch": 6.27,
+      "grad_norm": 9.477155685424805,
+      "learning_rate": 9.372932330827068e-06,
+      "loss": 0.5019,
+      "step": 4170
+    },
+    {
+      "epoch": 6.29,
+      "grad_norm": 7.294814109802246,
+      "learning_rate": 9.371428571428572e-06,
+      "loss": 0.556,
+      "step": 4180
+    },
+    {
+      "epoch": 6.3,
+      "grad_norm": 8.124314308166504,
+      "learning_rate": 9.369924812030075e-06,
+      "loss": 0.548,
+      "step": 4190
+    },
+    {
+      "epoch": 6.32,
+      "grad_norm": 6.1076483726501465,
+      "learning_rate": 9.36842105263158e-06,
+      "loss": 0.5228,
+      "step": 4200
+    },
+    {
+      "epoch": 6.33,
+      "grad_norm": 7.360411643981934,
+      "learning_rate": 9.366917293233084e-06,
+      "loss": 0.6335,
+      "step": 4210
+    },
+    {
+      "epoch": 6.35,
+      "grad_norm": 6.000509738922119,
+      "learning_rate": 9.365413533834588e-06,
+      "loss": 0.5727,
+      "step": 4220
+    },
+    {
+      "epoch": 6.36,
+      "grad_norm": 8.538400650024414,
+      "learning_rate": 9.363909774436091e-06,
+      "loss": 0.5135,
+      "step": 4230
+    },
+    {
+      "epoch": 6.38,
+      "grad_norm": 6.543038845062256,
+      "learning_rate": 9.362406015037595e-06,
+      "loss": 0.4859,
+      "step": 4240
+    },
+    {
+      "epoch": 6.39,
+      "grad_norm": 7.515405178070068,
+      "learning_rate": 9.360902255639098e-06,
+      "loss": 0.6172,
+      "step": 4250
+    },
+    {
+      "epoch": 6.41,
+      "grad_norm": 4.859574317932129,
+      "learning_rate": 9.359398496240603e-06,
+      "loss": 0.407,
+      "step": 4260
+    },
+    {
+      "epoch": 6.42,
+      "grad_norm": 5.684931755065918,
+      "learning_rate": 9.357894736842105e-06,
+      "loss": 0.5748,
+      "step": 4270
+    },
+    {
+      "epoch": 6.44,
+      "grad_norm": 5.063128471374512,
+      "learning_rate": 9.35639097744361e-06,
+      "loss": 0.5363,
+      "step": 4280
+    },
+    {
+      "epoch": 6.45,
+      "grad_norm": 9.34011173248291,
+      "learning_rate": 9.354887218045114e-06,
+      "loss": 0.5072,
+      "step": 4290
+    },
+    {
+      "epoch": 6.47,
+      "grad_norm": 6.302648544311523,
+      "learning_rate": 9.353383458646617e-06,
+      "loss": 0.4891,
+      "step": 4300
+    },
+    {
+      "epoch": 6.48,
+      "grad_norm": 6.268799781799316,
+      "learning_rate": 9.351879699248121e-06,
+      "loss": 0.6215,
+      "step": 4310
+    },
+    {
+      "epoch": 6.5,
+      "grad_norm": 5.54179573059082,
+      "learning_rate": 9.350375939849624e-06,
+      "loss": 0.5827,
+      "step": 4320
+    },
+    {
+      "epoch": 6.51,
+      "grad_norm": 6.042153835296631,
+      "learning_rate": 9.348872180451128e-06,
+      "loss": 0.5205,
+      "step": 4330
+    },
+    {
+      "epoch": 6.53,
+      "grad_norm": 7.558413982391357,
+      "learning_rate": 9.347368421052633e-06,
+      "loss": 0.5357,
+      "step": 4340
+    },
+    {
+      "epoch": 6.54,
+      "grad_norm": 7.838019847869873,
+      "learning_rate": 9.345864661654137e-06,
+      "loss": 0.5719,
+      "step": 4350
+    },
+    {
+      "epoch": 6.56,
+      "grad_norm": 10.056818008422852,
+      "learning_rate": 9.34436090225564e-06,
+      "loss": 0.542,
+      "step": 4360
+    },
+    {
+      "epoch": 6.57,
+      "grad_norm": 7.325047492980957,
+      "learning_rate": 9.342857142857144e-06,
+      "loss": 0.5564,
+      "step": 4370
+    },
+    {
+      "epoch": 6.59,
+      "grad_norm": 8.13595199584961,
+      "learning_rate": 9.341353383458647e-06,
+      "loss": 0.5106,
+      "step": 4380
+    },
+    {
+      "epoch": 6.6,
+      "grad_norm": 7.225549221038818,
+      "learning_rate": 9.33984962406015e-06,
+      "loss": 0.5714,
+      "step": 4390
+    },
+    {
+      "epoch": 6.62,
+      "grad_norm": 9.103632926940918,
+      "learning_rate": 9.338345864661656e-06,
+      "loss": 0.559,
+      "step": 4400
+    },
+    {
+      "epoch": 6.63,
+      "grad_norm": 5.579386234283447,
+      "learning_rate": 9.336842105263158e-06,
+      "loss": 0.5789,
+      "step": 4410
+    },
+    {
+      "epoch": 6.65,
+      "grad_norm": 9.875541687011719,
+      "learning_rate": 9.335338345864663e-06,
+      "loss": 0.5435,
+      "step": 4420
+    },
+    {
+      "epoch": 6.66,
+      "grad_norm": 10.33945083618164,
+      "learning_rate": 9.333834586466166e-06,
+      "loss": 0.4678,
+      "step": 4430
+    },
+    {
+      "epoch": 6.68,
+      "grad_norm": 8.180964469909668,
+      "learning_rate": 9.33233082706767e-06,
+      "loss": 0.5476,
+      "step": 4440
+    },
+    {
+      "epoch": 6.69,
+      "grad_norm": 8.327938079833984,
+      "learning_rate": 9.330827067669174e-06,
+      "loss": 0.5605,
+      "step": 4450
+    },
+    {
+      "epoch": 6.71,
+      "grad_norm": 6.9362311363220215,
+      "learning_rate": 9.329323308270679e-06,
+      "loss": 0.6099,
+      "step": 4460
+    },
+    {
+      "epoch": 6.72,
+      "grad_norm": 4.627447128295898,
+      "learning_rate": 9.32781954887218e-06,
+      "loss": 0.4521,
+      "step": 4470
+    },
+    {
+      "epoch": 6.74,
+      "grad_norm": 5.713562488555908,
+      "learning_rate": 9.326315789473684e-06,
+      "loss": 0.4532,
+      "step": 4480
+    },
+    {
+      "epoch": 6.75,
+      "grad_norm": 6.893897533416748,
+      "learning_rate": 9.32481203007519e-06,
+      "loss": 0.5298,
+      "step": 4490
+    },
+    {
+      "epoch": 6.77,
+      "grad_norm": 6.630578994750977,
+      "learning_rate": 9.323308270676693e-06,
+      "loss": 0.5226,
+      "step": 4500
+    },
+    {
+      "epoch": 6.78,
+      "grad_norm": 7.725119113922119,
+      "learning_rate": 9.321804511278196e-06,
+      "loss": 0.4581,
+      "step": 4510
+    },
+    {
+      "epoch": 6.8,
+      "grad_norm": 8.243720054626465,
+      "learning_rate": 9.3203007518797e-06,
+      "loss": 0.4821,
+      "step": 4520
+    },
+    {
+      "epoch": 6.81,
+      "grad_norm": 7.254865646362305,
+      "learning_rate": 9.318796992481203e-06,
+      "loss": 0.4623,
+      "step": 4530
+    },
+    {
+      "epoch": 6.83,
+      "grad_norm": 6.510406494140625,
+      "learning_rate": 9.317293233082707e-06,
+      "loss": 0.5073,
+      "step": 4540
+    },
+    {
+      "epoch": 6.84,
+      "grad_norm": 5.531012058258057,
+      "learning_rate": 9.315789473684212e-06,
+      "loss": 0.4842,
+      "step": 4550
+    },
+    {
+      "epoch": 6.86,
+      "grad_norm": 9.50185489654541,
+      "learning_rate": 9.314285714285714e-06,
+      "loss": 0.5485,
+      "step": 4560
+    },
+    {
+      "epoch": 6.87,
+      "grad_norm": 5.591551780700684,
+      "learning_rate": 9.312781954887219e-06,
+      "loss": 0.5397,
+      "step": 4570
+    },
+    {
+      "epoch": 6.89,
+      "grad_norm": 3.7708208560943604,
+      "learning_rate": 9.311278195488723e-06,
+      "loss": 0.4746,
+      "step": 4580
+    },
+    {
+      "epoch": 6.9,
+      "grad_norm": 5.826446533203125,
+      "learning_rate": 9.309774436090226e-06,
+      "loss": 0.503,
+      "step": 4590
+    },
+    {
+      "epoch": 6.92,
+      "grad_norm": 14.129280090332031,
+      "learning_rate": 9.30827067669173e-06,
+      "loss": 0.5729,
+      "step": 4600
+    },
+    {
+      "epoch": 6.93,
+      "grad_norm": 5.19706392288208,
+      "learning_rate": 9.306766917293233e-06,
+      "loss": 0.4967,
+      "step": 4610
+    },
+    {
+      "epoch": 6.95,
+      "grad_norm": 6.513811111450195,
+      "learning_rate": 9.305263157894737e-06,
+      "loss": 0.587,
+      "step": 4620
+    },
+    {
+      "epoch": 6.96,
+      "grad_norm": 7.2199506759643555,
+      "learning_rate": 9.303759398496242e-06,
+      "loss": 0.5733,
+      "step": 4630
+    },
+    {
+      "epoch": 6.98,
+      "grad_norm": 6.173489570617676,
+      "learning_rate": 9.302255639097745e-06,
+      "loss": 0.6391,
+      "step": 4640
+    },
+    {
+      "epoch": 6.99,
+      "grad_norm": 4.977587699890137,
+      "learning_rate": 9.300751879699249e-06,
+      "loss": 0.5506,
+      "step": 4650
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.9302,
+      "eval_loss": 0.2616922855377197,
+      "eval_runtime": 85.0513,
+      "eval_samples_per_second": 117.576,
+      "eval_steps_per_second": 0.47,
+      "step": 4655
+    },
+    {
+      "epoch": 7.01,
+      "grad_norm": 7.2303266525268555,
+      "learning_rate": 9.299248120300753e-06,
+      "loss": 0.4889,
+      "step": 4660
+    },
+    {
+      "epoch": 7.02,
+      "grad_norm": 5.8191914558410645,
+      "learning_rate": 9.297744360902256e-06,
+      "loss": 0.6557,
+      "step": 4670
+    },
+    {
+      "epoch": 7.04,
+      "grad_norm": 7.453029632568359,
+      "learning_rate": 9.29624060150376e-06,
+      "loss": 0.5975,
+      "step": 4680
+    },
+    {
+      "epoch": 7.05,
+      "grad_norm": 9.434555053710938,
+      "learning_rate": 9.294736842105265e-06,
+      "loss": 0.4769,
+      "step": 4690
+    },
+    {
+      "epoch": 7.07,
+      "grad_norm": 9.096846580505371,
+      "learning_rate": 9.293233082706767e-06,
+      "loss": 0.5521,
+      "step": 4700
+    },
+    {
+      "epoch": 7.08,
+      "grad_norm": 6.146598815917969,
+      "learning_rate": 9.291729323308272e-06,
+      "loss": 0.5191,
+      "step": 4710
+    },
+    {
+      "epoch": 7.1,
+      "grad_norm": 5.138683319091797,
+      "learning_rate": 9.290225563909775e-06,
+      "loss": 0.4883,
+      "step": 4720
+    },
+    {
+      "epoch": 7.11,
+      "grad_norm": 5.840444564819336,
+      "learning_rate": 9.288721804511279e-06,
+      "loss": 0.4901,
+      "step": 4730
+    },
+    {
+      "epoch": 7.13,
+      "grad_norm": 5.589585781097412,
+      "learning_rate": 9.287218045112782e-06,
+      "loss": 0.4574,
+      "step": 4740
+    },
+    {
+      "epoch": 7.14,
+      "grad_norm": 7.447097301483154,
+      "learning_rate": 9.285714285714288e-06,
+      "loss": 0.4672,
+      "step": 4750
+    },
+    {
+      "epoch": 7.16,
+      "grad_norm": 6.8820295333862305,
+      "learning_rate": 9.28421052631579e-06,
+      "loss": 0.5641,
+      "step": 4760
+    },
+    {
+      "epoch": 7.17,
+      "grad_norm": 5.46907901763916,
+      "learning_rate": 9.282706766917295e-06,
+      "loss": 0.4948,
+      "step": 4770
+    },
+    {
+      "epoch": 7.19,
+      "grad_norm": 5.4233527183532715,
+      "learning_rate": 9.281203007518798e-06,
+      "loss": 0.5507,
+      "step": 4780
+    },
+    {
+      "epoch": 7.2,
+      "grad_norm": 6.316089153289795,
+      "learning_rate": 9.279699248120302e-06,
+      "loss": 0.5791,
+      "step": 4790
+    },
+    {
+      "epoch": 7.22,
+      "grad_norm": 3.7618801593780518,
+      "learning_rate": 9.278195488721805e-06,
+      "loss": 0.4846,
+      "step": 4800
+    },
+    {
+      "epoch": 7.23,
+      "grad_norm": 6.426711082458496,
+      "learning_rate": 9.276691729323309e-06,
+      "loss": 0.4886,
+      "step": 4810
+    },
+    {
+      "epoch": 7.25,
+      "grad_norm": 6.98826265335083,
+      "learning_rate": 9.275187969924812e-06,
+      "loss": 0.4627,
+      "step": 4820
+    },
+    {
+      "epoch": 7.26,
+      "grad_norm": 6.147061824798584,
+      "learning_rate": 9.273684210526317e-06,
+      "loss": 0.4577,
+      "step": 4830
+    },
+    {
+      "epoch": 7.28,
+      "grad_norm": 7.308942794799805,
+      "learning_rate": 9.272180451127821e-06,
+      "loss": 0.4893,
+      "step": 4840
+    },
+    {
+      "epoch": 7.29,
+      "grad_norm": 8.406046867370605,
+      "learning_rate": 9.270676691729324e-06,
+      "loss": 0.4968,
+      "step": 4850
+    },
+    {
+      "epoch": 7.31,
+      "grad_norm": 4.631737232208252,
+      "learning_rate": 9.269172932330828e-06,
+      "loss": 0.4654,
+      "step": 4860
+    },
+    {
+      "epoch": 7.32,
+      "grad_norm": 3.802255868911743,
+      "learning_rate": 9.267669172932331e-06,
+      "loss": 0.5018,
+      "step": 4870
+    },
+    {
+      "epoch": 7.34,
+      "grad_norm": 7.958065986633301,
+      "learning_rate": 9.266165413533835e-06,
+      "loss": 0.5358,
+      "step": 4880
+    },
+    {
+      "epoch": 7.35,
+      "grad_norm": 4.825588703155518,
+      "learning_rate": 9.26466165413534e-06,
+      "loss": 0.5201,
+      "step": 4890
+    },
+    {
+      "epoch": 7.37,
+      "grad_norm": 4.964457035064697,
+      "learning_rate": 9.263157894736842e-06,
+      "loss": 0.4819,
+      "step": 4900
+    },
+    {
+      "epoch": 7.38,
+      "grad_norm": 11.642394065856934,
+      "learning_rate": 9.261654135338347e-06,
+      "loss": 0.5989,
+      "step": 4910
+    },
+    {
+      "epoch": 7.4,
+      "grad_norm": 9.31828498840332,
+      "learning_rate": 9.26015037593985e-06,
+      "loss": 0.5455,
+      "step": 4920
+    },
+    {
+      "epoch": 7.41,
+      "grad_norm": 7.8129963874816895,
+      "learning_rate": 9.258646616541354e-06,
+      "loss": 0.5542,
+      "step": 4930
+    },
+    {
+      "epoch": 7.43,
+      "grad_norm": 7.043788909912109,
+      "learning_rate": 9.257142857142858e-06,
+      "loss": 0.5153,
+      "step": 4940
+    },
+    {
+      "epoch": 7.44,
+      "grad_norm": 8.311758995056152,
+      "learning_rate": 9.255639097744363e-06,
+      "loss": 0.4802,
+      "step": 4950
+    },
+    {
+      "epoch": 7.46,
+      "grad_norm": 10.970717430114746,
+      "learning_rate": 9.254135338345865e-06,
+      "loss": 0.4495,
+      "step": 4960
+    },
+    {
+      "epoch": 7.47,
+      "grad_norm": 5.547107219696045,
+      "learning_rate": 9.252631578947368e-06,
+      "loss": 0.4824,
+      "step": 4970
+    },
+    {
+      "epoch": 7.49,
+      "grad_norm": 5.658668518066406,
+      "learning_rate": 9.251127819548874e-06,
+      "loss": 0.5722,
+      "step": 4980
+    },
+    {
+      "epoch": 7.5,
+      "grad_norm": 4.896615982055664,
+      "learning_rate": 9.249624060150375e-06,
+      "loss": 0.4936,
+      "step": 4990
+    },
+    {
+      "epoch": 7.52,
+      "grad_norm": 7.777392864227295,
+      "learning_rate": 9.24812030075188e-06,
+      "loss": 0.5078,
+      "step": 5000
+    },
+    {
+      "epoch": 7.53,
+      "grad_norm": 11.22333812713623,
+      "learning_rate": 9.246616541353384e-06,
+      "loss": 0.575,
+      "step": 5010
+    },
+    {
+      "epoch": 7.55,
+      "grad_norm": 6.031052589416504,
+      "learning_rate": 9.245112781954888e-06,
+      "loss": 0.4742,
+      "step": 5020
+    },
+    {
+      "epoch": 7.56,
+      "grad_norm": 11.427336692810059,
+      "learning_rate": 9.243609022556391e-06,
+      "loss": 0.5519,
+      "step": 5030
+    },
+    {
+      "epoch": 7.58,
+      "grad_norm": 6.76407527923584,
+      "learning_rate": 9.242105263157896e-06,
+      "loss": 0.5202,
+      "step": 5040
+    },
+    {
+      "epoch": 7.59,
+      "grad_norm": 7.091256618499756,
+      "learning_rate": 9.240601503759398e-06,
+      "loss": 0.5255,
+      "step": 5050
+    },
+    {
+      "epoch": 7.61,
+      "grad_norm": 6.818326473236084,
+      "learning_rate": 9.239097744360903e-06,
+      "loss": 0.442,
+      "step": 5060
+    },
+    {
+      "epoch": 7.62,
+      "grad_norm": 7.494906902313232,
+      "learning_rate": 9.237593984962407e-06,
+      "loss": 0.5126,
+      "step": 5070
+    },
+    {
+      "epoch": 7.64,
+      "grad_norm": 5.982577800750732,
+      "learning_rate": 9.23609022556391e-06,
+      "loss": 0.5056,
+      "step": 5080
+    },
+    {
+      "epoch": 7.65,
+      "grad_norm": 4.815781116485596,
+      "learning_rate": 9.234586466165414e-06,
+      "loss": 0.4896,
+      "step": 5090
+    },
+    {
+      "epoch": 7.67,
+      "grad_norm": 10.674721717834473,
+      "learning_rate": 9.233082706766918e-06,
+      "loss": 0.5209,
+      "step": 5100
+    },
+    {
+      "epoch": 7.68,
+      "grad_norm": 5.937568187713623,
+      "learning_rate": 9.231578947368421e-06,
+      "loss": 0.4772,
+      "step": 5110
+    },
+    {
+      "epoch": 7.7,
+      "grad_norm": 5.146367073059082,
+      "learning_rate": 9.230075187969926e-06,
+      "loss": 0.5032,
+      "step": 5120
+    },
+    {
+      "epoch": 7.71,
+      "grad_norm": 8.03272819519043,
+      "learning_rate": 9.22857142857143e-06,
+      "loss": 0.5072,
+      "step": 5130
+    },
+    {
+      "epoch": 7.73,
+      "grad_norm": 5.361180782318115,
+      "learning_rate": 9.227067669172933e-06,
+      "loss": 0.5057,
+      "step": 5140
+    },
+    {
+      "epoch": 7.74,
+      "grad_norm": 5.487973213195801,
+      "learning_rate": 9.225563909774437e-06,
+      "loss": 0.5253,
+      "step": 5150
+    },
+    {
