diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,10213 @@
+{
+  "best_metric": 0.9940476190476191,
+  "best_model_checkpoint": "vit-base-patch16-224-finetuned-main-gpu-30e-final/checkpoint-15979",
+  "epoch": 29.99909338168631,
+  "global_step": 16530,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.0248033877797943e-07,
+      "loss": 1.4755,
+      "step": 10
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.049606775559589e-07,
+      "loss": 1.4788,
+      "step": 20
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.074410163339384e-07,
+      "loss": 1.4525,
+      "step": 30
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.2099213551119177e-06,
+      "loss": 1.3702,
+      "step": 40
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.5124016938898973e-06,
+      "loss": 1.3018,
+      "step": 50
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.8148820326678768e-06,
+      "loss": 1.2006,
+      "step": 60
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2.1173623714458564e-06,
+      "loss": 1.1279,
+      "step": 70
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4198427102238355e-06,
+      "loss": 1.0402,
+      "step": 80
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2.722323049001815e-06,
+      "loss": 0.9743,
+      "step": 90
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.0248033877797946e-06,
+      "loss": 0.9197,
+      "step": 100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.327283726557774e-06,
+      "loss": 0.876,
+      "step": 110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.6297640653357536e-06,
+      "loss": 0.8733,
+      "step": 120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.932244404113733e-06,
+      "loss": 0.8408,
+      "step": 130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.234724742891713e-06,
+      "loss": 0.8088,
+      "step": 140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.537205081669692e-06,
+      "loss": 0.7937,
+      "step": 150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.839685420447671e-06,
+      "loss": 0.7814,
+      "step": 160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 5.142165759225651e-06,
+      "loss": 0.7276,
+      "step": 170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 5.44464609800363e-06,
+      "loss": 0.7215,
+      "step": 180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 5.747126436781609e-06,
+      "loss": 0.7441,
+      "step": 190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 6.049606775559589e-06,
+      "loss": 0.7327,
+      "step": 200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.352087114337568e-06,
+      "loss": 0.7344,
+      "step": 210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.654567453115548e-06,
+      "loss": 0.7162,
+      "step": 220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.957047791893527e-06,
+      "loss": 0.6783,
+      "step": 230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 7.259528130671507e-06,
+      "loss": 0.6718,
+      "step": 240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.5620084694494855e-06,
+      "loss": 0.6663,
+      "step": 250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 7.864488808227466e-06,
+      "loss": 0.6409,
+      "step": 260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 8.166969147005445e-06,
+      "loss": 0.6604,
+      "step": 270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 8.469449485783425e-06,
+      "loss": 0.6318,
+      "step": 280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 8.771929824561403e-06,
+      "loss": 0.6526,
+      "step": 290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.074410163339384e-06,
+      "loss": 0.6502,
+      "step": 300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.376890502117363e-06,
+      "loss": 0.6449,
+      "step": 310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.679370840895342e-06,
+      "loss": 0.601,
+      "step": 320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.981851179673321e-06,
+      "loss": 0.5727,
+      "step": 330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.0284331518451302e-05,
+      "loss": 0.6081,
+      "step": 340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.0586811857229281e-05,
+      "loss": 0.5953,
+      "step": 350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.088929219600726e-05,
+      "loss": 0.5768,
+      "step": 360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.1191772534785241e-05,
+      "loss": 0.5715,
+      "step": 370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.1494252873563218e-05,
+      "loss": 0.5973,
+      "step": 380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.1796733212341199e-05,
+      "loss": 0.5973,
+      "step": 390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.2099213551119178e-05,
+      "loss": 0.5756,
+      "step": 400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2401693889897157e-05,
+      "loss": 0.5622,
+      "step": 410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2704174228675136e-05,
+      "loss": 0.596,
+      "step": 420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.3006654567453117e-05,
+      "loss": 0.5729,
+      "step": 430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.3309134906231096e-05,
+      "loss": 0.5337,
+      "step": 440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.3611615245009074e-05,
+      "loss": 0.5514,
+      "step": 450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3914095583787055e-05,
+      "loss": 0.5525,
+      "step": 460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4216575922565034e-05,
+      "loss": 0.5209,
+      "step": 470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4519056261343015e-05,
+      "loss": 0.5393,
+      "step": 480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4821536600120992e-05,
+      "loss": 0.5156,
+      "step": 490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.5124016938898971e-05,
+      "loss": 0.5396,
+      "step": 500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.542649727767695e-05,
+      "loss": 0.5734,
+      "step": 510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.572897761645493e-05,
+      "loss": 0.5133,
+      "step": 520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6031457955232912e-05,