+      "epoch": 7.76,
+      "grad_norm": 6.845251560211182,
+      "learning_rate": 9.22406015037594e-06,
+      "loss": 0.5436,
+      "step": 5160
+    },
+    {
+      "epoch": 7.77,
+      "grad_norm": 4.931974411010742,
+      "learning_rate": 9.222556390977444e-06,
+      "loss": 0.4227,
+      "step": 5170
+    },
+    {
+      "epoch": 7.79,
+      "grad_norm": 7.382147312164307,
+      "learning_rate": 9.221052631578949e-06,
+      "loss": 0.5022,
+      "step": 5180
+    },
+    {
+      "epoch": 7.8,
+      "grad_norm": 8.380685806274414,
+      "learning_rate": 9.219548872180451e-06,
+      "loss": 0.5259,
+      "step": 5190
+    },
+    {
+      "epoch": 7.82,
+      "grad_norm": 6.625802993774414,
+      "learning_rate": 9.218045112781956e-06,
+      "loss": 0.5213,
+      "step": 5200
+    },
+    {
+      "epoch": 7.83,
+      "grad_norm": 6.7276692390441895,
+      "learning_rate": 9.21654135338346e-06,
+      "loss": 0.4831,
+      "step": 5210
+    },
+    {
+      "epoch": 7.85,
+      "grad_norm": 5.930064678192139,
+      "learning_rate": 9.215037593984963e-06,
+      "loss": 0.4278,
+      "step": 5220
+    },
+    {
+      "epoch": 7.86,
+      "grad_norm": 5.958808422088623,
+      "learning_rate": 9.213533834586467e-06,
+      "loss": 0.5039,
+      "step": 5230
+    },
+    {
+      "epoch": 7.88,
+      "grad_norm": 8.592114448547363,
+      "learning_rate": 9.21203007518797e-06,
+      "loss": 0.48,
+      "step": 5240
+    },
+    {
+      "epoch": 7.89,
+      "grad_norm": 7.286666393280029,
+      "learning_rate": 9.210526315789474e-06,
+      "loss": 0.5513,
+      "step": 5250
+    },
+    {
+      "epoch": 7.91,
+      "grad_norm": 3.8542234897613525,
+      "learning_rate": 9.209022556390979e-06,
+      "loss": 0.4452,
+      "step": 5260
+    },
+    {
+      "epoch": 7.92,
+      "grad_norm": 8.812358856201172,
+      "learning_rate": 9.207518796992482e-06,
+      "loss": 0.4796,
+      "step": 5270
+    },
+    {
+      "epoch": 7.94,
+      "grad_norm": 7.729457378387451,
+      "learning_rate": 9.206015037593986e-06,
+      "loss": 0.5283,
+      "step": 5280
+    },
+    {
+      "epoch": 7.95,
+      "grad_norm": 5.370766639709473,
+      "learning_rate": 9.20451127819549e-06,
+      "loss": 0.557,
+      "step": 5290
+    },
+    {
+      "epoch": 7.97,
+      "grad_norm": 7.00390625,
+      "learning_rate": 9.203007518796993e-06,
+      "loss": 0.5266,
+      "step": 5300
+    },
+    {
+      "epoch": 7.98,
+      "grad_norm": 5.945902347564697,
+      "learning_rate": 9.201503759398496e-06,
+      "loss": 0.505,
+      "step": 5310
+    },
+    {
+      "epoch": 8.0,
+      "grad_norm": 39.0330924987793,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 0.5673,
+      "step": 5320
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.9324,
+      "eval_loss": 0.2575855553150177,
+      "eval_runtime": 85.1441,
+      "eval_samples_per_second": 117.448,
+      "eval_steps_per_second": 0.47,
+      "step": 5320
+    },
+    {
+      "epoch": 8.02,
+      "grad_norm": 7.23183012008667,
+      "learning_rate": 9.198496240601504e-06,
+      "loss": 0.4209,
+      "step": 5330
+    },
+    {
+      "epoch": 8.03,
+      "grad_norm": 5.357458591461182,
+      "learning_rate": 9.196992481203009e-06,
+      "loss": 0.4701,
+      "step": 5340
+    },
+    {
+      "epoch": 8.05,
+      "grad_norm": 9.471860885620117,
+      "learning_rate": 9.195488721804512e-06,
+      "loss": 0.4924,
+      "step": 5350
+    },
+    {
+      "epoch": 8.06,
+      "grad_norm": 7.7437214851379395,
+      "learning_rate": 9.193984962406016e-06,
+      "loss": 0.4997,
+      "step": 5360
+    },
+    {
+      "epoch": 8.08,
+      "grad_norm": 6.697991371154785,
+      "learning_rate": 9.19248120300752e-06,
+      "loss": 0.4365,
+      "step": 5370
+    },
+    {
+      "epoch": 8.09,
+      "grad_norm": 9.78630542755127,
+      "learning_rate": 9.190977443609025e-06,
+      "loss": 0.5572,
+      "step": 5380
+    },
+    {
+      "epoch": 8.11,
+      "grad_norm": 6.197582244873047,
+      "learning_rate": 9.189473684210526e-06,
+      "loss": 0.4581,
+      "step": 5390
+    },
+    {
+      "epoch": 8.12,
+      "grad_norm": 5.565506458282471,
+      "learning_rate": 9.187969924812032e-06,
+      "loss": 0.5219,
+      "step": 5400
+    },
+    {
+      "epoch": 8.14,
+      "grad_norm": 5.3856353759765625,
+      "learning_rate": 9.186466165413535e-06,
+      "loss": 0.5744,
+      "step": 5410
+    },
+    {
+      "epoch": 8.15,
+      "grad_norm": 6.119697093963623,
+      "learning_rate": 9.184962406015039e-06,
+      "loss": 0.3964,
+      "step": 5420
+    },
+    {
+      "epoch": 8.17,
+      "grad_norm": 5.344573497772217,
+      "learning_rate": 9.183458646616542e-06,
+      "loss": 0.4816,
+      "step": 5430
+    },
+    {
+      "epoch": 8.18,
+      "grad_norm": 5.624027729034424,
+      "learning_rate": 9.181954887218046e-06,
+      "loss": 0.4341,
+      "step": 5440
+    },
+    {
+      "epoch": 8.2,
+      "grad_norm": 4.847854137420654,
+      "learning_rate": 9.180451127819549e-06,
+      "loss": 0.5471,
+      "step": 5450
+    },
+    {
+      "epoch": 8.21,
+      "grad_norm": 6.051102638244629,
+      "learning_rate": 9.178947368421053e-06,
+      "loss": 0.5593,
+      "step": 5460
+    },
+    {
+      "epoch": 8.23,
+      "grad_norm": 8.345075607299805,
+      "learning_rate": 9.177443609022558e-06,
+      "loss": 0.4132,
+      "step": 5470
+    },
+    {
+      "epoch": 8.24,
+      "grad_norm": 5.369195938110352,
+      "learning_rate": 9.17593984962406e-06,
+      "loss": 0.4844,
+      "step": 5480
+    },
+    {
+      "epoch": 8.26,
+      "grad_norm": 6.0165228843688965,
+      "learning_rate": 9.174436090225565e-06,
+      "loss": 0.4712,
+      "step": 5490
+    },
+    {
+      "epoch": 8.27,
+      "grad_norm": 4.54939079284668,
+      "learning_rate": 9.172932330827068e-06,
+      "loss": 0.4956,
+      "step": 5500
+    },
+    {
+      "epoch": 8.29,
+      "grad_norm": 5.021441459655762,
+      "learning_rate": 9.171428571428572e-06,
+      "loss": 0.4428,
+      "step": 5510
+    },
+    {
+      "epoch": 8.3,
+      "grad_norm": 7.0452446937561035,
+      "learning_rate": 9.169924812030075e-06,
+      "loss": 0.5036,
+      "step": 5520
+    },
+    {
+      "epoch": 8.32,
+      "grad_norm": 8.039512634277344,
+      "learning_rate": 9.168421052631579e-06,
+      "loss": 0.478,
+      "step": 5530
+    },
+    {
+      "epoch": 8.33,
+      "grad_norm": 6.608015537261963,
+      "learning_rate": 9.166917293233083e-06,
+      "loss": 0.4338,
+      "step": 5540
+    },
+    {
+      "epoch": 8.35,
+      "grad_norm": 6.659971714019775,
+      "learning_rate": 9.165413533834588e-06,
+      "loss": 0.5033,
+      "step": 5550
+    },
+    {
+      "epoch": 8.36,
+      "grad_norm": 5.663388252258301,
+      "learning_rate": 9.163909774436091e-06,
+      "loss": 0.4779,
+      "step": 5560
+    },
+    {
+      "epoch": 8.38,
+      "grad_norm": 5.978389739990234,
+      "learning_rate": 9.162406015037595e-06,
+      "loss": 0.4615,
+      "step": 5570
+    },
+    {
+      "epoch": 8.39,
+      "grad_norm": 5.943080425262451,
+      "learning_rate": 9.160902255639098e-06,
+      "loss": 0.4282,
+      "step": 5580
+    },
+    {
+      "epoch": 8.41,
+      "grad_norm": 3.245058536529541,
+      "learning_rate": 9.159398496240602e-06,
+      "loss": 0.4423,
+      "step": 5590
+    },
+    {
+      "epoch": 8.42,
+      "grad_norm": 5.73254919052124,
+      "learning_rate": 9.157894736842105e-06,
+      "loss": 0.4777,
+      "step": 5600
+    },
+    {
+      "epoch": 8.44,
+      "grad_norm": 6.487976551055908,
+      "learning_rate": 9.15639097744361e-06,
+      "loss": 0.5132,
+      "step": 5610
+    },
+    {
+      "epoch": 8.45,
+      "grad_norm": 4.1268463134765625,
+      "learning_rate": 9.154887218045112e-06,
+      "loss": 0.3942,
+      "step": 5620
+    },
+    {
+      "epoch": 8.47,
+      "grad_norm": 7.593535900115967,
+      "learning_rate": 9.153383458646618e-06,
+      "loss": 0.5109,
+      "step": 5630
+    },
+    {
+      "epoch": 8.48,
+      "grad_norm": 4.127936840057373,
+      "learning_rate": 9.151879699248121e-06,
+      "loss": 0.4554,
+      "step": 5640
+    },
+    {
+      "epoch": 8.5,
+      "grad_norm": 12.721508026123047,
+      "learning_rate": 9.150375939849625e-06,
+      "loss": 0.5204,
+      "step": 5650
+    },
+    {
+      "epoch": 8.51,
+      "grad_norm": 7.618612289428711,
+      "learning_rate": 9.148872180451128e-06,
+      "loss": 0.5095,
+      "step": 5660
+    },
+    {
+      "epoch": 8.53,
+      "grad_norm": 5.089692115783691,
+      "learning_rate": 9.147368421052633e-06,
+      "loss": 0.478,
+      "step": 5670
+    },
+    {
+      "epoch": 8.54,
+      "grad_norm": 6.890159606933594,
+      "learning_rate": 9.145864661654135e-06,
+      "loss": 0.4634,
+      "step": 5680
+    },
+    {
+      "epoch": 8.56,
+      "grad_norm": 9.275102615356445,
+      "learning_rate": 9.14436090225564e-06,
+      "loss": 0.528,
+      "step": 5690
+    },
+    {
+      "epoch": 8.57,
+      "grad_norm": 4.839653015136719,
+      "learning_rate": 9.142857142857144e-06,
+      "loss": 0.4558,
+      "step": 5700
+    },
+    {
+      "epoch": 8.59,
+      "grad_norm": 7.7605791091918945,
+      "learning_rate": 9.141353383458647e-06,
+      "loss": 0.5086,
+      "step": 5710
+    },
+    {
+      "epoch": 8.6,
+      "grad_norm": 6.753016948699951,
+      "learning_rate": 9.139849624060151e-06,
+      "loss": 0.4953,
+      "step": 5720
+    },
+    {
+      "epoch": 8.62,
+      "grad_norm": 10.297369003295898,
+      "learning_rate": 9.138345864661654e-06,
+      "loss": 0.454,
+      "step": 5730
+    },
+    {
+      "epoch": 8.63,
+      "grad_norm": 5.704922676086426,
+      "learning_rate": 9.136842105263158e-06,
+      "loss": 0.4504,
+      "step": 5740
+    },
+    {
+      "epoch": 8.65,
+      "grad_norm": 5.4977030754089355,
+      "learning_rate": 9.135338345864663e-06,
+      "loss": 0.5203,
+      "step": 5750
+    },
+    {
+      "epoch": 8.66,
+      "grad_norm": 7.531189918518066,
+      "learning_rate": 9.133834586466167e-06,
+      "loss": 0.4534,
+      "step": 5760
+    },
+    {
+      "epoch": 8.68,
+      "grad_norm": 6.660569667816162,
+      "learning_rate": 9.13233082706767e-06,
+      "loss": 0.577,
+      "step": 5770
+    },
+    {
+      "epoch": 8.69,
+      "grad_norm": 6.752721309661865,
+      "learning_rate": 9.130827067669174e-06,
+      "loss": 0.4574,
+      "step": 5780
+    },
+    {
+      "epoch": 8.71,
+      "grad_norm": 7.526786804199219,
+      "learning_rate": 9.129323308270677e-06,
+      "loss": 0.4511,
+      "step": 5790
+    },
+    {
+      "epoch": 8.72,
+      "grad_norm": 6.5963239669799805,
+      "learning_rate": 9.12781954887218e-06,
+      "loss": 0.4917,
+      "step": 5800
+    },
+    {
+      "epoch": 8.74,
+      "grad_norm": 4.203681468963623,
+      "learning_rate": 9.126315789473686e-06,
+      "loss": 0.5218,
+      "step": 5810
+    },
+    {
+      "epoch": 8.75,
+      "grad_norm": 4.367255210876465,
+      "learning_rate": 9.124812030075188e-06,
+      "loss": 0.4162,
+      "step": 5820
+    },
+    {
+      "epoch": 8.77,
+      "grad_norm": 3.8670730590820312,
+      "learning_rate": 9.123308270676693e-06,
+      "loss": 0.3913,
+      "step": 5830
+    },
+    {
+      "epoch": 8.78,
+      "grad_norm": 9.634267807006836,
+      "learning_rate": 9.121804511278197e-06,
+      "loss": 0.5062,
+      "step": 5840
+    },
+    {
+      "epoch": 8.8,
+      "grad_norm": 2.509295701980591,
+      "learning_rate": 9.1203007518797e-06,
+      "loss": 0.4422,
+      "step": 5850
+    },
+    {
+      "epoch": 8.81,
+      "grad_norm": 7.6244659423828125,
+      "learning_rate": 9.118796992481204e-06,
+      "loss": 0.4566,
+      "step": 5860
+    },
+    {
+      "epoch": 8.83,
+      "grad_norm": 6.837118625640869,
+      "learning_rate": 9.117293233082709e-06,
+      "loss": 0.5095,
+      "step": 5870
+    },
+    {
+      "epoch": 8.84,
+      "grad_norm": 4.819979667663574,
+      "learning_rate": 9.11578947368421e-06,
+      "loss": 0.4471,
+      "step": 5880
+    },
+    {
+      "epoch": 8.86,
+      "grad_norm": 6.350512504577637,
+      "learning_rate": 9.114285714285716e-06,
+      "loss": 0.4751,
+      "step": 5890
+    },
+    {
+      "epoch": 8.87,
+      "grad_norm": 3.4793074131011963,
+      "learning_rate": 9.11278195488722e-06,
+      "loss": 0.4089,
+      "step": 5900
+    },
+    {
+      "epoch": 8.89,
+      "grad_norm": 5.1062774658203125,
+      "learning_rate": 9.111278195488723e-06,
+      "loss": 0.5624,
+      "step": 5910
+    },
+    {
+      "epoch": 8.9,
+      "grad_norm": 8.126543998718262,
+      "learning_rate": 9.109774436090226e-06,
+      "loss": 0.5146,
+      "step": 5920
+    },
+    {
+      "epoch": 8.92,
+      "grad_norm": 7.661808967590332,
+      "learning_rate": 9.10827067669173e-06,
+      "loss": 0.4602,
+      "step": 5930
+    },
+    {
+      "epoch": 8.93,
+      "grad_norm": 6.820888996124268,
+      "learning_rate": 9.106766917293233e-06,
+      "loss": 0.4827,
+      "step": 5940
+    },
+    {
+      "epoch": 8.95,
+      "grad_norm": 24.169485092163086,
+      "learning_rate": 9.105263157894739e-06,
+      "loss": 0.5302,
+      "step": 5950
+    },
+    {
+      "epoch": 8.96,
+      "grad_norm": 5.068043231964111,
+      "learning_rate": 9.103759398496242e-06,
+      "loss": 0.5453,
+      "step": 5960
+    },
+    {
+      "epoch": 8.98,
+      "grad_norm": 5.819450378417969,
+      "learning_rate": 9.102255639097744e-06,
+      "loss": 0.4811,
+      "step": 5970
+    },
+    {
+      "epoch": 8.99,
+      "grad_norm": 4.129781723022461,
+      "learning_rate": 9.10075187969925e-06,
+      "loss": 0.4613,
+      "step": 5980
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9311,
+      "eval_loss": 0.25862327218055725,
+      "eval_runtime": 84.8621,
+      "eval_samples_per_second": 117.838,
+      "eval_steps_per_second": 0.471,
+      "step": 5985
+    },
+    {
+      "epoch": 9.01,
+      "grad_norm": 6.844424247741699,
+      "learning_rate": 9.099248120300753e-06,
+      "loss": 0.4154,
+      "step": 5990
+    },
+    {
+      "epoch": 9.02,
+      "grad_norm": 5.792689323425293,
+      "learning_rate": 9.097744360902256e-06,
+      "loss": 0.4345,
+      "step": 6000
+    },
+    {
+      "epoch": 9.04,
+      "grad_norm": 5.300471305847168,
+      "learning_rate": 9.09624060150376e-06,
+      "loss": 0.3986,
+      "step": 6010
+    },
+    {
+      "epoch": 9.05,
+      "grad_norm": 11.9564208984375,
+      "learning_rate": 9.094736842105263e-06,
+      "loss": 0.4255,
+      "step": 6020
+    },
+    {
+      "epoch": 9.07,
+      "grad_norm": 6.798022270202637,
+      "learning_rate": 9.093233082706767e-06,
+      "loss": 0.4296,
+      "step": 6030
+    },
+    {
+      "epoch": 9.08,
+      "grad_norm": 7.78212308883667,
+      "learning_rate": 9.091729323308272e-06,
+      "loss": 0.4373,
+      "step": 6040
+    },
+    {
+      "epoch": 9.1,
+      "grad_norm": 6.719583988189697,
+      "learning_rate": 9.090225563909776e-06,
+      "loss": 0.4943,
+      "step": 6050
+    },
+    {
+      "epoch": 9.11,
+      "grad_norm": 8.298834800720215,
+      "learning_rate": 9.088721804511279e-06,
+      "loss": 0.3937,
+      "step": 6060
+    },
+    {
+      "epoch": 9.13,
+      "grad_norm": 4.731727600097656,
+      "learning_rate": 9.087218045112783e-06,
+      "loss": 0.4711,
+      "step": 6070
+    },
+    {
+      "epoch": 9.14,
+      "grad_norm": 6.207810878753662,
+      "learning_rate": 9.085714285714286e-06,
+      "loss": 0.4565,
+      "step": 6080
+    },
+    {
+      "epoch": 9.16,
+      "grad_norm": 6.939966678619385,
+      "learning_rate": 9.08421052631579e-06,
+      "loss": 0.5276,
+      "step": 6090
+    },
+    {
+      "epoch": 9.17,
+      "grad_norm": 9.00831127166748,
+      "learning_rate": 9.082706766917295e-06,
+      "loss": 0.5558,
+      "step": 6100
+    },
+    {
+      "epoch": 9.19,
+      "grad_norm": 4.730199813842773,
+      "learning_rate": 9.081203007518797e-06,
+      "loss": 0.4826,
+      "step": 6110
+    },
+    {
+      "epoch": 9.2,
+      "grad_norm": 4.198337078094482,
+      "learning_rate": 9.079699248120302e-06,
+      "loss": 0.3978,
+      "step": 6120
+    },
+    {
+      "epoch": 9.22,
+      "grad_norm": 5.722704887390137,
+      "learning_rate": 9.078195488721805e-06,
+      "loss": 0.4622,
+      "step": 6130
+    },
+    {
+      "epoch": 9.23,
+      "grad_norm": 8.497228622436523,
+      "learning_rate": 9.076691729323309e-06,
+      "loss": 0.4726,
+      "step": 6140
+    },
+    {
+      "epoch": 9.25,
+      "grad_norm": 7.046009063720703,
+      "learning_rate": 9.075187969924812e-06,
+      "loss": 0.527,
+      "step": 6150
+    },
+    {
+      "epoch": 9.26,
+      "grad_norm": 7.972896099090576,
+      "learning_rate": 9.073684210526316e-06,
+      "loss": 0.3072,
+      "step": 6160
+    },
+    {
+      "epoch": 9.28,
+      "grad_norm": 8.850788116455078,
+      "learning_rate": 9.07218045112782e-06,
+      "loss": 0.4753,
+      "step": 6170
+    },
+    {
+      "epoch": 9.29,
+      "grad_norm": 6.064061641693115,
+      "learning_rate": 9.070676691729325e-06,
+      "loss": 0.4009,
+      "step": 6180
+    },
+    {
+      "epoch": 9.31,
+      "grad_norm": 6.12713098526001,
+      "learning_rate": 9.069172932330828e-06,
+      "loss": 0.4786,
+      "step": 6190
+    },
+    {
+      "epoch": 9.32,
+      "grad_norm": 8.643204689025879,
+      "learning_rate": 9.067669172932332e-06,
+      "loss": 0.5134,
+      "step": 6200
+    },
+    {
+      "epoch": 9.34,
+      "grad_norm": 7.257277488708496,
+      "learning_rate": 9.066165413533835e-06,
+      "loss": 0.4474,
+      "step": 6210
+    },
+    {
+      "epoch": 9.35,
+      "grad_norm": 7.62333869934082,
+      "learning_rate": 9.064661654135339e-06,
+      "loss": 0.4579,
+      "step": 6220
+    },
+    {
+      "epoch": 9.37,
+      "grad_norm": 9.346735954284668,
+      "learning_rate": 9.063157894736842e-06,
+      "loss": 0.4714,
+      "step": 6230
+    },
+    {
+      "epoch": 9.38,
+      "grad_norm": 3.8007750511169434,
+      "learning_rate": 9.061654135338347e-06,
+      "loss": 0.4334,
+      "step": 6240
+    },
+    {
+      "epoch": 9.4,
+      "grad_norm": 6.266302108764648,
+      "learning_rate": 9.06015037593985e-06,
+      "loss": 0.4704,
+      "step": 6250
+    },
+    {
+      "epoch": 9.41,
+      "grad_norm": 6.959786891937256,
+      "learning_rate": 9.058646616541355e-06,
+      "loss": 0.5353,
+      "step": 6260
+    },
+    {
+      "epoch": 9.43,
+      "grad_norm": 6.572616100311279,
+      "learning_rate": 9.057142857142858e-06,
+      "loss": 0.4629,
+      "step": 6270
+    },
+    {
+      "epoch": 9.44,
+      "grad_norm": 5.961916446685791,
+      "learning_rate": 9.055639097744362e-06,
+      "loss": 0.5119,
+      "step": 6280
+    },
+    {
+      "epoch": 9.46,
+      "grad_norm": 6.547915935516357,
+      "learning_rate": 9.054135338345865e-06,
+      "loss": 0.4576,
+      "step": 6290
+    },
+    {
+      "epoch": 9.47,
+      "grad_norm": 6.359402179718018,
+      "learning_rate": 9.05263157894737e-06,
+      "loss": 0.3692,
+      "step": 6300
+    },
+    {
+      "epoch": 9.49,
+      "grad_norm": 7.048614501953125,
+      "learning_rate": 9.051127819548872e-06,
+      "loss": 0.3911,
+      "step": 6310
+    },
+    {
+      "epoch": 9.5,
+      "grad_norm": 5.198198318481445,
+      "learning_rate": 9.049624060150377e-06,
+      "loss": 0.475,
+      "step": 6320
+    },
+    {
+      "epoch": 9.52,
+      "grad_norm": 5.797221660614014,
+      "learning_rate": 9.04812030075188e-06,
+      "loss": 0.3771,
+      "step": 6330
+    },
+    {
+      "epoch": 9.53,
+      "grad_norm": 5.751585483551025,
+      "learning_rate": 9.046616541353384e-06,
+      "loss": 0.497,
+      "step": 6340
+    },
+    {
+      "epoch": 9.55,
+      "grad_norm": 9.54306697845459,
+      "learning_rate": 9.045112781954888e-06,
+      "loss": 0.3976,
+      "step": 6350
+    },
+    {
+      "epoch": 9.56,
+      "grad_norm": 8.968032836914062,
+      "learning_rate": 9.043609022556391e-06,
+      "loss": 0.4252,
+      "step": 6360
+    },
+    {
+      "epoch": 9.58,
+      "grad_norm": 13.218304634094238,
+      "learning_rate": 9.042105263157895e-06,
+      "loss": 0.5191,
+      "step": 6370
+    },
+    {
+      "epoch": 9.59,
+      "grad_norm": 4.405686855316162,
+      "learning_rate": 9.0406015037594e-06,
+      "loss": 0.4149,
+      "step": 6380
+    },
+    {
+      "epoch": 9.61,
+      "grad_norm": 4.863158702850342,
+      "learning_rate": 9.039097744360904e-06,
+      "loss": 0.4875,
+      "step": 6390
+    },
+    {
+      "epoch": 9.62,
+      "grad_norm": 6.247385501861572,
+      "learning_rate": 9.037593984962407e-06,
+      "loss": 0.4106,
+      "step": 6400
+    },
+    {
+      "epoch": 9.64,
+      "grad_norm": 6.554888725280762,
+      "learning_rate": 9.03609022556391e-06,
+      "loss": 0.4645,
+      "step": 6410
+    },
+    {
+      "epoch": 9.65,
+      "grad_norm": 7.249465465545654,
+      "learning_rate": 9.034586466165414e-06,
+      "loss": 0.4291,
+      "step": 6420
+    },
+    {
+      "epoch": 9.67,
+      "grad_norm": 6.810882568359375,
+      "learning_rate": 9.033082706766918e-06,
+      "loss": 0.4637,
+      "step": 6430
+    },
+    {
+      "epoch": 9.68,
+      "grad_norm": 2.65733003616333,
+      "learning_rate": 9.031578947368423e-06,
+      "loss": 0.4585,
+      "step": 6440
+    },
+    {
+      "epoch": 9.7,
+      "grad_norm": 8.63343334197998,
+      "learning_rate": 9.030075187969925e-06,
+      "loss": 0.5187,
+      "step": 6450
+    },
+    {
+      "epoch": 9.71,
+      "grad_norm": 5.558303356170654,
+      "learning_rate": 9.028571428571428e-06,
+      "loss": 0.4819,
+      "step": 6460
+    },
+    {
+      "epoch": 9.73,
+      "grad_norm": 8.711833000183105,
+      "learning_rate": 9.027067669172933e-06,
+      "loss": 0.4585,
+      "step": 6470
+    },
+    {
+      "epoch": 9.74,
+      "grad_norm": 7.882017135620117,
+      "learning_rate": 9.025563909774437e-06,
+      "loss": 0.5154,
+      "step": 6480
+    },
+    {
+      "epoch": 9.76,
+      "grad_norm": 5.849830150604248,
+      "learning_rate": 9.02406015037594e-06,
+      "loss": 0.5093,
+      "step": 6490
+    },
+    {
+      "epoch": 9.77,
+      "grad_norm": 7.546263217926025,
+      "learning_rate": 9.022556390977444e-06,
+      "loss": 0.3969,
+      "step": 6500
+    },
+    {
+      "epoch": 9.79,
+      "grad_norm": 7.114614963531494,
+      "learning_rate": 9.021052631578948e-06,
+      "loss": 0.4485,
+      "step": 6510
+    },
+    {
+      "epoch": 9.8,
+      "grad_norm": 7.6169209480285645,
+      "learning_rate": 9.019548872180451e-06,
+      "loss": 0.425,
+      "step": 6520
+    },
+    {
+      "epoch": 9.82,
+      "grad_norm": 5.7843403816223145,
+      "learning_rate": 9.018045112781956e-06,
+      "loss": 0.4171,
+      "step": 6530
+    },
+    {
+      "epoch": 9.83,
+      "grad_norm": 6.0503082275390625,
+      "learning_rate": 9.016541353383458e-06,
+      "loss": 0.483,
+      "step": 6540
+    },
+    {
+      "epoch": 9.85,
+      "grad_norm": 7.677584648132324,
+      "learning_rate": 9.015037593984963e-06,
+      "loss": 0.4747,
+      "step": 6550
+    },
+    {
+      "epoch": 9.86,
+      "grad_norm": 5.793139934539795,
+      "learning_rate": 9.013533834586467e-06,
+      "loss": 0.3621,
+      "step": 6560
+    },
+    {
+      "epoch": 9.88,
+      "grad_norm": 6.399969577789307,
+      "learning_rate": 9.01203007518797e-06,
+      "loss": 0.4373,
+      "step": 6570
+    },
+    {
+      "epoch": 9.89,
+      "grad_norm": 10.296338081359863,
+      "learning_rate": 9.010526315789474e-06,
+      "loss": 0.417,
+      "step": 6580
+    },
+    {
+      "epoch": 9.91,
+      "grad_norm": 6.193917274475098,
+      "learning_rate": 9.009022556390979e-06,
+      "loss": 0.4419,
+      "step": 6590
+    },
+    {
+      "epoch": 9.92,
+      "grad_norm": 3.921016216278076,
+      "learning_rate": 9.007518796992481e-06,
+      "loss": 0.3981,
+      "step": 6600
+    },
+    {
+      "epoch": 9.94,
+      "grad_norm": 6.30132532119751,
+      "learning_rate": 9.006015037593986e-06,
+      "loss": 0.4699,
+      "step": 6610
+    },
+    {
+      "epoch": 9.95,
+      "grad_norm": 8.901771545410156,
+      "learning_rate": 9.00451127819549e-06,
+      "loss": 0.4308,
+      "step": 6620
+    },
+    {
+      "epoch": 9.97,
+      "grad_norm": 5.031552314758301,
+      "learning_rate": 9.003007518796993e-06,
+      "loss": 0.48,
+      "step": 6630
+    },
+    {
+      "epoch": 9.98,
+      "grad_norm": 5.636510372161865,
+      "learning_rate": 9.001503759398497e-06,
+      "loss": 0.403,
+      "step": 6640
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 3.5294342041015625,
+      "learning_rate": 9e-06,
+      "loss": 0.4179,
+      "step": 6650
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.9285,
+      "eval_loss": 0.2555387318134308,
+      "eval_runtime": 84.7099,
+      "eval_samples_per_second": 118.05,
+      "eval_steps_per_second": 0.472,
+      "step": 6650
+    },
+    {
+      "epoch": 10.02,
+      "grad_norm": 6.522907257080078,
+      "learning_rate": 8.998496240601504e-06,
+      "loss": 0.4525,
+      "step": 6660
+    },
+    {
+      "epoch": 10.03,
+      "grad_norm": 6.142210006713867,
+      "learning_rate": 8.996992481203009e-06,
+      "loss": 0.3998,
+      "step": 6670
+    },
+    {
+      "epoch": 10.05,
+      "grad_norm": 7.781100749969482,
+      "learning_rate": 8.995488721804512e-06,
+      "loss": 0.4122,
+      "step": 6680
+    },
+    {
+      "epoch": 10.06,
+      "grad_norm": 5.448252201080322,
+      "learning_rate": 8.993984962406016e-06,
+      "loss": 0.461,
+      "step": 6690
+    },
+    {
+      "epoch": 10.08,
+      "grad_norm": 7.063671588897705,
+      "learning_rate": 8.99248120300752e-06,
+      "loss": 0.432,
+      "step": 6700
+    },
+    {
+      "epoch": 10.09,
+      "grad_norm": 6.696626663208008,
+      "learning_rate": 8.990977443609023e-06,
+      "loss": 0.4208,
+      "step": 6710
+    },
+    {
+      "epoch": 10.11,
+      "grad_norm": 6.5666656494140625,
+      "learning_rate": 8.989473684210527e-06,
+      "loss": 0.4527,
+      "step": 6720
+    },
+    {
+      "epoch": 10.12,
+      "grad_norm": 8.801324844360352,
+      "learning_rate": 8.987969924812032e-06,
+      "loss": 0.442,
+      "step": 6730
+    },
+    {
+      "epoch": 10.14,
+      "grad_norm": 6.743152141571045,
+      "learning_rate": 8.986466165413534e-06,
+      "loss": 0.4424,
+      "step": 6740
+    },
+    {
+      "epoch": 10.15,
+      "grad_norm": 5.408703327178955,
+      "learning_rate": 8.984962406015039e-06,
+      "loss": 0.4548,
+      "step": 6750
+    },
+    {
+      "epoch": 10.17,
+      "grad_norm": 8.466784477233887,
+      "learning_rate": 8.983458646616542e-06,
+      "loss": 0.4355,
+      "step": 6760
+    },
+    {
+      "epoch": 10.18,
+      "grad_norm": 5.309767723083496,
+      "learning_rate": 8.981954887218046e-06,
+      "loss": 0.4023,
+      "step": 6770
+    },
+    {
+      "epoch": 10.2,
+      "grad_norm": 3.3604421615600586,
+      "learning_rate": 8.98045112781955e-06,
+      "loss": 0.4367,
+      "step": 6780
+    },
+    {
+      "epoch": 10.21,
+      "grad_norm": 6.275347709655762,
+      "learning_rate": 8.978947368421055e-06,
+      "loss": 0.4764,
+      "step": 6790
+    },
+    {
+      "epoch": 10.23,
+      "grad_norm": 6.770579814910889,
+      "learning_rate": 8.977443609022556e-06,
+      "loss": 0.4414,
+      "step": 6800
+    },
+    {
+      "epoch": 10.24,
+      "grad_norm": 8.56733512878418,
+      "learning_rate": 8.975939849624062e-06,
+      "loss": 0.4426,
+      "step": 6810
+    },
+    {
+      "epoch": 10.26,
+      "grad_norm": 6.006712436676025,
+      "learning_rate": 8.974436090225565e-06,
+      "loss": 0.3702,
+      "step": 6820
+    },
+    {
+      "epoch": 10.27,
+      "grad_norm": 4.649052143096924,
+      "learning_rate": 8.972932330827069e-06,
+      "loss": 0.5371,
+      "step": 6830
+    },
+    {
+      "epoch": 10.29,
+      "grad_norm": 9.080769538879395,
+      "learning_rate": 8.971428571428572e-06,
+      "loss": 0.4866,
+      "step": 6840
+    },
+    {
+      "epoch": 10.3,
+      "grad_norm": 5.778624534606934,
+      "learning_rate": 8.969924812030076e-06,
+      "loss": 0.381,
+      "step": 6850
+    },
+    {
+      "epoch": 10.32,
+      "grad_norm": 7.814187049865723,
+      "learning_rate": 8.96842105263158e-06,
+      "loss": 0.5162,
+      "step": 6860
+    },
+    {
+      "epoch": 10.33,
+      "grad_norm": 5.049838542938232,
+      "learning_rate": 8.966917293233084e-06,
+      "loss": 0.4879,
+      "step": 6870
+    },
+    {
+      "epoch": 10.35,
+      "grad_norm": 8.096096992492676,
+      "learning_rate": 8.965413533834588e-06,
+      "loss": 0.4726,
+      "step": 6880
+    },
+    {
+      "epoch": 10.36,
+      "grad_norm": 7.028320789337158,
+      "learning_rate": 8.963909774436091e-06,
+      "loss": 0.4424,
+      "step": 6890
+    },
+    {
+      "epoch": 10.38,
+      "grad_norm": 4.826821804046631,
+      "learning_rate": 8.962406015037595e-06,
+      "loss": 0.4552,
+      "step": 6900
+    },
+    {
+      "epoch": 10.39,
+      "grad_norm": 8.392495155334473,
+      "learning_rate": 8.960902255639098e-06,
+      "loss": 0.4378,
+      "step": 6910
+    },
+    {
+      "epoch": 10.41,
+      "grad_norm": 4.868290424346924,
+      "learning_rate": 8.959398496240602e-06,
+      "loss": 0.4151,
+      "step": 6920
+    },
+    {
+      "epoch": 10.42,
+      "grad_norm": 6.117234230041504,
+      "learning_rate": 8.957894736842107e-06,
+      "loss": 0.5149,
+      "step": 6930
+    },
+    {
+      "epoch": 10.44,
+      "grad_norm": 9.33238697052002,
+      "learning_rate": 8.956390977443609e-06,
+      "loss": 0.3984,
+      "step": 6940
+    },
+    {
+      "epoch": 10.45,
+      "grad_norm": 9.559886932373047,
+      "learning_rate": 8.954887218045113e-06,
+      "loss": 0.4171,
+      "step": 6950
+    },
+    {
+      "epoch": 10.47,
+      "grad_norm": 4.344634056091309,
+      "learning_rate": 8.953383458646618e-06,
+      "loss": 0.4419,
+      "step": 6960
+    },
+    {
+      "epoch": 10.48,
+      "grad_norm": 5.508487701416016,
+      "learning_rate": 8.951879699248121e-06,
+      "loss": 0.4605,
+      "step": 6970
+    },
+    {
+      "epoch": 10.5,
+      "grad_norm": 5.529686450958252,
+      "learning_rate": 8.950375939849625e-06,
+      "loss": 0.4004,
+      "step": 6980
+    },
+    {
+      "epoch": 10.51,
+      "grad_norm": 5.424170970916748,
+      "learning_rate": 8.948872180451128e-06,
+      "loss": 0.4351,
+      "step": 6990
+    },
+    {
+      "epoch": 10.53,
+      "grad_norm": 6.121506690979004,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 0.4198,
+      "step": 7000
+    },
+    {
+      "epoch": 10.54,
+      "grad_norm": 4.664872169494629,
+      "learning_rate": 8.945864661654135e-06,
+      "loss": 0.3917,
+      "step": 7010
+    },
+    {
+      "epoch": 10.56,
+      "grad_norm": 5.378602027893066,
+      "learning_rate": 8.94436090225564e-06,
+      "loss": 0.47,
+      "step": 7020
+    },
+    {
+      "epoch": 10.57,
+      "grad_norm": 8.281057357788086,
+      "learning_rate": 8.942857142857142e-06,
+      "loss": 0.3779,
+      "step": 7030
+    },
+    {
+      "epoch": 10.59,
+      "grad_norm": 5.378328800201416,
+      "learning_rate": 8.941353383458648e-06,
+      "loss": 0.4878,
+      "step": 7040
+    },
+    {
+      "epoch": 10.6,
+      "grad_norm": 4.809008598327637,
+      "learning_rate": 8.939849624060151e-06,
+      "loss": 0.4409,
+      "step": 7050
+    },
+    {
+      "epoch": 10.62,
+      "grad_norm": 6.703794002532959,
+      "learning_rate": 8.938345864661655e-06,
+      "loss": 0.45,
+      "step": 7060
+    },
+    {
+      "epoch": 10.63,
+      "grad_norm": 10.097111701965332,
+      "learning_rate": 8.936842105263158e-06,
+      "loss": 0.4442,
+      "step": 7070
+    },
+    {
+      "epoch": 10.65,
+      "grad_norm": 5.404522895812988,
+      "learning_rate": 8.935338345864662e-06,
+      "loss": 0.4289,
+      "step": 7080
+    },
+    {
+      "epoch": 10.66,
+      "grad_norm": 2.983161449432373,
+      "learning_rate": 8.933834586466165e-06,
+      "loss": 0.4072,
+      "step": 7090
+    },
+    {
+      "epoch": 10.68,
+      "grad_norm": 6.501340389251709,
+      "learning_rate": 8.93233082706767e-06,
+      "loss": 0.4344,
+      "step": 7100
+    },
+    {
+      "epoch": 10.69,
+      "grad_norm": 7.439212322235107,
+      "learning_rate": 8.930827067669174e-06,
+      "loss": 0.4208,
+      "step": 7110
+    },
+    {
+      "epoch": 10.71,
+      "grad_norm": 6.9180192947387695,