+      "loss": 0.4997,
+      "step": 530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.633393829401089e-05,
+      "loss": 0.5118,
+      "step": 540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.663641863278887e-05,
+      "loss": 0.5113,
+      "step": 550
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.797108843537415,
+      "eval_loss": 0.4745405316352844,
+      "eval_runtime": 88.6392,
+      "eval_samples_per_second": 132.673,
+      "eval_steps_per_second": 4.152,
+      "step": 551
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.693889897156685e-05,
+      "loss": 0.5527,
+      "step": 560
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.7241379310344828e-05,
+      "loss": 0.4961,
+      "step": 570
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.7543859649122806e-05,
+      "loss": 0.4813,
+      "step": 580
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.7846339987900787e-05,
+      "loss": 0.4942,
+      "step": 590
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.8148820326678767e-05,
+      "loss": 0.5071,
+      "step": 600
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.8451300665456748e-05,
+      "loss": 0.49,
+      "step": 610
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.8753781004234726e-05,
+      "loss": 0.4658,
+      "step": 620
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.9056261343012703e-05,
+      "loss": 0.4801,
+      "step": 630
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.9358741681790684e-05,
+      "loss": 0.4607,
+      "step": 640
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.9661222020568665e-05,
+      "loss": 0.4489,
+      "step": 650
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.9963702359346642e-05,
+      "loss": 0.4627,
+      "step": 660
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2.0266182698124623e-05,
+      "loss": 0.465,
+      "step": 670
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2.0568663036902604e-05,
+      "loss": 0.4495,
+      "step": 680
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2.087114337568058e-05,
+      "loss": 0.4746,
+      "step": 690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2.1173623714458562e-05,
+      "loss": 0.4774,
+      "step": 700
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2.147610405323654e-05,
+      "loss": 0.4397,
+      "step": 710
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.177858439201452e-05,
+      "loss": 0.4547,
+      "step": 720
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.20810647307925e-05,
+      "loss": 0.4415,
+      "step": 730
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.2383545069570482e-05,
+      "loss": 0.4453,
+      "step": 740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.2686025408348456e-05,
+      "loss": 0.464,
+      "step": 750
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.2988505747126437e-05,
+      "loss": 0.3787,
+      "step": 760
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.3290986085904417e-05,
+      "loss": 0.4344,
+      "step": 770
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.3593466424682398e-05,
+      "loss": 0.426,
+      "step": 780
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3895946763460376e-05,
+      "loss": 0.4089,
+      "step": 790
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.4198427102238356e-05,
+      "loss": 0.4195,
+      "step": 800
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.4500907441016334e-05,
+      "loss": 0.4056,
+      "step": 810
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.4803387779794315e-05,
+      "loss": 0.4241,
+      "step": 820
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.5105868118572295e-05,
+      "loss": 0.3987,
+      "step": 830
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.5408348457350273e-05,
+      "loss": 0.4178,
+      "step": 840
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.571082879612825e-05,
+      "loss": 0.4039,
+      "step": 850
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.6013309134906235e-05,
+      "loss": 0.428,
+      "step": 860
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 0.4004,
+      "step": 870
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.6618269812462193e-05,
+      "loss": 0.4399,
+      "step": 880
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.692075015124017e-05,
+      "loss": 0.414,
+      "step": 890
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.7223230490018148e-05,
+      "loss": 0.3833,
+      "step": 900
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.7525710828796132e-05,
+      "loss": 0.4077,
+      "step": 910
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.782819116757411e-05,
+      "loss": 0.404,
+      "step": 920
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.813067150635209e-05,
+      "loss": 0.3873,
+      "step": 930
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.8433151845130067e-05,
+      "loss": 0.3897,
+      "step": 940
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8735632183908045e-05,
+      "loss": 0.4044,
+      "step": 950
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.903811252268603e-05,
+      "loss": 0.3742,
+      "step": 960
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.9340592861464007e-05,
+      "loss": 0.3996,
+      "step": 970
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.9643073200241984e-05,
+      "loss": 0.3959,
+      "step": 980
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.9945553539019965e-05,
+      "loss": 0.3588,
+      "step": 990
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.0248033877797942e-05,
+      "loss": 0.3762,
+      "step": 1000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.0550514216575926e-05,
+      "loss": 0.3864,
+      "step": 1010
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.08529945553539e-05,
+      "loss": 0.3552,
+      "step": 1020
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.115547489413188e-05,
+      "loss": 0.3519,
+      "step": 1030
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.145795523290986e-05,
+      "loss": 0.3989,
+      "step": 1040
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.176043557168784e-05,
+      "loss": 0.3591,
+      "step": 1050
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.2062915910465824e-05,