+      "learning_rate": 8.929323308270677e-06,
+      "loss": 0.4901,
+      "step": 7120
+    },
+    {
+      "epoch": 10.72,
+      "grad_norm": 4.9598212242126465,
+      "learning_rate": 8.927819548872181e-06,
+      "loss": 0.4551,
+      "step": 7130
+    },
+    {
+      "epoch": 10.74,
+      "grad_norm": 7.020519256591797,
+      "learning_rate": 8.926315789473685e-06,
+      "loss": 0.4469,
+      "step": 7140
+    },
+    {
+      "epoch": 10.75,
+      "grad_norm": 6.747496604919434,
+      "learning_rate": 8.924812030075188e-06,
+      "loss": 0.3626,
+      "step": 7150
+    },
+    {
+      "epoch": 10.77,
+      "grad_norm": 2.869495153427124,
+      "learning_rate": 8.923308270676693e-06,
+      "loss": 0.3794,
+      "step": 7160
+    },
+    {
+      "epoch": 10.78,
+      "grad_norm": 7.156761169433594,
+      "learning_rate": 8.921804511278195e-06,
+      "loss": 0.4909,
+      "step": 7170
+    },
+    {
+      "epoch": 10.8,
+      "grad_norm": 9.461006164550781,
+      "learning_rate": 8.9203007518797e-06,
+      "loss": 0.4487,
+      "step": 7180
+    },
+    {
+      "epoch": 10.81,
+      "grad_norm": 5.75421142578125,
+      "learning_rate": 8.918796992481204e-06,
+      "loss": 0.4953,
+      "step": 7190
+    },
+    {
+      "epoch": 10.83,
+      "grad_norm": 4.186371326446533,
+      "learning_rate": 8.917293233082707e-06,
+      "loss": 0.3788,
+      "step": 7200
+    },
+    {
+      "epoch": 10.84,
+      "grad_norm": 6.402685165405273,
+      "learning_rate": 8.915789473684211e-06,
+      "loss": 0.5009,
+      "step": 7210
+    },
+    {
+      "epoch": 10.86,
+      "grad_norm": 10.709757804870605,
+      "learning_rate": 8.914285714285716e-06,
+      "loss": 0.5308,
+      "step": 7220
+    },
+    {
+      "epoch": 10.87,
+      "grad_norm": 8.926152229309082,
+      "learning_rate": 8.912781954887218e-06,
+      "loss": 0.4461,
+      "step": 7230
+    },
+    {
+      "epoch": 10.89,
+      "grad_norm": 6.41901969909668,
+      "learning_rate": 8.911278195488723e-06,
+      "loss": 0.4188,
+      "step": 7240
+    },
+    {
+      "epoch": 10.9,
+      "grad_norm": 4.931794166564941,
+      "learning_rate": 8.909774436090227e-06,
+      "loss": 0.4089,
+      "step": 7250
+    },
+    {
+      "epoch": 10.92,
+      "grad_norm": 7.75593376159668,
+      "learning_rate": 8.90827067669173e-06,
+      "loss": 0.5261,
+      "step": 7260
+    },
+    {
+      "epoch": 10.93,
+      "grad_norm": 9.013036727905273,
+      "learning_rate": 8.906766917293234e-06,
+      "loss": 0.3274,
+      "step": 7270
+    },
+    {
+      "epoch": 10.95,
+      "grad_norm": 6.653579235076904,
+      "learning_rate": 8.905263157894737e-06,
+      "loss": 0.4646,
+      "step": 7280
+    },
+    {
+      "epoch": 10.96,
+      "grad_norm": 5.304203987121582,
+      "learning_rate": 8.90375939849624e-06,
+      "loss": 0.4399,
+      "step": 7290
+    },
+    {
+      "epoch": 10.98,
+      "grad_norm": 40.1646842956543,
+      "learning_rate": 8.902255639097746e-06,
+      "loss": 0.3007,
+      "step": 7300
+    },
+    {
+      "epoch": 10.99,
+      "grad_norm": 5.538785934448242,
+      "learning_rate": 8.90075187969925e-06,
+      "loss": 0.4438,
+      "step": 7310
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9316,
+      "eval_loss": 0.25541195273399353,
+      "eval_runtime": 84.8107,
+      "eval_samples_per_second": 117.91,
+      "eval_steps_per_second": 0.472,
+      "step": 7315
+    },
+    {
+      "epoch": 11.01,
+      "grad_norm": 8.498943328857422,
+      "learning_rate": 8.899248120300753e-06,
+      "loss": 0.4009,
+      "step": 7320
+    },
+    {
+      "epoch": 11.02,
+      "grad_norm": 6.2147040367126465,
+      "learning_rate": 8.897744360902256e-06,
+      "loss": 0.4283,
+      "step": 7330
+    },
+    {
+      "epoch": 11.04,
+      "grad_norm": 5.028774261474609,
+      "learning_rate": 8.89624060150376e-06,
+      "loss": 0.4145,
+      "step": 7340
+    },
+    {
+      "epoch": 11.05,
+      "grad_norm": 7.040588855743408,
+      "learning_rate": 8.894736842105264e-06,
+      "loss": 0.3753,
+      "step": 7350
+    },
+    {
+      "epoch": 11.07,
+      "grad_norm": 4.658559322357178,
+      "learning_rate": 8.893233082706769e-06,
+      "loss": 0.4139,
+      "step": 7360
+    },
+    {
+      "epoch": 11.08,
+      "grad_norm": 7.867548942565918,
+      "learning_rate": 8.89172932330827e-06,
+      "loss": 0.4854,
+      "step": 7370
+    },
+    {
+      "epoch": 11.1,
+      "grad_norm": 8.354945182800293,
+      "learning_rate": 8.890225563909776e-06,
+      "loss": 0.4186,
+      "step": 7380
+    },
+    {
+      "epoch": 11.11,
+      "grad_norm": 6.198273658752441,
+      "learning_rate": 8.88872180451128e-06,
+      "loss": 0.4486,
+      "step": 7390
+    },
+    {
+      "epoch": 11.13,
+      "grad_norm": 7.401607990264893,
+      "learning_rate": 8.887218045112783e-06,
+      "loss": 0.4015,
+      "step": 7400
+    },
+    {
+      "epoch": 11.14,
+      "grad_norm": 5.412950038909912,
+      "learning_rate": 8.885714285714286e-06,
+      "loss": 0.3654,
+      "step": 7410
+    },
+    {
+      "epoch": 11.16,
+      "grad_norm": 3.7357654571533203,
+      "learning_rate": 8.884210526315792e-06,
+      "loss": 0.4165,
+      "step": 7420
+    },
+    {
+      "epoch": 11.17,
+      "grad_norm": 7.468185901641846,
+      "learning_rate": 8.882706766917293e-06,
+      "loss": 0.3241,
+      "step": 7430
+    },
+    {
+      "epoch": 11.19,
+      "grad_norm": 5.967494487762451,
+      "learning_rate": 8.881203007518799e-06,
+      "loss": 0.4364,
+      "step": 7440
+    },
+    {
+      "epoch": 11.2,
+      "grad_norm": 8.94781494140625,
+      "learning_rate": 8.879699248120302e-06,
+      "loss": 0.4927,
+      "step": 7450
+    },
+    {
+      "epoch": 11.22,
+      "grad_norm": 7.6195969581604,
+      "learning_rate": 8.878195488721804e-06,
+      "loss": 0.3722,
+      "step": 7460
+    },
+    {
+      "epoch": 11.23,
+      "grad_norm": 9.522473335266113,
+      "learning_rate": 8.876691729323309e-06,
+      "loss": 0.3974,
+      "step": 7470
+    },
+    {
+      "epoch": 11.25,
+      "grad_norm": 9.590860366821289,
+      "learning_rate": 8.875187969924813e-06,
+      "loss": 0.3842,
+      "step": 7480
+    },
+    {
+      "epoch": 11.26,
+      "grad_norm": 6.479350566864014,
+      "learning_rate": 8.873684210526316e-06,
+      "loss": 0.4135,
+      "step": 7490
+    },
+    {
+      "epoch": 11.28,
+      "grad_norm": 8.100231170654297,
+      "learning_rate": 8.87218045112782e-06,
+      "loss": 0.4262,
+      "step": 7500
+    },
+    {
+      "epoch": 11.29,
+      "grad_norm": 9.401702880859375,
+      "learning_rate": 8.870676691729325e-06,
+      "loss": 0.3899,
+      "step": 7510
+    },
+    {
+      "epoch": 11.31,
+      "grad_norm": 7.8885626792907715,
+      "learning_rate": 8.869172932330827e-06,
+      "loss": 0.4738,
+      "step": 7520
+    },
+    {
+      "epoch": 11.32,
+      "grad_norm": 7.2377753257751465,
+      "learning_rate": 8.867669172932332e-06,
+      "loss": 0.3686,
+      "step": 7530
+    },
+    {
+      "epoch": 11.34,
+      "grad_norm": 5.0235209465026855,
+      "learning_rate": 8.866165413533835e-06,
+      "loss": 0.3939,
+      "step": 7540
+    },
+    {
+      "epoch": 11.35,
+      "grad_norm": 6.832250595092773,
+      "learning_rate": 8.864661654135339e-06,
+      "loss": 0.4485,
+      "step": 7550
+    },
+    {
+      "epoch": 11.37,
+      "grad_norm": 8.186062812805176,
+      "learning_rate": 8.863157894736842e-06,
+      "loss": 0.4242,
+      "step": 7560
+    },
+    {
+      "epoch": 11.38,
+      "grad_norm": 5.467780113220215,
+      "learning_rate": 8.861654135338346e-06,
+      "loss": 0.4599,
+      "step": 7570
+    },
+    {
+      "epoch": 11.4,
+      "grad_norm": 6.155720233917236,
+      "learning_rate": 8.86015037593985e-06,
+      "loss": 0.5285,
+      "step": 7580
+    },
+    {
+      "epoch": 11.41,
+      "grad_norm": 6.44677734375,
+      "learning_rate": 8.858646616541355e-06,
+      "loss": 0.4931,
+      "step": 7590
+    },
+    {
+      "epoch": 11.43,
+      "grad_norm": 15.308818817138672,
+      "learning_rate": 8.857142857142858e-06,
+      "loss": 0.3899,
+      "step": 7600
+    },
+    {
+      "epoch": 11.44,
+      "grad_norm": 6.691050052642822,
+      "learning_rate": 8.855639097744362e-06,
+      "loss": 0.4313,
+      "step": 7610
+    },
+    {
+      "epoch": 11.46,
+      "grad_norm": 5.215397357940674,
+      "learning_rate": 8.854135338345865e-06,
+      "loss": 0.3528,
+      "step": 7620
+    },
+    {
+      "epoch": 11.47,
+      "grad_norm": 7.355811595916748,
+      "learning_rate": 8.852631578947369e-06,
+      "loss": 0.4402,
+      "step": 7630
+    },
+    {
+      "epoch": 11.49,
+      "grad_norm": 4.864825248718262,
+      "learning_rate": 8.851127819548872e-06,
+      "loss": 0.3485,
+      "step": 7640
+    },
+    {
+      "epoch": 11.5,
+      "grad_norm": 7.4907755851745605,
+      "learning_rate": 8.849624060150378e-06,
+      "loss": 0.4522,
+      "step": 7650
+    },
+    {
+      "epoch": 11.52,
+      "grad_norm": 6.480433464050293,
+      "learning_rate": 8.84812030075188e-06,
+      "loss": 0.4655,
+      "step": 7660
+    },
+    {
+      "epoch": 11.53,
+      "grad_norm": 5.072092056274414,
+      "learning_rate": 8.846616541353385e-06,
+      "loss": 0.3735,
+      "step": 7670
+    },
+    {
+      "epoch": 11.55,
+      "grad_norm": 10.207109451293945,
+      "learning_rate": 8.845112781954888e-06,
+      "loss": 0.3884,
+      "step": 7680
+    },
+    {
+      "epoch": 11.56,
+      "grad_norm": 5.795559883117676,
+      "learning_rate": 8.843609022556392e-06,
+      "loss": 0.4115,
+      "step": 7690
+    },
+    {
+      "epoch": 11.58,
+      "grad_norm": 7.781355381011963,
+      "learning_rate": 8.842105263157895e-06,
+      "loss": 0.4617,
+      "step": 7700
+    },
+    {
+      "epoch": 11.59,
+      "grad_norm": 6.770030975341797,
+      "learning_rate": 8.8406015037594e-06,
+      "loss": 0.4218,
+      "step": 7710
+    },
+    {
+      "epoch": 11.61,
+      "grad_norm": 7.052707672119141,
+      "learning_rate": 8.839097744360902e-06,
+      "loss": 0.3897,
+      "step": 7720
+    },
+    {
+      "epoch": 11.62,
+      "grad_norm": 8.882899284362793,
+      "learning_rate": 8.837593984962407e-06,
+      "loss": 0.4546,
+      "step": 7730
+    },
+    {
+      "epoch": 11.64,
+      "grad_norm": 7.858944892883301,
+      "learning_rate": 8.836090225563911e-06,
+      "loss": 0.4287,
+      "step": 7740
+    },
+    {
+      "epoch": 11.65,
+      "grad_norm": 6.74614953994751,
+      "learning_rate": 8.834586466165414e-06,
+      "loss": 0.3326,
+      "step": 7750
+    },
+    {
+      "epoch": 11.67,
+      "grad_norm": 8.970141410827637,
+      "learning_rate": 8.833082706766918e-06,
+      "loss": 0.4863,
+      "step": 7760
+    },
+    {
+      "epoch": 11.68,
+      "grad_norm": 6.568352699279785,
+      "learning_rate": 8.831578947368421e-06,
+      "loss": 0.4248,
+      "step": 7770
+    },
+    {
+      "epoch": 11.7,
+      "grad_norm": 6.05830717086792,
+      "learning_rate": 8.830075187969925e-06,
+      "loss": 0.4829,
+      "step": 7780
+    },
+    {
+      "epoch": 11.71,
+      "grad_norm": 7.666469097137451,
+      "learning_rate": 8.82857142857143e-06,
+      "loss": 0.5319,
+      "step": 7790
+    },
+    {
+      "epoch": 11.73,
+      "grad_norm": 5.955508708953857,
+      "learning_rate": 8.827067669172934e-06,
+      "loss": 0.4309,
+      "step": 7800
+    },
+    {
+      "epoch": 11.74,
+      "grad_norm": 6.9883270263671875,
+      "learning_rate": 8.825563909774437e-06,
+      "loss": 0.4157,
+      "step": 7810
+    },
+    {
+      "epoch": 11.76,
+      "grad_norm": 6.703571319580078,
+      "learning_rate": 8.82406015037594e-06,
+      "loss": 0.3219,
+      "step": 7820
+    },
+    {
+      "epoch": 11.77,
+      "grad_norm": 7.131542682647705,
+      "learning_rate": 8.822556390977444e-06,
+      "loss": 0.4492,
+      "step": 7830
+    },
+    {
+      "epoch": 11.79,
+      "grad_norm": 5.014946460723877,
+      "learning_rate": 8.821052631578948e-06,
+      "loss": 0.4189,
+      "step": 7840
+    },
+    {
+      "epoch": 11.8,
+      "grad_norm": 4.254874229431152,
+      "learning_rate": 8.819548872180453e-06,
+      "loss": 0.484,
+      "step": 7850
+    },
+    {
+      "epoch": 11.82,
+      "grad_norm": 4.319407939910889,
+      "learning_rate": 8.818045112781955e-06,
+      "loss": 0.3861,
+      "step": 7860
+    },
+    {
+      "epoch": 11.83,
+      "grad_norm": 7.9686408042907715,
+      "learning_rate": 8.81654135338346e-06,
+      "loss": 0.4264,
+      "step": 7870
+    },
+    {
+      "epoch": 11.85,
+      "grad_norm": 5.5855326652526855,
+      "learning_rate": 8.815037593984964e-06,
+      "loss": 0.4532,
+      "step": 7880
+    },
+    {
+      "epoch": 11.86,
+      "grad_norm": 6.914451599121094,
+      "learning_rate": 8.813533834586467e-06,
+      "loss": 0.4355,
+      "step": 7890
+    },
+    {
+      "epoch": 11.88,
+      "grad_norm": 7.542539596557617,
+      "learning_rate": 8.81203007518797e-06,
+      "loss": 0.3671,
+      "step": 7900
+    },
+    {
+      "epoch": 11.89,
+      "grad_norm": 7.947263717651367,
+      "learning_rate": 8.810526315789474e-06,
+      "loss": 0.373,
+      "step": 7910
+    },
+    {
+      "epoch": 11.91,
+      "grad_norm": 7.884321689605713,
+      "learning_rate": 8.809022556390978e-06,
+      "loss": 0.4827,
+      "step": 7920
+    },
+    {
+      "epoch": 11.92,
+      "grad_norm": 5.361155986785889,
+      "learning_rate": 8.807518796992483e-06,
+      "loss": 0.4485,
+      "step": 7930
+    },
+    {
+      "epoch": 11.94,
+      "grad_norm": 7.507490158081055,
+      "learning_rate": 8.806015037593986e-06,
+      "loss": 0.4446,
+      "step": 7940
+    },
+    {
+      "epoch": 11.95,
+      "grad_norm": 7.053649425506592,
+      "learning_rate": 8.804511278195488e-06,
+      "loss": 0.4112,
+      "step": 7950
+    },
+    {
+      "epoch": 11.97,
+      "grad_norm": 8.394134521484375,
+      "learning_rate": 8.803007518796993e-06,
+      "loss": 0.4221,
+      "step": 7960
+    },
+    {
+      "epoch": 11.98,
+      "grad_norm": 9.852388381958008,
+      "learning_rate": 8.801503759398497e-06,
+      "loss": 0.4178,
+      "step": 7970
+    },
+    {
+      "epoch": 12.0,
+      "grad_norm": 17.5406551361084,
+      "learning_rate": 8.8e-06,
+      "loss": 0.4869,
+      "step": 7980
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.9298,
+      "eval_loss": 0.2563527822494507,
+      "eval_runtime": 84.9156,
+      "eval_samples_per_second": 117.764,
+      "eval_steps_per_second": 0.471,
+      "step": 7980
+    },
+    {
+      "epoch": 12.02,
+      "grad_norm": 4.651547908782959,
+      "learning_rate": 8.798496240601504e-06,
+      "loss": 0.4307,
+      "step": 7990
+    },
+    {
+      "epoch": 12.03,
+      "grad_norm": 5.47507905960083,
+      "learning_rate": 8.796992481203007e-06,
+      "loss": 0.3957,
+      "step": 8000
+    },
+    {
+      "epoch": 12.05,
+      "grad_norm": 7.309173583984375,
+      "learning_rate": 8.795488721804511e-06,
+      "loss": 0.3848,
+      "step": 8010
+    },
+    {
+      "epoch": 12.06,
+      "grad_norm": 5.4073591232299805,
+      "learning_rate": 8.793984962406016e-06,
+      "loss": 0.347,
+      "step": 8020
+    },
+    {
+      "epoch": 12.08,
+      "grad_norm": 9.495542526245117,
+      "learning_rate": 8.79248120300752e-06,
+      "loss": 0.3963,
+      "step": 8030
+    },
+    {
+      "epoch": 12.09,
+      "grad_norm": 7.175304412841797,
+      "learning_rate": 8.790977443609023e-06,
+      "loss": 0.4028,
+      "step": 8040
+    },
+    {
+      "epoch": 12.11,
+      "grad_norm": 5.7672624588012695,
+      "learning_rate": 8.789473684210527e-06,
+      "loss": 0.4336,
+      "step": 8050
+    },
+    {
+      "epoch": 12.12,
+      "grad_norm": 5.373271942138672,
+      "learning_rate": 8.78796992481203e-06,
+      "loss": 0.4214,
+      "step": 8060
+    },
+    {
+      "epoch": 12.14,
+      "grad_norm": 7.81503963470459,
+      "learning_rate": 8.786466165413534e-06,
+      "loss": 0.3362,
+      "step": 8070
+    },
+    {
+      "epoch": 12.15,
+      "grad_norm": 5.352240085601807,
+      "learning_rate": 8.784962406015039e-06,
+      "loss": 0.438,
+      "step": 8080
+    },
+    {
+      "epoch": 12.17,
+      "grad_norm": 4.825592994689941,
+      "learning_rate": 8.783458646616541e-06,
+      "loss": 0.3996,
+      "step": 8090
+    },
+    {
+      "epoch": 12.18,
+      "grad_norm": 4.875209808349609,
+      "learning_rate": 8.781954887218046e-06,
+      "loss": 0.4056,
+      "step": 8100
+    },
+    {
+      "epoch": 12.2,
+      "grad_norm": 6.405061721801758,
+      "learning_rate": 8.78045112781955e-06,
+      "loss": 0.404,
+      "step": 8110
+    },
+    {
+      "epoch": 12.21,
+      "grad_norm": 5.762337684631348,
+      "learning_rate": 8.778947368421053e-06,
+      "loss": 0.3609,
+      "step": 8120
+    },
+    {
+      "epoch": 12.23,
+      "grad_norm": 8.700191497802734,
+      "learning_rate": 8.777443609022557e-06,
+      "loss": 0.4316,
+      "step": 8130
+    },
+    {
+      "epoch": 12.24,
+      "grad_norm": 5.509273052215576,
+      "learning_rate": 8.775939849624062e-06,
+      "loss": 0.3814,
+      "step": 8140
+    },
+    {
+      "epoch": 12.26,
+      "grad_norm": 6.949098587036133,
+      "learning_rate": 8.774436090225564e-06,
+      "loss": 0.3588,
+      "step": 8150
+    },
+    {
+      "epoch": 12.27,
+      "grad_norm": 5.564908981323242,
+      "learning_rate": 8.772932330827069e-06,
+      "loss": 0.353,
+      "step": 8160
+    },
+    {
+      "epoch": 12.29,
+      "grad_norm": 6.935297012329102,
+      "learning_rate": 8.771428571428572e-06,
+      "loss": 0.4212,
+      "step": 8170
+    },
+    {
+      "epoch": 12.3,
+      "grad_norm": 4.811358451843262,
+      "learning_rate": 8.769924812030076e-06,
+      "loss": 0.3865,
+      "step": 8180
+    },
+    {
+      "epoch": 12.32,
+      "grad_norm": 6.4804368019104,
+      "learning_rate": 8.76842105263158e-06,
+      "loss": 0.3656,
+      "step": 8190
+    },
+    {
+      "epoch": 12.33,
+      "grad_norm": 3.9236013889312744,
+      "learning_rate": 8.766917293233083e-06,
+      "loss": 0.4885,
+      "step": 8200
+    },
+    {
+      "epoch": 12.35,
+      "grad_norm": 7.50891637802124,
+      "learning_rate": 8.765413533834586e-06,
+      "loss": 0.3962,
+      "step": 8210
+    },
+    {
+      "epoch": 12.36,
+      "grad_norm": 4.313982963562012,
+      "learning_rate": 8.763909774436092e-06,
+      "loss": 0.4023,
+      "step": 8220
+    },
+    {
+      "epoch": 12.38,
+      "grad_norm": 4.385167121887207,
+      "learning_rate": 8.762406015037595e-06,
+      "loss": 0.4841,
+      "step": 8230
+    },
+    {
+      "epoch": 12.39,
+      "grad_norm": 5.977277755737305,
+      "learning_rate": 8.760902255639099e-06,
+      "loss": 0.399,
+      "step": 8240
+    },
+    {
+      "epoch": 12.41,
+      "grad_norm": 8.858118057250977,
+      "learning_rate": 8.759398496240602e-06,
+      "loss": 0.451,
+      "step": 8250
+    },
+    {
+      "epoch": 12.42,
+      "grad_norm": 6.294662952423096,
+      "learning_rate": 8.757894736842106e-06,
+      "loss": 0.416,
+      "step": 8260
+    },
+    {
+      "epoch": 12.44,
+      "grad_norm": 4.536668300628662,
+      "learning_rate": 8.75639097744361e-06,
+      "loss": 0.407,
+      "step": 8270
+    },
+    {
+      "epoch": 12.45,
+      "grad_norm": 5.644812107086182,
+      "learning_rate": 8.754887218045114e-06,
+      "loss": 0.3685,
+      "step": 8280
+    },
+    {
+      "epoch": 12.47,
+      "grad_norm": 5.488842010498047,
+      "learning_rate": 8.753383458646616e-06,
+      "loss": 0.4136,
+      "step": 8290
+    },
+    {
+      "epoch": 12.48,
+      "grad_norm": 4.548142910003662,
+      "learning_rate": 8.751879699248122e-06,
+      "loss": 0.4502,
+      "step": 8300
+    },
+    {
+      "epoch": 12.5,
+      "grad_norm": 3.41457200050354,
+      "learning_rate": 8.750375939849625e-06,
+      "loss": 0.3598,
+      "step": 8310
+    },
+    {
+      "epoch": 12.51,
+      "grad_norm": 6.259812831878662,
+      "learning_rate": 8.748872180451129e-06,
+      "loss": 0.3843,
+      "step": 8320
+    },
+    {
+      "epoch": 12.53,
+      "grad_norm": 5.301551342010498,
+      "learning_rate": 8.747368421052632e-06,
+      "loss": 0.4038,
+      "step": 8330
+    },
+    {
+      "epoch": 12.54,
+      "grad_norm": 14.684255599975586,
+      "learning_rate": 8.745864661654137e-06,
+      "loss": 0.4115,
+      "step": 8340
+    },
+    {
+      "epoch": 12.56,
+      "grad_norm": 6.711531162261963,
+      "learning_rate": 8.744360902255639e-06,
+      "loss": 0.4112,
+      "step": 8350
+    },
+    {
+      "epoch": 12.57,
+      "grad_norm": 8.990388870239258,
+      "learning_rate": 8.742857142857144e-06,
+      "loss": 0.3547,
+      "step": 8360
+    },
+    {
+      "epoch": 12.59,
+      "grad_norm": 4.513948440551758,
+      "learning_rate": 8.741353383458648e-06,
+      "loss": 0.3776,
+      "step": 8370
+    },
+    {
+      "epoch": 12.6,
+      "grad_norm": 6.088433742523193,
+      "learning_rate": 8.739849624060151e-06,
+      "loss": 0.4116,
+      "step": 8380
+    },
+    {
+      "epoch": 12.62,
+      "grad_norm": 7.882970809936523,
+      "learning_rate": 8.738345864661655e-06,
+      "loss": 0.3712,
+      "step": 8390
+    },
+    {
+      "epoch": 12.63,
+      "grad_norm": 6.829627990722656,
+      "learning_rate": 8.736842105263158e-06,
+      "loss": 0.3842,
+      "step": 8400
+    },
+    {
+      "epoch": 12.65,
+      "grad_norm": 6.185722351074219,
+      "learning_rate": 8.735338345864662e-06,
+      "loss": 0.424,
+      "step": 8410
+    },
+    {
+      "epoch": 12.66,
+      "grad_norm": 4.945958137512207,
+      "learning_rate": 8.733834586466167e-06,
+      "loss": 0.377,
+      "step": 8420
+    },
+    {
+      "epoch": 12.68,
+      "grad_norm": 6.356648921966553,
+      "learning_rate": 8.73233082706767e-06,
+      "loss": 0.4256,
+      "step": 8430
+    },
+    {
+      "epoch": 12.69,
+      "grad_norm": 6.276622295379639,
+      "learning_rate": 8.730827067669172e-06,
+      "loss": 0.3733,
+      "step": 8440
+    },
+    {
+      "epoch": 12.71,
+      "grad_norm": 7.50572395324707,
+      "learning_rate": 8.729323308270678e-06,
+      "loss": 0.4407,
+      "step": 8450
+    },
+    {
+      "epoch": 12.72,
+      "grad_norm": 7.089003086090088,
+      "learning_rate": 8.727819548872181e-06,
+      "loss": 0.3948,
+      "step": 8460
+    },
+    {
+      "epoch": 12.74,
+      "grad_norm": 6.90725564956665,
+      "learning_rate": 8.726315789473685e-06,
+      "loss": 0.4511,
+      "step": 8470
+    },
+    {
+      "epoch": 12.75,
+      "grad_norm": 4.369374752044678,
+      "learning_rate": 8.724812030075188e-06,
+      "loss": 0.3559,
+      "step": 8480
+    },
+    {
+      "epoch": 12.77,
+      "grad_norm": 2.895493507385254,
+      "learning_rate": 8.723308270676692e-06,
+      "loss": 0.349,
+      "step": 8490
+    },
+    {
+      "epoch": 12.78,
+      "grad_norm": 8.638984680175781,
+      "learning_rate": 8.721804511278195e-06,
+      "loss": 0.3406,
+      "step": 8500
+    },
+    {
+      "epoch": 12.8,
+      "grad_norm": 7.664207458496094,
+      "learning_rate": 8.7203007518797e-06,
+      "loss": 0.3619,
+      "step": 8510
+    },
+    {
+      "epoch": 12.81,
+      "grad_norm": 4.544347286224365,
+      "learning_rate": 8.718796992481204e-06,
+      "loss": 0.3109,
+      "step": 8520
+    },
+    {
+      "epoch": 12.83,
+      "grad_norm": 6.640614032745361,
+      "learning_rate": 8.717293233082708e-06,
+      "loss": 0.4116,
+      "step": 8530
+    },
+    {
+      "epoch": 12.84,
+      "grad_norm": 7.840051174163818,
+      "learning_rate": 8.715789473684211e-06,
+      "loss": 0.4027,
+      "step": 8540
+    },
+    {
+      "epoch": 12.86,
+      "grad_norm": 10.355204582214355,
+      "learning_rate": 8.714285714285715e-06,
+      "loss": 0.4013,
+      "step": 8550
+    },
+    {
+      "epoch": 12.87,
+      "grad_norm": 7.472030162811279,
+      "learning_rate": 8.712781954887218e-06,
+      "loss": 0.4119,
+      "step": 8560
+    },
+    {
+      "epoch": 12.89,
+      "grad_norm": 7.9360246658325195,
+      "learning_rate": 8.711278195488723e-06,
+      "loss": 0.3472,
+      "step": 8570
+    },
+    {
+      "epoch": 12.9,
+      "grad_norm": 5.889431953430176,
+      "learning_rate": 8.709774436090225e-06,
+      "loss": 0.4009,
+      "step": 8580
+    },
+    {
+      "epoch": 12.92,
+      "grad_norm": 5.548401355743408,
+      "learning_rate": 8.70827067669173e-06,
+      "loss": 0.4261,
+      "step": 8590
+    },
+    {
+      "epoch": 12.93,
+      "grad_norm": 5.590747833251953,
+      "learning_rate": 8.706766917293234e-06,
+      "loss": 0.4437,
+      "step": 8600
+    },
+    {
+      "epoch": 12.95,
+      "grad_norm": 6.401696681976318,
+      "learning_rate": 8.705263157894737e-06,
+      "loss": 0.3746,
+      "step": 8610
+    },
+    {
+      "epoch": 12.96,
+      "grad_norm": 9.315383911132812,
+      "learning_rate": 8.703759398496241e-06,
+      "loss": 0.3803,
+      "step": 8620
+    },
+    {
+      "epoch": 12.98,
+      "grad_norm": 3.9589388370513916,
+      "learning_rate": 8.702255639097746e-06,
+      "loss": 0.4448,
+      "step": 8630
+    },
+    {
+      "epoch": 12.99,
+      "grad_norm": 4.445014953613281,
+      "learning_rate": 8.700751879699248e-06,
+      "loss": 0.4289,
+      "step": 8640
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9288,
+      "eval_loss": 0.2712935507297516,
+      "eval_runtime": 84.8607,
+      "eval_samples_per_second": 117.84,
+      "eval_steps_per_second": 0.471,
+      "step": 8645
+    },
+    {
+      "epoch": 13.01,
+      "grad_norm": 5.444362163543701,
+      "learning_rate": 8.699248120300753e-06,
+      "loss": 0.3994,
+      "step": 8650
+    },
+    {
+      "epoch": 13.02,
+      "grad_norm": 5.8952178955078125,
+      "learning_rate": 8.697744360902257e-06,
+      "loss": 0.3819,
+      "step": 8660
+    },
+    {
+      "epoch": 13.04,
+      "grad_norm": 5.363025188446045,
+      "learning_rate": 8.69624060150376e-06,
+      "loss": 0.4251,
+      "step": 8670
+    },
+    {
+      "epoch": 13.05,
+      "grad_norm": 6.1266961097717285,
+      "learning_rate": 8.694736842105264e-06,
+      "loss": 0.4236,
+      "step": 8680
+    },
+    {
+      "epoch": 13.07,
+      "grad_norm": 6.096094131469727,
+      "learning_rate": 8.693233082706767e-06,
+      "loss": 0.4411,
+      "step": 8690
+    },
+    {
+      "epoch": 13.08,
+      "grad_norm": 6.0483293533325195,
+      "learning_rate": 8.69172932330827e-06,
+      "loss": 0.3538,
+      "step": 8700
+    },
+    {
+      "epoch": 13.1,
+      "grad_norm": 8.619955062866211,
+      "learning_rate": 8.690225563909776e-06,
+      "loss": 0.4698,
+      "step": 8710
+    },
+    {
+      "epoch": 13.11,
+      "grad_norm": 5.028072834014893,
+      "learning_rate": 8.68872180451128e-06,
+      "loss": 0.3883,
+      "step": 8720
+    },
+    {
+      "epoch": 13.13,
+      "grad_norm": 7.43666934967041,
+      "learning_rate": 8.687218045112783e-06,
+      "loss": 0.3552,
+      "step": 8730
+    },
+    {
+      "epoch": 13.14,
+      "grad_norm": 9.520151138305664,
+      "learning_rate": 8.685714285714287e-06,
+      "loss": 0.4079,
+      "step": 8740
+    },
+    {
+      "epoch": 13.16,
+      "grad_norm": 7.852067947387695,
+      "learning_rate": 8.68421052631579e-06,
+      "loss": 0.3607,
+      "step": 8750
+    },
+    {
+      "epoch": 13.17,
+      "grad_norm": 5.92877721786499,
+      "learning_rate": 8.682706766917294e-06,
+      "loss": 0.3739,
+      "step": 8760
+    },
+    {
+      "epoch": 13.19,
+      "grad_norm": 4.25166130065918,
+      "learning_rate": 8.681203007518799e-06,
+      "loss": 0.4621,
+      "step": 8770
+    },
+    {
+      "epoch": 13.2,
+      "grad_norm": 7.073912143707275,
+      "learning_rate": 8.6796992481203e-06,
+      "loss": 0.4465,
+      "step": 8780
+    },
+    {
+      "epoch": 13.22,
+      "grad_norm": 7.39524507522583,
+      "learning_rate": 8.678195488721806e-06,
+      "loss": 0.4303,
+      "step": 8790
+    },
+    {
+      "epoch": 13.23,
+      "grad_norm": 6.938388824462891,
+      "learning_rate": 8.67669172932331e-06,
+      "loss": 0.3535,
+      "step": 8800
+    },
+    {
+      "epoch": 13.25,
+      "grad_norm": 5.0067524909973145,
+      "learning_rate": 8.675187969924813e-06,
+      "loss": 0.4399,
+      "step": 8810
+    },
+    {
+      "epoch": 13.26,
+      "grad_norm": 6.340808391571045,
+      "learning_rate": 8.673684210526316e-06,
+      "loss": 0.4199,
+      "step": 8820
+    },
+    {
+      "epoch": 13.28,
+      "grad_norm": 4.246801853179932,
+      "learning_rate": 8.67218045112782e-06,
+      "loss": 0.4039,
+      "step": 8830
+    },
+    {
+      "epoch": 13.29,
+      "grad_norm": 4.85552453994751,
+      "learning_rate": 8.670676691729323e-06,
+      "loss": 0.3753,
+      "step": 8840
+    },
+    {
+      "epoch": 13.31,
+      "grad_norm": 6.020550727844238,
+      "learning_rate": 8.669172932330829e-06,
+      "loss": 0.4039,
+      "step": 8850
+    },
+    {
+      "epoch": 13.32,
+      "grad_norm": 3.4875411987304688,
+      "learning_rate": 8.667669172932332e-06,
+      "loss": 0.3829,
+      "step": 8860
+    },
+    {
+      "epoch": 13.34,
+      "grad_norm": 6.239095211029053,
+      "learning_rate": 8.666165413533836e-06,
+      "loss": 0.3511,
+      "step": 8870
+    },
+    {
+      "epoch": 13.35,
+      "grad_norm": 4.244966983795166,
+      "learning_rate": 8.66466165413534e-06,
+      "loss": 0.4268,
+      "step": 8880
+    },
+    {
+      "epoch": 13.37,
+      "grad_norm": 12.684317588806152,
+      "learning_rate": 8.663157894736843e-06,
+      "loss": 0.3471,
+      "step": 8890
+    },
+    {
+      "epoch": 13.38,
+      "grad_norm": 8.664961814880371,
+      "learning_rate": 8.661654135338346e-06,
+      "loss": 0.3822,
+      "step": 8900
+    },
+    {
+      "epoch": 13.4,
+      "grad_norm": 5.7766804695129395,
+      "learning_rate": 8.660150375939851e-06,
+      "loss": 0.4199,
+      "step": 8910
+    },
+    {
+      "epoch": 13.41,
+      "grad_norm": 4.019351959228516,
+      "learning_rate": 8.658646616541353e-06,
+      "loss": 0.4137,
+      "step": 8920
+    },
+    {
+      "epoch": 13.43,
+      "grad_norm": 6.156152248382568,
+      "learning_rate": 8.657142857142858e-06,
+      "loss": 0.4606,
+      "step": 8930
+    },
+    {
+      "epoch": 13.44,
+      "grad_norm": 5.74890661239624,
+      "learning_rate": 8.655639097744362e-06,
+      "loss": 0.3221,
+      "step": 8940
+    },
+    {
+      "epoch": 13.46,
+      "grad_norm": 6.321985721588135,
+      "learning_rate": 8.654135338345866e-06,
+      "loss": 0.415,
+      "step": 8950
+    },
+    {
+      "epoch": 13.47,
+      "grad_norm": 8.88508129119873,
+      "learning_rate": 8.652631578947369e-06,
+      "loss": 0.3714,
+      "step": 8960
+    },
+    {
+      "epoch": 13.49,
+      "grad_norm": 6.999327659606934,
+      "learning_rate": 8.651127819548873e-06,
+      "loss": 0.3576,
+      "step": 8970
+    },
+    {
+      "epoch": 13.5,
+      "grad_norm": 7.313613414764404,
+      "learning_rate": 8.649624060150376e-06,
+      "loss": 0.4096,
+      "step": 8980
+    },
+    {
+      "epoch": 13.52,
+      "grad_norm": 6.363276958465576,
+      "learning_rate": 8.64812030075188e-06,
+      "loss": 0.3356,
+      "step": 8990
+    },
+    {
+      "epoch": 13.53,
+      "grad_norm": 7.81085729598999,
+      "learning_rate": 8.646616541353385e-06,
+      "loss": 0.4216,
+      "step": 9000
+    },
+    {
+      "epoch": 13.55,
+      "grad_norm": 8.093158721923828,
+      "learning_rate": 8.645112781954887e-06,
+      "loss": 0.5105,
+      "step": 9010
+    },
+    {
+      "epoch": 13.56,
+      "grad_norm": 3.801630735397339,
+      "learning_rate": 8.643609022556392e-06,
+      "loss": 0.439,
+      "step": 9020
+    },
+    {
+      "epoch": 13.58,
+      "grad_norm": 5.564939975738525,
+      "learning_rate": 8.642105263157895e-06,
+      "loss": 0.3854,
+      "step": 9030
+    },
+    {
+      "epoch": 13.59,
+      "grad_norm": 9.847439765930176,
+      "learning_rate": 8.640601503759399e-06,
+      "loss": 0.4034,
+      "step": 9040
+    },
+    {
+      "epoch": 13.61,
+      "grad_norm": 9.21834659576416,
+      "learning_rate": 8.639097744360902e-06,
+      "loss": 0.4448,
+      "step": 9050
+    },
+    {
+      "epoch": 13.62,
+      "grad_norm": 4.98524808883667,
+      "learning_rate": 8.637593984962408e-06,
+      "loss": 0.3646,
+      "step": 9060
+    },
+    {
+      "epoch": 13.64,
+      "grad_norm": 6.707414150238037,
+      "learning_rate": 8.63609022556391e-06,
+      "loss": 0.3618,
+      "step": 9070
+    },
+    {
+      "epoch": 13.65,
+      "grad_norm": 5.5840840339660645,
+      "learning_rate": 8.634586466165415e-06,
+      "loss": 0.3628,
+      "step": 9080
+    },
+    {
+      "epoch": 13.67,
+      "grad_norm": 4.939608097076416,
+      "learning_rate": 8.633082706766918e-06,
+      "loss": 0.3785,
+      "step": 9090
+    },
+    {
+      "epoch": 13.68,
+      "grad_norm": 7.449197769165039,
+      "learning_rate": 8.631578947368422e-06,
+      "loss": 0.4354,
+      "step": 9100
+    },
+    {
+      "epoch": 13.7,
+      "grad_norm": 9.470358848571777,
+      "learning_rate": 8.630075187969925e-06,
+      "loss": 0.4075,
+      "step": 9110
+    },
+    {
+      "epoch": 13.71,
+      "grad_norm": 7.6183085441589355,
+      "learning_rate": 8.628571428571429e-06,
+      "loss": 0.4067,
+      "step": 9120
+    },
+    {
+      "epoch": 13.73,
+      "grad_norm": 3.0916943550109863,
+      "learning_rate": 8.627067669172932e-06,
+      "loss": 0.365,
+      "step": 9130
+    },
+    {
+      "epoch": 13.74,
+      "grad_norm": 4.251070499420166,
+      "learning_rate": 8.625563909774437e-06,
+      "loss": 0.4255,
+      "step": 9140
+    },
+    {
+      "epoch": 13.76,
+      "grad_norm": 6.8059282302856445,
+      "learning_rate": 8.624060150375941e-06,
+      "loss": 0.32,
+      "step": 9150
+    },
+    {
+      "epoch": 13.77,
+      "grad_norm": 7.302189826965332,
+      "learning_rate": 8.622556390977444e-06,
+      "loss": 0.4131,
+      "step": 9160
+    },
+    {
+      "epoch": 13.79,
+      "grad_norm": 6.402463436126709,
+      "learning_rate": 8.621052631578948e-06,
+      "loss": 0.3848,
+      "step": 9170
+    },
+    {
+      "epoch": 13.8,
+      "grad_norm": 4.343325138092041,
+      "learning_rate": 8.619548872180452e-06,
+      "loss": 0.4014,
+      "step": 9180
+    },
+    {
+      "epoch": 13.82,
+      "grad_norm": 9.013459205627441,
+      "learning_rate": 8.618045112781955e-06,
+      "loss": 0.3739,
+      "step": 9190
+    },
+    {
+      "epoch": 13.83,
+      "grad_norm": 7.037381172180176,
+      "learning_rate": 8.61654135338346e-06,
+      "loss": 0.4189,
+      "step": 9200
+    },
+    {
+      "epoch": 13.85,
+      "grad_norm": 4.7024760246276855,
+      "learning_rate": 8.615037593984962e-06,
+      "loss": 0.38,
+      "step": 9210
+    },
+    {
+      "epoch": 13.86,
+      "grad_norm": 4.808414936065674,
+      "learning_rate": 8.613533834586467e-06,
+      "loss": 0.4414,
+      "step": 9220
+    },
+    {
+      "epoch": 13.88,
+      "grad_norm": 8.237750053405762,
+      "learning_rate": 8.61203007518797e-06,
+      "loss": 0.4215,
+      "step": 9230
+    },
+    {
+      "epoch": 13.89,
+      "grad_norm": 7.862570285797119,
+      "learning_rate": 8.610526315789474e-06,
+      "loss": 0.4727,
+      "step": 9240
+    },
+    {
+      "epoch": 13.91,
+      "grad_norm": 7.045783519744873,
+      "learning_rate": 8.609022556390978e-06,
+      "loss": 0.4109,
+      "step": 9250
+    },
+    {
+      "epoch": 13.92,
+      "grad_norm": 5.3544135093688965,
+      "learning_rate": 8.607518796992483e-06,
+      "loss": 0.3824,
+      "step": 9260
+    },
+    {
+      "epoch": 13.94,
+      "grad_norm": 14.21022891998291,
+      "learning_rate": 8.606015037593985e-06,
+      "loss": 0.3659,
+      "step": 9270
+    },
+    {
+      "epoch": 13.95,
+      "grad_norm": 7.408153533935547,
+      "learning_rate": 8.60451127819549e-06,
+      "loss": 0.3508,
+      "step": 9280
+    },
+    {
+      "epoch": 13.97,
+      "grad_norm": 3.206442356109619,
+      "learning_rate": 8.603007518796994e-06,
+      "loss": 0.4451,
+      "step": 9290
+    },
+    {
+      "epoch": 13.98,
+      "grad_norm": 4.974185466766357,
+      "learning_rate": 8.601503759398497e-06,
+      "loss": 0.3888,
+      "step": 9300
+    },
+    {
+      "epoch": 14.0,
+      "grad_norm": 15.39065170288086,
+      "learning_rate": 8.6e-06,
+      "loss": 0.4003,
+      "step": 9310
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.932,
+      "eval_loss": 0.2616865336894989,
+      "eval_runtime": 84.8808,
+      "eval_samples_per_second": 117.812,
+      "eval_steps_per_second": 0.471,
+      "step": 9310
+    },
+    {
+      "epoch": 14.02,
+      "grad_norm": 5.1268534660339355,
+      "learning_rate": 8.598496240601504e-06,
+      "loss": 0.4195,
+      "step": 9320
+    },
+    {
+      "epoch": 14.03,
+      "grad_norm": 6.874084949493408,
+      "learning_rate": 8.596992481203008e-06,
+      "loss": 0.3901,
+      "step": 9330
+    },
+    {
+      "epoch": 14.05,
+      "grad_norm": 11.405204772949219,
+      "learning_rate": 8.595488721804513e-06,
+      "loss": 0.3234,
+      "step": 9340
+    },
+    {
+      "epoch": 14.06,
+      "grad_norm": 4.844882965087891,
+      "learning_rate": 8.593984962406016e-06,
+      "loss": 0.3437,
+      "step": 9350
+    },
+    {
+      "epoch": 14.08,
+      "grad_norm": 7.187948226928711,
+      "learning_rate": 8.59248120300752e-06,
+      "loss": 0.3895,
+      "step": 9360
+    },
+    {
+      "epoch": 14.09,
+      "grad_norm": 3.7594106197357178,
+      "learning_rate": 8.590977443609023e-06,
+      "loss": 0.3329,
+      "step": 9370
+    },
+    {
+      "epoch": 14.11,
+      "grad_norm": 4.245199203491211,
+      "learning_rate": 8.589473684210527e-06,
+      "loss": 0.3644,
+      "step": 9380
+    },
+    {
+      "epoch": 14.12,
+      "grad_norm": 6.302145004272461,
+      "learning_rate": 8.58796992481203e-06,
+      "loss": 0.4615,
+      "step": 9390
+    },
+    {
+      "epoch": 14.14,
+      "grad_norm": 6.26497220993042,
+      "learning_rate": 8.586466165413536e-06,
+      "loss": 0.3983,
+      "step": 9400
+    },
+    {
+      "epoch": 14.15,
+      "grad_norm": 7.315799236297607,
+      "learning_rate": 8.584962406015038e-06,
+      "loss": 0.3474,
+      "step": 9410
+    },
+    {
+      "epoch": 14.17,
+      "grad_norm": 8.584407806396484,
+      "learning_rate": 8.583458646616543e-06,
+      "loss": 0.3858,
+      "step": 9420
+    },
+    {
+      "epoch": 14.18,
+      "grad_norm": 6.192986488342285,
+      "learning_rate": 8.581954887218046e-06,
+      "loss": 0.4653,
+      "step": 9430
+    },
+    {
+      "epoch": 14.2,
+      "grad_norm": 6.261072635650635,
+      "learning_rate": 8.58045112781955e-06,
+      "loss": 0.3686,
+      "step": 9440
+    },
+    {
+      "epoch": 14.21,
+      "grad_norm": 6.7162017822265625,
+      "learning_rate": 8.578947368421053e-06,
+      "loss": 0.4143,
+      "step": 9450
+    },
+    {
+      "epoch": 14.23,
+      "grad_norm": 5.550053119659424,
+      "learning_rate": 8.577443609022557e-06,
+      "loss": 0.4262,
+      "step": 9460
+    },
+    {
+      "epoch": 14.24,
+      "grad_norm": 6.601341247558594,
+      "learning_rate": 8.57593984962406e-06,
+      "loss": 0.3615,
+      "step": 9470
+    },
+    {
+      "epoch": 14.26,
+      "grad_norm": 6.859097957611084,
+      "learning_rate": 8.574436090225564e-06,
+      "loss": 0.3584,
+      "step": 9480
+    },
+    {
+      "epoch": 14.27,
+      "grad_norm": 3.824615478515625,
+      "learning_rate": 8.572932330827069e-06,
+      "loss": 0.3941,
+      "step": 9490
+    },
+    {
+      "epoch": 14.29,
+      "grad_norm": 6.923836708068848,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 0.4016,
+      "step": 9500
+    },
+    {
+      "epoch": 14.3,
+      "grad_norm": 6.395806789398193,
+      "learning_rate": 8.569924812030076e-06,
+      "loss": 0.368,
+      "step": 9510
+    },
+    {
+      "epoch": 14.32,
+      "grad_norm": 6.522418022155762,
+      "learning_rate": 8.56842105263158e-06,
+      "loss": 0.377,
+      "step": 9520
+    },
+    {
+      "epoch": 14.33,
+      "grad_norm": 7.502889633178711,
+      "learning_rate": 8.566917293233083e-06,
+      "loss": 0.4179,
+      "step": 9530
+    },
+    {
+      "epoch": 14.35,
+      "grad_norm": 6.025669574737549,
+      "learning_rate": 8.565413533834587e-06,
+      "loss": 0.3395,
+      "step": 9540
+    },
+    {
+      "epoch": 14.36,
+      "grad_norm": 7.751435279846191,
+      "learning_rate": 8.563909774436092e-06,
+      "loss": 0.3531,
+      "step": 9550
+    },
+    {
+      "epoch": 14.38,
+      "grad_norm": 6.964672088623047,
+      "learning_rate": 8.562406015037594e-06,
+      "loss": 0.3905,
+      "step": 9560
+    },
+    {
+      "epoch": 14.39,
+      "grad_norm": 2.502666473388672,
+      "learning_rate": 8.560902255639099e-06,
+      "loss": 0.3172,
+      "step": 9570
+    },
+    {
+      "epoch": 14.41,
+      "grad_norm": 7.133659839630127,
+      "learning_rate": 8.559398496240602e-06,
+      "loss": 0.442,
+      "step": 9580
+    },
+    {
+      "epoch": 14.42,
+      "grad_norm": 4.32753324508667,
+      "learning_rate": 8.557894736842106e-06,
+      "loss": 0.3458,
+      "step": 9590
+    },
+    {
+      "epoch": 14.44,
+      "grad_norm": 6.268803119659424,
+      "learning_rate": 8.55639097744361e-06,
+      "loss": 0.3431,
+      "step": 9600
+    },
+    {
+      "epoch": 14.45,
+      "grad_norm": 7.018800735473633,
+      "learning_rate": 8.554887218045113e-06,
+      "loss": 0.4369,
+      "step": 9610
+    },
+    {
+      "epoch": 14.47,
+      "grad_norm": 7.435917377471924,
+      "learning_rate": 8.553383458646617e-06,
+      "loss": 0.3861,
+      "step": 9620
+    },
+    {
+      "epoch": 14.48,
+      "grad_norm": 7.8388752937316895,
+      "learning_rate": 8.551879699248122e-06,
+      "loss": 0.4239,
+      "step": 9630
+    },
+    {
+      "epoch": 14.5,
+      "grad_norm": 7.880455493927002,
+      "learning_rate": 8.550375939849625e-06,
+      "loss": 0.3829,
+      "step": 9640
+    },
+    {
+      "epoch": 14.51,
+      "grad_norm": 5.568830490112305,
+      "learning_rate": 8.548872180451129e-06,
+      "loss": 0.4011,
+      "step": 9650
+    },
+    {
+      "epoch": 14.53,
+      "grad_norm": 5.6487274169921875,
+      "learning_rate": 8.547368421052632e-06,
+      "loss": 0.4091,
+      "step": 9660
+    },
+    {
+      "epoch": 14.54,
+      "grad_norm": 4.51718807220459,
+      "learning_rate": 8.545864661654136e-06,
+      "loss": 0.4087,
+      "step": 9670
+    },
+    {
+      "epoch": 14.56,
+      "grad_norm": 7.874798774719238,
+      "learning_rate": 8.54436090225564e-06,
+      "loss": 0.3494,
+      "step": 9680
+    },
+    {
+      "epoch": 14.57,
+      "grad_norm": 4.497681617736816,
+      "learning_rate": 8.542857142857145e-06,
+      "loss": 0.3783,
+      "step": 9690
+    },
+    {
+      "epoch": 14.59,
+      "grad_norm": 5.411101818084717,
+      "learning_rate": 8.541353383458646e-06,
+      "loss": 0.3395,
+      "step": 9700
+    },
+    {
+      "epoch": 14.6,
+      "grad_norm": 5.714541435241699,
+      "learning_rate": 8.539849624060152e-06,
+      "loss": 0.4507,
+      "step": 9710
+    },
+    {
+      "epoch": 14.62,
+      "grad_norm": 7.042336940765381,
+      "learning_rate": 8.538345864661655e-06,
+      "loss": 0.3363,
+      "step": 9720
+    },
+    {
+      "epoch": 14.63,
+      "grad_norm": 6.70949649810791,
+      "learning_rate": 8.536842105263159e-06,
+      "loss": 0.3559,
+      "step": 9730
+    },
+    {
+      "epoch": 14.65,
+      "grad_norm": 5.84644889831543,
+      "learning_rate": 8.535338345864662e-06,
+      "loss": 0.3936,
+      "step": 9740
+    },
+    {
+      "epoch": 14.66,
+      "grad_norm": 5.070087432861328,
+      "learning_rate": 8.533834586466166e-06,
+      "loss": 0.4014,
+      "step": 9750
+    },
+    {
+      "epoch": 14.68,
+      "grad_norm": 5.353463172912598,
+      "learning_rate": 8.53233082706767e-06,
+      "loss": 0.4091,
+      "step": 9760
+    },
+    {
+      "epoch": 14.69,
+      "grad_norm": 9.118497848510742,
+      "learning_rate": 8.530827067669174e-06,
+      "loss": 0.3922,
+      "step": 9770
+    },
+    {
+      "epoch": 14.71,
+      "grad_norm": 7.667191982269287,
+      "learning_rate": 8.529323308270678e-06,
+      "loss": 0.4087,
+      "step": 9780
+    },
+    {
+      "epoch": 14.72,
+      "grad_norm": 7.210267066955566,
+      "learning_rate": 8.527819548872181e-06,
+      "loss": 0.3458,
+      "step": 9790
+    },
+    {
+      "epoch": 14.74,
+      "grad_norm": 5.242373466491699,
+      "learning_rate": 8.526315789473685e-06,
+      "loss": 0.3293,
+      "step": 9800
+    },
+    {
+      "epoch": 14.75,
+      "grad_norm": 7.6933393478393555,
+      "learning_rate": 8.524812030075188e-06,
+      "loss": 0.4406,
+      "step": 9810
+    },
+    {
+      "epoch": 14.77,
+      "grad_norm": 5.179628372192383,
+      "learning_rate": 8.523308270676692e-06,
+      "loss": 0.3827,
+      "step": 9820
+    },
+    {
+      "epoch": 14.78,
+      "grad_norm": 9.525907516479492,
+      "learning_rate": 8.521804511278197e-06,
+      "loss": 0.4095,
+      "step": 9830
+    },
+    {
+      "epoch": 14.8,
+      "grad_norm": 6.132147789001465,
+      "learning_rate": 8.520300751879699e-06,
+      "loss": 0.3778,
+      "step": 9840
+    },
+    {
+      "epoch": 14.81,
+      "grad_norm": 8.04976749420166,
+      "learning_rate": 8.518796992481204e-06,
+      "loss": 0.3996,
+      "step": 9850
+    },
+    {
+      "epoch": 14.83,
+      "grad_norm": 9.131913185119629,
+      "learning_rate": 8.517293233082708e-06,
+      "loss": 0.4103,
+      "step": 9860
+    },
+    {
+      "epoch": 14.84,
+      "grad_norm": 5.724211692810059,
+      "learning_rate": 8.515789473684211e-06,
+      "loss": 0.4166,
+      "step": 9870
+    },
+    {
+      "epoch": 14.86,
+      "grad_norm": 4.537842750549316,
+      "learning_rate": 8.514285714285715e-06,
+      "loss": 0.4357,
+      "step": 9880
+    },
+    {
+      "epoch": 14.87,
+      "grad_norm": 9.75554370880127,
+      "learning_rate": 8.51278195488722e-06,
+      "loss": 0.351,
+      "step": 9890
+    },
+    {
+      "epoch": 14.89,
+      "grad_norm": 5.427340030670166,
+      "learning_rate": 8.511278195488722e-06,
+      "loss": 0.3488,
+      "step": 9900
+    },
+    {
+      "epoch": 14.9,
+      "grad_norm": 4.465277671813965,
+      "learning_rate": 8.509774436090227e-06,
+      "loss": 0.32,
+      "step": 9910
+    },
+    {
+      "epoch": 14.92,
+      "grad_norm": 4.331689834594727,
+      "learning_rate": 8.50827067669173e-06,
+      "loss": 0.4442,
+      "step": 9920
+    },
+    {
+      "epoch": 14.93,
+      "grad_norm": 5.798705577850342,
+      "learning_rate": 8.506766917293232e-06,
+      "loss": 0.4003,
+      "step": 9930
+    },
+    {
+      "epoch": 14.95,
+      "grad_norm": 9.353456497192383,
+      "learning_rate": 8.505263157894738e-06,
+      "loss": 0.3969,
+      "step": 9940
+    },
+    {
+      "epoch": 14.96,
+      "grad_norm": 7.088143825531006,
+      "learning_rate": 8.503759398496241e-06,
+      "loss": 0.3749,
+      "step": 9950
+    },
+    {
+      "epoch": 14.98,
+      "grad_norm": 6.825297832489014,
+      "learning_rate": 8.502255639097745e-06,
+      "loss": 0.3532,
+      "step": 9960
+    },
+    {
+      "epoch": 14.99,
+      "grad_norm": 3.3977503776550293,
+      "learning_rate": 8.500751879699248e-06,
+      "loss": 0.3227,
+      "step": 9970
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.9335,
+      "eval_loss": 0.2566547989845276,
+      "eval_runtime": 84.4818,
+      "eval_samples_per_second": 118.369,
+      "eval_steps_per_second": 0.473,
+      "step": 9975
+    },
+    {
+      "epoch": 15.01,
+      "grad_norm": 3.752779483795166,
+      "learning_rate": 8.499248120300753e-06,
+      "loss": 0.2741,
+      "step": 9980
+    },
+    {
+      "epoch": 15.02,
+      "grad_norm": 5.626667499542236,
+      "learning_rate": 8.497744360902255e-06,
+      "loss": 0.3584,
+      "step": 9990
+    },
+    {
+      "epoch": 15.04,
+      "grad_norm": 5.8750834465026855,
+      "learning_rate": 8.49624060150376e-06,
+      "loss": 0.3563,
+      "step": 10000
+    },
+    {
+      "epoch": 15.05,
+      "grad_norm": 6.653073787689209,
+      "learning_rate": 8.494736842105264e-06,
+      "loss": 0.3808,
+      "step": 10010
+    },
+    {
+      "epoch": 15.07,
+      "grad_norm": 6.017663955688477,
+      "learning_rate": 8.493233082706767e-06,
+      "loss": 0.3165,
+      "step": 10020
+    },
+    {
+      "epoch": 15.08,
+      "grad_norm": 4.427550792694092,
+      "learning_rate": 8.491729323308271e-06,
+      "loss": 0.4417,
+      "step": 10030
+    },
+    {
+      "epoch": 15.1,
+      "grad_norm": 7.968047142028809,
+      "learning_rate": 8.490225563909775e-06,
+      "loss": 0.3807,
+      "step": 10040
+    },
+    {
+      "epoch": 15.11,
+      "grad_norm": 6.646710395812988,
+      "learning_rate": 8.488721804511278e-06,
+      "loss": 0.3014,
+      "step": 10050
+    },
+    {
+      "epoch": 15.13,
+      "grad_norm": 5.180790424346924,
+      "learning_rate": 8.487218045112783e-06,
+      "loss": 0.3737,
+      "step": 10060
+    },
+    {
+      "epoch": 15.14,
+      "grad_norm": 8.142125129699707,
+      "learning_rate": 8.485714285714287e-06,
+      "loss": 0.3953,
+      "step": 10070
+    },
+    {
+      "epoch": 15.16,
+      "grad_norm": 5.258510589599609,
+      "learning_rate": 8.48421052631579e-06,
+      "loss": 0.2742,
+      "step": 10080
+    },
+    {
+      "epoch": 15.17,
+      "grad_norm": 7.299388408660889,
+      "learning_rate": 8.482706766917294e-06,
+      "loss": 0.388,
+      "step": 10090
+    },
+    {
+      "epoch": 15.19,
+      "grad_norm": 9.05027961730957,
+      "learning_rate": 8.481203007518797e-06,
+      "loss": 0.3364,
+      "step": 10100
+    },
+    {
+      "epoch": 15.2,
+      "grad_norm": 8.713417053222656,
+      "learning_rate": 8.4796992481203e-06,
+      "loss": 0.2847,
+      "step": 10110
+    },
+    {
+      "epoch": 15.22,
+      "grad_norm": 4.598002910614014,
+      "learning_rate": 8.478195488721806e-06,
+      "loss": 0.3536,
+      "step": 10120
+    },
+    {
+      "epoch": 15.23,
+      "grad_norm": 8.883098602294922,
+      "learning_rate": 8.476691729323308e-06,
+      "loss": 0.3917,
+      "step": 10130
+    },
+    {
+      "epoch": 15.25,
+      "grad_norm": 9.343592643737793,
+      "learning_rate": 8.475187969924813e-06,
+      "loss": 0.3912,
+      "step": 10140
+    },
+    {
+      "epoch": 15.26,
+      "grad_norm": 11.456267356872559,
+      "learning_rate": 8.473684210526317e-06,
+      "loss": 0.3171,
+      "step": 10150
+    },
+    {
+      "epoch": 15.28,
+      "grad_norm": 7.874906539916992,
+      "learning_rate": 8.47218045112782e-06,
+      "loss": 0.4443,
+      "step": 10160
+    },
+    {
+      "epoch": 15.29,
+      "grad_norm": 7.280247211456299,
+      "learning_rate": 8.470676691729324e-06,
+      "loss": 0.4121,
+      "step": 10170
+    },
+    {
+      "epoch": 15.31,
+      "grad_norm": 7.839987754821777,
+      "learning_rate": 8.469172932330829e-06,
+      "loss": 0.4284,
+      "step": 10180
+    },
+    {
+      "epoch": 15.32,
+      "grad_norm": 5.39335298538208,
+      "learning_rate": 8.46766917293233e-06,
+      "loss": 0.4311,
+      "step": 10190
+    },
+    {
+      "epoch": 15.34,
+      "grad_norm": 8.08191204071045,
+      "learning_rate": 8.466165413533836e-06,
+      "loss": 0.3789,
+      "step": 10200
+    },
+    {
+      "epoch": 15.35,
+      "grad_norm": 3.493443250656128,
+      "learning_rate": 8.46466165413534e-06,
+      "loss": 0.3526,
+      "step": 10210
+    },
+    {
+      "epoch": 15.37,
+      "grad_norm": 7.41270637512207,
+      "learning_rate": 8.463157894736843e-06,
+      "loss": 0.4283,
+      "step": 10220
+    },
+    {
+      "epoch": 15.38,
+      "grad_norm": 6.4891486167907715,
+      "learning_rate": 8.461654135338346e-06,
+      "loss": 0.298,
+      "step": 10230
+    },
+    {
+      "epoch": 15.4,
+      "grad_norm": 6.028573989868164,
+      "learning_rate": 8.46015037593985e-06,
+      "loss": 0.3996,
+      "step": 10240
+    },
+    {
+      "epoch": 15.41,
+      "grad_norm": 3.365438938140869,
+      "learning_rate": 8.458646616541353e-06,
+      "loss": 0.334,
+      "step": 10250
+    },
+    {
+      "epoch": 15.43,
+      "grad_norm": 5.2097015380859375,
+      "learning_rate": 8.457142857142859e-06,
+      "loss": 0.3555,
+      "step": 10260
+    },
+    {
+      "epoch": 15.44,
+      "grad_norm": 4.656721591949463,
+      "learning_rate": 8.455639097744362e-06,
+      "loss": 0.398,
+      "step": 10270
+    },
+    {
+      "epoch": 15.46,
+      "grad_norm": 11.353671073913574,
+      "learning_rate": 8.454135338345866e-06,
+      "loss": 0.3341,
+      "step": 10280
+    },
+    {
+      "epoch": 15.47,
+      "grad_norm": 6.971073627471924,
+      "learning_rate": 8.45263157894737e-06,
+      "loss": 0.3569,
+      "step": 10290
+    },
+    {
+      "epoch": 15.49,
+      "grad_norm": 5.809013843536377,
+      "learning_rate": 8.451127819548873e-06,
+      "loss": 0.3642,
+      "step": 10300
+    },
+    {
+      "epoch": 15.5,
+      "grad_norm": 7.3322834968566895,
+      "learning_rate": 8.449624060150376e-06,
+      "loss": 0.3814,
+      "step": 10310
+    },
+    {
+      "epoch": 15.52,
+      "grad_norm": 9.322681427001953,
+      "learning_rate": 8.448120300751882e-06,
+      "loss": 0.3455,
+      "step": 10320
+    },
+    {
+      "epoch": 15.53,
+      "grad_norm": 7.197205066680908,
+      "learning_rate": 8.446616541353383e-06,
+      "loss": 0.3652,
+      "step": 10330
+    },
+    {
+      "epoch": 15.55,
+      "grad_norm": 7.1502766609191895,
+      "learning_rate": 8.445112781954889e-06,
+      "loss": 0.3882,
+      "step": 10340
+    },
+    {
+      "epoch": 15.56,
+      "grad_norm": 4.826005935668945,
+      "learning_rate": 8.443609022556392e-06,
+      "loss": 0.3878,
+      "step": 10350
+    },
+    {
+      "epoch": 15.58,
+      "grad_norm": 8.432343482971191,
+      "learning_rate": 8.442105263157896e-06,
+      "loss": 0.3778,
+      "step": 10360
+    },
+    {
+      "epoch": 15.59,
+      "grad_norm": 6.166329383850098,
+      "learning_rate": 8.440601503759399e-06,
+      "loss": 0.3675,
+      "step": 10370
+    },
+    {
+      "epoch": 15.61,
+      "grad_norm": 9.84304428100586,
+      "learning_rate": 8.439097744360903e-06,
+      "loss": 0.385,
+      "step": 10380
+    },
+    {
+      "epoch": 15.62,
+      "grad_norm": 4.937039375305176,
+      "learning_rate": 8.437593984962406e-06,
+      "loss": 0.3558,
+      "step": 10390
+    },
+    {
+      "epoch": 15.64,
+      "grad_norm": 5.817636966705322,
+      "learning_rate": 8.436090225563911e-06,
+      "loss": 0.3605,
+      "step": 10400
+    },
+    {
+      "epoch": 15.65,
+      "grad_norm": 6.982740879058838,
+      "learning_rate": 8.434586466165415e-06,
+      "loss": 0.2966,
+      "step": 10410
+    },
+    {
+      "epoch": 15.67,
+      "grad_norm": 5.2945098876953125,
+      "learning_rate": 8.433082706766918e-06,
+      "loss": 0.3279,
+      "step": 10420
+    },
+    {
+      "epoch": 15.68,
+      "grad_norm": 7.526950359344482,
+      "learning_rate": 8.431578947368422e-06,
+      "loss": 0.414,
+      "step": 10430
+    },
+    {
+      "epoch": 15.7,
+      "grad_norm": 8.325518608093262,
+      "learning_rate": 8.430075187969925e-06,
+      "loss": 0.314,
+      "step": 10440
+    },
+    {
+      "epoch": 15.71,
+      "grad_norm": 2.876897096633911,
+      "learning_rate": 8.428571428571429e-06,
+      "loss": 0.354,
+      "step": 10450
+    },
+    {
+      "epoch": 15.73,
+      "grad_norm": 6.776325702667236,
+      "learning_rate": 8.427067669172932e-06,
+      "loss": 0.3462,
+      "step": 10460
+    },
+    {
+      "epoch": 15.74,
+      "grad_norm": 8.158499717712402,
+      "learning_rate": 8.425563909774438e-06,
+      "loss": 0.3307,
+      "step": 10470
+    },
+    {
+      "epoch": 15.76,
+      "grad_norm": 10.01845645904541,
+      "learning_rate": 8.42406015037594e-06,
+      "loss": 0.3196,
+      "step": 10480
+    },
+    {
+      "epoch": 15.77,
+      "grad_norm": 4.778624534606934,
+      "learning_rate": 8.422556390977445e-06,
+      "loss": 0.3412,
+      "step": 10490
+    },
+    {
+      "epoch": 15.79,
+      "grad_norm": 5.8504157066345215,
+      "learning_rate": 8.421052631578948e-06,
+      "loss": 0.4183,
+      "step": 10500
+    },
+    {
+      "epoch": 15.8,
+      "grad_norm": 5.351130962371826,
+      "learning_rate": 8.419548872180452e-06,
+      "loss": 0.2639,
+      "step": 10510
+    },
+    {
+      "epoch": 15.82,
+      "grad_norm": 7.211291313171387,
+      "learning_rate": 8.418045112781955e-06,
+      "loss": 0.3021,
+      "step": 10520
+    },
+    {
+      "epoch": 15.83,
+      "grad_norm": 6.899810791015625,
+      "learning_rate": 8.416541353383459e-06,
+      "loss": 0.4069,
+      "step": 10530
+    },
+    {
+      "epoch": 15.85,
+      "grad_norm": 4.64746618270874,
+      "learning_rate": 8.415037593984962e-06,
+      "loss": 0.2798,
+      "step": 10540
+    },
+    {
+      "epoch": 15.86,
+      "grad_norm": 9.508644104003906,
+      "learning_rate": 8.413533834586468e-06,
+      "loss": 0.3791,
+      "step": 10550
+    },
+    {
+      "epoch": 15.88,
+      "grad_norm": 6.956771373748779,
+      "learning_rate": 8.412030075187971e-06,
+      "loss": 0.4188,
+      "step": 10560
+    },
+    {
+      "epoch": 15.89,
+      "grad_norm": 6.4203667640686035,
+      "learning_rate": 8.410526315789475e-06,
+      "loss": 0.3749,
+      "step": 10570
+    },
+    {
+      "epoch": 15.91,
+      "grad_norm": 8.310030937194824,
+      "learning_rate": 8.409022556390978e-06,
+      "loss": 0.4605,
+      "step": 10580
+    },
+    {
+      "epoch": 15.92,
+      "grad_norm": 8.788355827331543,
+      "learning_rate": 8.407518796992482e-06,
+      "loss": 0.3811,
+      "step": 10590
+    },
+    {
+      "epoch": 15.94,
+      "grad_norm": 5.931136131286621,
+      "learning_rate": 8.406015037593985e-06,
+      "loss": 0.4129,
+      "step": 10600
+    },
+    {
+      "epoch": 15.95,
+      "grad_norm": 7.846260070800781,
+      "learning_rate": 8.40451127819549e-06,
+      "loss": 0.3889,
+      "step": 10610
+    },
+    {
+      "epoch": 15.97,
+      "grad_norm": 6.834481239318848,
+      "learning_rate": 8.403007518796992e-06,
+      "loss": 0.4303,
+      "step": 10620
+    },
+    {
+      "epoch": 15.98,
+      "grad_norm": 3.6619720458984375,
+      "learning_rate": 8.401503759398497e-06,
+      "loss": 0.3483,
+      "step": 10630
+    },
+    {
+      "epoch": 16.0,
+      "grad_norm": 12.63433837890625,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.386,
+      "step": 10640
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.931,
+      "eval_loss": 0.25709524750709534,
+      "eval_runtime": 84.9959,
+      "eval_samples_per_second": 117.653,
+      "eval_steps_per_second": 0.471,
+      "step": 10640
+    },
+    {
+      "epoch": 16.02,
+      "grad_norm": 7.7511796951293945,
+      "learning_rate": 8.398496240601504e-06,
+      "loss": 0.3916,
+      "step": 10650
+    },
+    {
+      "epoch": 16.03,
+      "grad_norm": 4.640677452087402,
+      "learning_rate": 8.396992481203008e-06,
+      "loss": 0.395,
+      "step": 10660
+    },
+    {
+      "epoch": 16.05,
+      "grad_norm": 7.270589828491211,
+      "learning_rate": 8.395488721804511e-06,
+      "loss": 0.4299,
+      "step": 10670
+    },
+    {
+      "epoch": 16.06,
+      "grad_norm": 5.698379993438721,
+      "learning_rate": 8.393984962406015e-06,
+      "loss": 0.3792,
+      "step": 10680
+    },
+    {
+      "epoch": 16.08,
+      "grad_norm": 29.69732093811035,
+      "learning_rate": 8.39248120300752e-06,
+      "loss": 0.3564,
+      "step": 10690
+    },
+    {
+      "epoch": 16.09,
+      "grad_norm": 5.11942720413208,
+      "learning_rate": 8.390977443609024e-06,
+      "loss": 0.3284,
+      "step": 10700
+    },
+    {
+      "epoch": 16.11,
+      "grad_norm": 6.748551368713379,
+      "learning_rate": 8.389473684210527e-06,
+      "loss": 0.3582,
+      "step": 10710
+    },
+    {
+      "epoch": 16.12,
+      "grad_norm": 6.748464584350586,
+      "learning_rate": 8.38796992481203e-06,
+      "loss": 0.3707,
+      "step": 10720
+    },
+    {
+      "epoch": 16.14,
+      "grad_norm": 6.992805480957031,
+      "learning_rate": 8.386466165413534e-06,
+      "loss": 0.4388,
+      "step": 10730
+    },
+    {
+      "epoch": 16.15,
+      "grad_norm": 4.689752578735352,
+      "learning_rate": 8.384962406015038e-06,
+      "loss": 0.3394,
+      "step": 10740
+    },
+    {
+      "epoch": 16.17,
+      "grad_norm": 7.47608757019043,
+      "learning_rate": 8.383458646616543e-06,
+      "loss": 0.3662,
+      "step": 10750
+    },
+    {
+      "epoch": 16.18,
+      "grad_norm": 8.161937713623047,
+      "learning_rate": 8.381954887218045e-06,
+      "loss": 0.3729,
+      "step": 10760
+    },
+    {
+      "epoch": 16.2,
+      "grad_norm": 4.649080276489258,
+      "learning_rate": 8.38045112781955e-06,
+      "loss": 0.4012,
+      "step": 10770
+    },
+    {
+      "epoch": 16.21,
+      "grad_norm": 5.3081512451171875,
+      "learning_rate": 8.378947368421054e-06,
+      "loss": 0.3275,
+      "step": 10780
+    },
+    {
+      "epoch": 16.23,
+      "grad_norm": 8.424674034118652,
+      "learning_rate": 8.377443609022557e-06,
+      "loss": 0.3169,
+      "step": 10790
+    },
+    {
+      "epoch": 16.24,
+      "grad_norm": 7.213728427886963,
+      "learning_rate": 8.37593984962406e-06,
+      "loss": 0.4158,
+      "step": 10800
+    },
+    {
+      "epoch": 16.26,
+      "grad_norm": 4.228058815002441,
+      "learning_rate": 8.374436090225566e-06,
+      "loss": 0.2967,
+      "step": 10810
+    },
+    {
+      "epoch": 16.27,
+      "grad_norm": 9.091151237487793,
+      "learning_rate": 8.372932330827068e-06,
+      "loss": 0.3631,
+      "step": 10820
+    },
+    {
+      "epoch": 16.29,
+      "grad_norm": 7.325952053070068,
+      "learning_rate": 8.371428571428573e-06,
+      "loss": 0.4045,
+      "step": 10830
+    },
+    {
+      "epoch": 16.3,
+      "grad_norm": 8.557323455810547,
+      "learning_rate": 8.369924812030076e-06,
+      "loss": 0.3657,
+      "step": 10840
+    },
+    {
+      "epoch": 16.32,
+      "grad_norm": 8.98796558380127,
+      "learning_rate": 8.36842105263158e-06,
+      "loss": 0.3812,
+      "step": 10850
+    },
+    {
+      "epoch": 16.33,
+      "grad_norm": 14.74909496307373,
+      "learning_rate": 8.366917293233083e-06,
+      "loss": 0.3966,
+      "step": 10860
+    },
+    {
+      "epoch": 16.35,
+      "grad_norm": 6.872273921966553,
+      "learning_rate": 8.365413533834587e-06,
+      "loss": 0.3797,
+      "step": 10870
+    },
+    {
+      "epoch": 16.36,
+      "grad_norm": 8.392032623291016,
+      "learning_rate": 8.36390977443609e-06,
+      "loss": 0.3864,
+      "step": 10880
+    },
+    {
+      "epoch": 16.38,
+      "grad_norm": 8.61768627166748,
+      "learning_rate": 8.362406015037596e-06,
+      "loss": 0.3434,
+      "step": 10890
+    },
+    {
+      "epoch": 16.39,
+      "grad_norm": 6.622392177581787,
+      "learning_rate": 8.3609022556391e-06,
+      "loss": 0.3751,
+      "step": 10900
+    },
+    {
+      "epoch": 16.41,
+      "grad_norm": 7.108027458190918,
+      "learning_rate": 8.359398496240603e-06,
+      "loss": 0.3542,
+      "step": 10910
+    },
+    {
+      "epoch": 16.42,
+      "grad_norm": 7.145939826965332,
+      "learning_rate": 8.357894736842106e-06,
+      "loss": 0.3586,
+      "step": 10920
+    },
+    {
+      "epoch": 16.44,
+      "grad_norm": 16.715761184692383,
+      "learning_rate": 8.35639097744361e-06,
+      "loss": 0.4365,
+      "step": 10930
+    },
+    {
+      "epoch": 16.45,
+      "grad_norm": 3.3526134490966797,
+      "learning_rate": 8.354887218045113e-06,
+      "loss": 0.38,
+      "step": 10940
+    },
+    {
+      "epoch": 16.47,
+      "grad_norm": 4.425145149230957,
+      "learning_rate": 8.353383458646617e-06,
+      "loss": 0.2944,
+      "step": 10950
+    },
+    {
+      "epoch": 16.48,
+      "grad_norm": 4.0147552490234375,
+      "learning_rate": 8.35187969924812e-06,
+      "loss": 0.3836,
+      "step": 10960
+    },
+    {
+      "epoch": 16.5,
+      "grad_norm": 6.398830413818359,
+      "learning_rate": 8.350375939849624e-06,
+      "loss": 0.3402,
+      "step": 10970
+    },
+    {
+      "epoch": 16.51,
+      "grad_norm": 5.556189060211182,
+      "learning_rate": 8.348872180451129e-06,
+      "loss": 0.3728,
+      "step": 10980
+    },
+    {
+      "epoch": 16.53,
+      "grad_norm": 4.928891658782959,
+      "learning_rate": 8.347368421052633e-06,
+      "loss": 0.2993,
+      "step": 10990
+    },
+    {
+      "epoch": 16.54,
+      "grad_norm": 8.429544448852539,
+      "learning_rate": 8.345864661654136e-06,
+      "loss": 0.3896,
+      "step": 11000
+    },
+    {
+      "epoch": 16.56,
+      "grad_norm": 9.285510063171387,
+      "learning_rate": 8.34436090225564e-06,
+      "loss": 0.3428,
+      "step": 11010
+    },
+    {
+      "epoch": 16.57,
+      "grad_norm": 5.118491172790527,
+      "learning_rate": 8.342857142857143e-06,
+      "loss": 0.3583,
+      "step": 11020
+    },
+    {
+      "epoch": 16.59,
+      "grad_norm": 9.391587257385254,
+      "learning_rate": 8.341353383458647e-06,
+      "loss": 0.312,
+      "step": 11030
+    },
+    {
+      "epoch": 16.6,
+      "grad_norm": 22.18227767944336,
+      "learning_rate": 8.339849624060152e-06,
+      "loss": 0.3549,
+      "step": 11040
+    },
+    {
+      "epoch": 16.62,
+      "grad_norm": 5.834057331085205,
+      "learning_rate": 8.338345864661654e-06,
+      "loss": 0.4242,
+      "step": 11050
+    },
+    {
+      "epoch": 16.63,
+      "grad_norm": 5.136647701263428,
+      "learning_rate": 8.336842105263159e-06,
+      "loss": 0.3623,
+      "step": 11060
+    },
+    {
+      "epoch": 16.65,
+      "grad_norm": 5.481499195098877,
+      "learning_rate": 8.335338345864662e-06,
+      "loss": 0.3745,
+      "step": 11070
+    },
+    {
+      "epoch": 16.66,
+      "grad_norm": 3.6383814811706543,
+      "learning_rate": 8.333834586466166e-06,
+      "loss": 0.3526,
+      "step": 11080
+    },
+    {
+      "epoch": 16.68,
+      "grad_norm": 4.198364734649658,
+      "learning_rate": 8.33233082706767e-06,
+      "loss": 0.3247,
+      "step": 11090
+    },
+    {
+      "epoch": 16.69,
+      "grad_norm": 6.066871166229248,
+      "learning_rate": 8.330827067669175e-06,
+      "loss": 0.346,
+      "step": 11100
+    },
+    {
+      "epoch": 16.71,
+      "grad_norm": 4.822031497955322,
+      "learning_rate": 8.329323308270676e-06,
+      "loss": 0.3421,
+      "step": 11110
+    },
+    {
+      "epoch": 16.72,
+      "grad_norm": 8.0927152633667,
+      "learning_rate": 8.327819548872182e-06,
+      "loss": 0.3212,
+      "step": 11120
+    },
+    {
+      "epoch": 16.74,
+      "grad_norm": 5.737279415130615,
+      "learning_rate": 8.326315789473685e-06,
+      "loss": 0.338,
+      "step": 11130
+    },
+    {
+      "epoch": 16.75,
+      "grad_norm": 6.6013994216918945,
+      "learning_rate": 8.324812030075189e-06,
+      "loss": 0.4221,
+      "step": 11140
+    },
+    {
+      "epoch": 16.77,
+      "grad_norm": 3.22804856300354,
+      "learning_rate": 8.323308270676692e-06,
+      "loss": 0.2824,
+      "step": 11150
+    },
+    {
+      "epoch": 16.78,
+      "grad_norm": 7.42767333984375,
+      "learning_rate": 8.321804511278196e-06,
+      "loss": 0.3654,
+      "step": 11160
+    },
+    {
+      "epoch": 16.8,
+      "grad_norm": 6.393558979034424,
+      "learning_rate": 8.3203007518797e-06,
+      "loss": 0.3939,
+      "step": 11170
+    },
+    {
+      "epoch": 16.81,
+      "grad_norm": 6.1491241455078125,
+      "learning_rate": 8.318796992481204e-06,
+      "loss": 0.3454,
+      "step": 11180
+    },
+    {
+      "epoch": 16.83,
+      "grad_norm": 4.844079971313477,
+      "learning_rate": 8.317293233082708e-06,
+      "loss": 0.3744,
+      "step": 11190
+    },
+    {
+      "epoch": 16.84,
+      "grad_norm": 7.577675819396973,
+      "learning_rate": 8.315789473684212e-06,
+      "loss": 0.4259,
+      "step": 11200
+    },
+    {
+      "epoch": 16.86,
+      "grad_norm": 5.073234558105469,
+      "learning_rate": 8.314285714285715e-06,
+      "loss": 0.343,
+      "step": 11210
+    },
+    {
+      "epoch": 16.87,
+      "grad_norm": 4.934657096862793,
+      "learning_rate": 8.312781954887219e-06,
+      "loss": 0.3834,
+      "step": 11220
+    },
+    {
+      "epoch": 16.89,
+      "grad_norm": 4.744530200958252,
+      "learning_rate": 8.311278195488722e-06,
+      "loss": 0.2937,
+      "step": 11230
+    },
+    {
+      "epoch": 16.9,
+      "grad_norm": 7.569250583648682,
+      "learning_rate": 8.309774436090227e-06,
+      "loss": 0.3719,
+      "step": 11240
+    },
+    {
+      "epoch": 16.92,
+      "grad_norm": 7.076653480529785,
+      "learning_rate": 8.308270676691729e-06,
+      "loss": 0.3489,
+      "step": 11250
+    },
+    {
+      "epoch": 16.93,
+      "grad_norm": 7.1391520500183105,
+      "learning_rate": 8.306766917293234e-06,
+      "loss": 0.3325,
+      "step": 11260
+    },
+    {
+      "epoch": 16.95,
+      "grad_norm": 4.94738245010376,
+      "learning_rate": 8.305263157894738e-06,
+      "loss": 0.3421,
+      "step": 11270
+    },
+    {
+      "epoch": 16.96,
+      "grad_norm": 6.052053451538086,
+      "learning_rate": 8.303759398496241e-06,
+      "loss": 0.4267,
+      "step": 11280
+    },
+    {
+      "epoch": 16.98,
+      "grad_norm": 6.822144985198975,
+      "learning_rate": 8.302255639097745e-06,
+      "loss": 0.4932,
+      "step": 11290
+    },
+    {
+      "epoch": 16.99,
+      "grad_norm": 7.0719218254089355,
+      "learning_rate": 8.300751879699248e-06,
+      "loss": 0.3688,
+      "step": 11300
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.9346,
+      "eval_loss": 0.25758126378059387,
+      "eval_runtime": 84.7562,
+      "eval_samples_per_second": 117.986,
+      "eval_steps_per_second": 0.472,
+      "step": 11305
+    },
+    {
+      "epoch": 17.01,
+      "grad_norm": 4.741504669189453,
+      "learning_rate": 8.299248120300752e-06,
+      "loss": 0.3837,
+      "step": 11310
+    },
+    {
+      "epoch": 17.02,
+      "grad_norm": 6.912674427032471,
+      "learning_rate": 8.297744360902257e-06,
+      "loss": 0.2938,
+      "step": 11320
+    },
+    {
+      "epoch": 17.04,
+      "grad_norm": 5.881788730621338,
+      "learning_rate": 8.29624060150376e-06,
+      "loss": 0.3984,
+      "step": 11330
+    },
+    {
+      "epoch": 17.05,
+      "grad_norm": 5.705871105194092,
+      "learning_rate": 8.294736842105264e-06,
+      "loss": 0.3793,
+      "step": 11340
+    },
+    {
+      "epoch": 17.07,
+      "grad_norm": 5.036585330963135,
+      "learning_rate": 8.293233082706768e-06,
+      "loss": 0.365,
+      "step": 11350
+    },
+    {
+      "epoch": 17.08,
+      "grad_norm": 5.040714740753174,
+      "learning_rate": 8.291729323308271e-06,
+      "loss": 0.3662,
+      "step": 11360
+    },
+    {
+      "epoch": 17.1,
+      "grad_norm": 7.933087348937988,
+      "learning_rate": 8.290225563909775e-06,
+      "loss": 0.3204,
+      "step": 11370
+    },
+    {
+      "epoch": 17.11,
+      "grad_norm": 8.096324920654297,
+      "learning_rate": 8.28872180451128e-06,
+      "loss": 0.3383,
+      "step": 11380
+    },
+    {
+      "epoch": 17.13,
+      "grad_norm": 5.693844318389893,
+      "learning_rate": 8.287218045112782e-06,
+      "loss": 0.3455,
+      "step": 11390
+    },
+    {
+      "epoch": 17.14,
+      "grad_norm": 5.274537086486816,
+      "learning_rate": 8.285714285714287e-06,
+      "loss": 0.3695,
+      "step": 11400
+    },
+    {
+      "epoch": 17.16,
+      "grad_norm": 5.249573230743408,
+      "learning_rate": 8.28421052631579e-06,
+      "loss": 0.3801,
+      "step": 11410
+    },
+    {
+      "epoch": 17.17,
+      "grad_norm": 6.644190311431885,
+      "learning_rate": 8.282706766917294e-06,
+      "loss": 0.3355,
+      "step": 11420
+    },
+    {
+      "epoch": 17.19,
+      "grad_norm": 8.79143238067627,
+      "learning_rate": 8.281203007518798e-06,
+      "loss": 0.3857,
+      "step": 11430
+    },
+    {
+      "epoch": 17.2,
+      "grad_norm": 4.0490851402282715,
+      "learning_rate": 8.279699248120301e-06,
+      "loss": 0.3224,
+      "step": 11440
+    },
+    {
+      "epoch": 17.22,
+      "grad_norm": 8.35059928894043,
+      "learning_rate": 8.278195488721805e-06,
+      "loss": 0.372,
+      "step": 11450
+    },
+    {
+      "epoch": 17.23,
+      "grad_norm": 5.733313083648682,
+      "learning_rate": 8.276691729323308e-06,
+      "loss": 0.3511,
+      "step": 11460
+    },
+    {
+      "epoch": 17.25,
+      "grad_norm": 4.974298477172852,
+      "learning_rate": 8.275187969924813e-06,
+      "loss": 0.332,
+      "step": 11470
+    },
+    {
+      "epoch": 17.26,
+      "grad_norm": 6.859165191650391,
+      "learning_rate": 8.273684210526317e-06,
+      "loss": 0.3053,
+      "step": 11480
+    },
+    {
+      "epoch": 17.28,
+      "grad_norm": 7.4038591384887695,
+      "learning_rate": 8.27218045112782e-06,
+      "loss": 0.3419,
+      "step": 11490
+    },
+    {
+      "epoch": 17.29,
+      "grad_norm": 4.045393943786621,
+      "learning_rate": 8.270676691729324e-06,
+      "loss": 0.3656,
+      "step": 11500
+    },
+    {
+      "epoch": 17.31,
+      "grad_norm": 2.965898036956787,
+      "learning_rate": 8.269172932330827e-06,
+      "loss": 0.4222,
+      "step": 11510
+    },
+    {
+      "epoch": 17.32,
+      "grad_norm": 6.6445746421813965,
+      "learning_rate": 8.267669172932331e-06,
+      "loss": 0.4351,
+      "step": 11520
+    },
+    {
+      "epoch": 17.34,
+      "grad_norm": 11.220673561096191,
+      "learning_rate": 8.266165413533836e-06,
+      "loss": 0.3738,
+      "step": 11530
+    },
+    {
+      "epoch": 17.35,
+      "grad_norm": 3.7990376949310303,
+      "learning_rate": 8.264661654135338e-06,
+      "loss": 0.2779,
+      "step": 11540
+    },
+    {
+      "epoch": 17.37,
+      "grad_norm": 6.193857669830322,
+      "learning_rate": 8.263157894736843e-06,
+      "loss": 0.3864,
+      "step": 11550
+    },
+    {
+      "epoch": 17.38,
+      "grad_norm": 7.089908123016357,
+      "learning_rate": 8.261654135338347e-06,
+      "loss": 0.3243,
+      "step": 11560
+    },
+    {
+      "epoch": 17.4,
+      "grad_norm": 10.148313522338867,
+      "learning_rate": 8.26015037593985e-06,
+      "loss": 0.3848,
+      "step": 11570
+    },
+    {
+      "epoch": 17.41,
+      "grad_norm": 7.47261905670166,
+      "learning_rate": 8.258646616541354e-06,
+      "loss": 0.3958,
+      "step": 11580
+    },
+    {
+      "epoch": 17.43,
+      "grad_norm": 8.237654685974121,
+      "learning_rate": 8.257142857142857e-06,
+      "loss": 0.4104,
+      "step": 11590
+    },
+    {
+      "epoch": 17.44,
+      "grad_norm": 7.028960227966309,
+      "learning_rate": 8.25563909774436e-06,
+      "loss": 0.3515,
+      "step": 11600
+    },
+    {
+      "epoch": 17.46,
+      "grad_norm": 6.804955959320068,
+      "learning_rate": 8.254135338345866e-06,
+      "loss": 0.3555,
+      "step": 11610
+    },
+    {
+      "epoch": 17.47,
+      "grad_norm": 8.740710258483887,
+      "learning_rate": 8.25263157894737e-06,
+      "loss": 0.328,
+      "step": 11620
+    },
+    {
+      "epoch": 17.49,
+      "grad_norm": 6.383413314819336,
+      "learning_rate": 8.251127819548873e-06,
+      "loss": 0.3352,
+      "step": 11630
+    },
+    {
+      "epoch": 17.5,
+      "grad_norm": 8.289705276489258,
+      "learning_rate": 8.249624060150377e-06,
+      "loss": 0.4067,
+      "step": 11640
+    },
+    {
+      "epoch": 17.52,
+      "grad_norm": 6.0566911697387695,
+      "learning_rate": 8.24812030075188e-06,
+      "loss": 0.3556,
+      "step": 11650
+    },
+    {
+      "epoch": 17.53,
+      "grad_norm": 9.813027381896973,
+      "learning_rate": 8.246616541353384e-06,
+      "loss": 0.3856,
+      "step": 11660
+    },
+    {
+      "epoch": 17.55,
+      "grad_norm": 6.0970988273620605,
+      "learning_rate": 8.245112781954889e-06,
+      "loss": 0.396,
+      "step": 11670
+    },
+    {
+      "epoch": 17.56,
+      "grad_norm": 4.837037086486816,
+      "learning_rate": 8.24360902255639e-06,
+      "loss": 0.3194,
+      "step": 11680
+    },
+    {
+      "epoch": 17.58,
+      "grad_norm": 10.178328514099121,
+      "learning_rate": 8.242105263157896e-06,
+      "loss": 0.4108,
+      "step": 11690
+    },
+    {
+      "epoch": 17.59,
+      "grad_norm": 3.696746826171875,
+      "learning_rate": 8.2406015037594e-06,
+      "loss": 0.3198,
+      "step": 11700
+    },
+    {
+      "epoch": 17.61,
+      "grad_norm": 6.4773993492126465,
+      "learning_rate": 8.239097744360903e-06,
+      "loss": 0.2889,
+      "step": 11710
+    },
+    {
+      "epoch": 17.62,
+      "grad_norm": 9.036526679992676,
+      "learning_rate": 8.237593984962406e-06,
+      "loss": 0.4616,
+      "step": 11720
+    },
+    {
+      "epoch": 17.64,
+      "grad_norm": 5.1061320304870605,
+      "learning_rate": 8.236090225563912e-06,
+      "loss": 0.3941,
+      "step": 11730
+    },
+    {
+      "epoch": 17.65,
+      "grad_norm": 5.17496919631958,
+      "learning_rate": 8.234586466165413e-06,
+      "loss": 0.3883,
+      "step": 11740
+    },
+    {
+      "epoch": 17.67,
+      "grad_norm": 4.007594585418701,
+      "learning_rate": 8.233082706766919e-06,
+      "loss": 0.3097,
+      "step": 11750
+    },
+    {
+      "epoch": 17.68,
+      "grad_norm": 3.903956174850464,
+      "learning_rate": 8.231578947368422e-06,
+      "loss": 0.3473,
+      "step": 11760
+    },
+    {
+      "epoch": 17.7,
+      "grad_norm": 9.270066261291504,
+      "learning_rate": 8.230075187969926e-06,
+      "loss": 0.3255,
+      "step": 11770
+    },
+    {
+      "epoch": 17.71,
+      "grad_norm": 4.118042469024658,
+      "learning_rate": 8.22857142857143e-06,
+      "loss": 0.335,
+      "step": 11780
+    },
+    {
+      "epoch": 17.73,
+      "grad_norm": 5.715611457824707,
+      "learning_rate": 8.227067669172933e-06,
+      "loss": 0.422,
+      "step": 11790
+    },
+    {
+      "epoch": 17.74,
+      "grad_norm": 5.848507881164551,
+      "learning_rate": 8.225563909774436e-06,
+      "loss": 0.3465,
+      "step": 11800
+    },
+    {
+      "epoch": 17.76,
+      "grad_norm": 5.273082733154297,
+      "learning_rate": 8.224060150375941e-06,
+      "loss": 0.3479,
+      "step": 11810
+    },
+    {
+      "epoch": 17.77,
+      "grad_norm": 4.500287055969238,
+      "learning_rate": 8.222556390977445e-06,
+      "loss": 0.3813,
+      "step": 11820
+    },
+    {
+      "epoch": 17.79,
+      "grad_norm": 7.676726341247559,
+      "learning_rate": 8.221052631578948e-06,
+      "loss": 0.3973,
+      "step": 11830
+    },
+    {
+      "epoch": 17.8,
+      "grad_norm": 6.1550211906433105,
+      "learning_rate": 8.219548872180452e-06,
+      "loss": 0.3209,
+      "step": 11840
+    },
+    {
+      "epoch": 17.82,
+      "grad_norm": 3.796853542327881,
+      "learning_rate": 8.218045112781955e-06,
+      "loss": 0.3241,
+      "step": 11850
+    },
+    {
+      "epoch": 17.83,
+      "grad_norm": 7.188779354095459,
+      "learning_rate": 8.216541353383459e-06,
+      "loss": 0.3923,
+      "step": 11860
+    },
+    {
+      "epoch": 17.85,
+      "grad_norm": 5.088048934936523,
+      "learning_rate": 8.215037593984964e-06,
+      "loss": 0.3528,
+      "step": 11870
+    },
+    {
+      "epoch": 17.86,
+      "grad_norm": 6.49263334274292,
+      "learning_rate": 8.213533834586466e-06,
+      "loss": 0.3724,
+      "step": 11880
+    },
+    {
+      "epoch": 17.88,
+      "grad_norm": 8.370095252990723,
+      "learning_rate": 8.212030075187971e-06,
+      "loss": 0.3697,
+      "step": 11890
+    },
+    {
+      "epoch": 17.89,
+      "grad_norm": 7.578341007232666,
+      "learning_rate": 8.210526315789475e-06,
+      "loss": 0.3317,
+      "step": 11900
+    },
+    {
+      "epoch": 17.91,
+      "grad_norm": 5.1709723472595215,
+      "learning_rate": 8.209022556390978e-06,
+      "loss": 0.3304,
+      "step": 11910
+    },
+    {
+      "epoch": 17.92,
+      "grad_norm": 4.586398124694824,
+      "learning_rate": 8.207518796992482e-06,
+      "loss": 0.3942,
+      "step": 11920
+    },
+    {
+      "epoch": 17.94,
+      "grad_norm": 3.6240298748016357,
+      "learning_rate": 8.206015037593985e-06,
+      "loss": 0.2839,
+      "step": 11930
+    },
+    {
+      "epoch": 17.95,
+      "grad_norm": 7.672499179840088,
+      "learning_rate": 8.204511278195489e-06,
+      "loss": 0.3539,
+      "step": 11940
+    },
+    {
+      "epoch": 17.97,
+      "grad_norm": 5.807362079620361,
+      "learning_rate": 8.203007518796992e-06,
+      "loss": 0.3006,
+      "step": 11950
+    },
+    {
+      "epoch": 17.98,
+      "grad_norm": 5.875560283660889,
+      "learning_rate": 8.201503759398498e-06,
+      "loss": 0.3857,
+      "step": 11960
+    },
+    {
+      "epoch": 18.0,
+      "grad_norm": 47.7349967956543,
+      "learning_rate": 8.2e-06,
+      "loss": 0.3985,
+      "step": 11970
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.9356,
+      "eval_loss": 0.25322866439819336,
+      "eval_runtime": 84.6682,
+      "eval_samples_per_second": 118.108,
+      "eval_steps_per_second": 0.472,
+      "step": 11970
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 66500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 100,
+  "save_steps": 500,
+  "total_flos": 2.0962905086398464e+20,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}