+      "loss": 0.3787,
+      "step": 1060
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.2365396249243804e-05,
+      "loss": 0.3306,
+      "step": 1070
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.266787658802178e-05,
+      "loss": 0.3387,
+      "step": 1080
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.297035692679976e-05,
+      "loss": 0.359,
+      "step": 1090
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.327283726557774e-05,
+      "loss": 0.3409,
+      "step": 1100
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8960884353741496,
+      "eval_loss": 0.26969781517982483,
+      "eval_runtime": 90.8852,
+      "eval_samples_per_second": 129.394,
+      "eval_steps_per_second": 4.049,
+      "step": 1102
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.3575317604355714e-05,
+      "loss": 0.3485,
+      "step": 1110
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.38777979431337e-05,
+      "loss": 0.3683,
+      "step": 1120
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.4180278281911676e-05,
+      "loss": 0.3749,
+      "step": 1130
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.4482758620689657e-05,
+      "loss": 0.3904,
+      "step": 1140
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.478523895946764e-05,
+      "loss": 0.3264,
+      "step": 1150
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.508771929824561e-05,
+      "loss": 0.3502,
+      "step": 1160
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.53901996370236e-05,
+      "loss": 0.3429,
+      "step": 1170
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.569267997580157e-05,
+      "loss": 0.3766,
+      "step": 1180
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.5995160314579554e-05,
+      "loss": 0.3469,
+      "step": 1190
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.6297640653357535e-05,
+      "loss": 0.3546,
+      "step": 1200
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.660012099213551e-05,
+      "loss": 0.3374,
+      "step": 1210
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.6902601330913496e-05,
+      "loss": 0.3609,
+      "step": 1220
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.720508166969147e-05,
+      "loss": 0.3443,
+      "step": 1230
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.750756200846945e-05,
+      "loss": 0.3053,
+      "step": 1240
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.781004234724743e-05,
+      "loss": 0.323,
+      "step": 1250
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.8112522686025406e-05,
+      "loss": 0.3398,
+      "step": 1260
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.8415003024803394e-05,
+      "loss": 0.305,
+      "step": 1270
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.871748336358137e-05,
+      "loss": 0.3456,
+      "step": 1280
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.901996370235935e-05,
+      "loss": 0.3201,
+      "step": 1290
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.932244404113733e-05,
+      "loss": 0.3078,
+      "step": 1300
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.962492437991531e-05,
+      "loss": 0.3326,
+      "step": 1310
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9927404718693284e-05,
+      "loss": 0.2847,
+      "step": 1320
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.0229885057471265e-05,
+      "loss": 0.3121,
+      "step": 1330
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.0532365396249246e-05,
+      "loss": 0.28,
+      "step": 1340
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.0834845735027227e-05,
+      "loss": 0.3172,
+      "step": 1350
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.113732607380521e-05,
+      "loss": 0.336,
+      "step": 1360
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.143980641258318e-05,
+      "loss": 0.3164,
+      "step": 1370
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.174228675136116e-05,
+      "loss": 0.3434,
+      "step": 1380
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.204476709013914e-05,
+      "loss": 0.3078,
+      "step": 1390
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.2347247428917124e-05,
+      "loss": 0.2977,
+      "step": 1400
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.2649727767695105e-05,
+      "loss": 0.3033,
+      "step": 1410
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.295220810647308e-05,
+      "loss": 0.326,
+      "step": 1420
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.325468844525106e-05,
+      "loss": 0.2928,
+      "step": 1430
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.355716878402904e-05,
+      "loss": 0.2733,
+      "step": 1440
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.3859649122807014e-05,
+      "loss": 0.2843,
+      "step": 1450
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.4162129461585e-05,
+      "loss": 0.321,
+      "step": 1460
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.4464609800362976e-05,
+      "loss": 0.3248,
+      "step": 1470
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.4767090139140964e-05,
+      "loss": 0.321,
+      "step": 1480
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.506957047791894e-05,
+      "loss": 0.2956,
+      "step": 1490
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.537205081669691e-05,
+      "loss": 0.2895,
+      "step": 1500
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.56745311554749e-05,
+      "loss": 0.3339,
+      "step": 1510
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.597701149425287e-05,
+      "loss": 0.3157,
+      "step": 1520
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.6279491833030854e-05,
+      "loss": 0.2886,
+      "step": 1530
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.6581972171808835e-05,
+      "loss": 0.2915,
+      "step": 1540
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.6884452510586816e-05,
+      "loss": 0.2604,
+      "step": 1550
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.7186932849364796e-05,
+      "loss": 0.2793,
+      "step": 1560
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.748941318814277e-05,
+      "loss": 0.2717,
+      "step": 1570
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.779189352692075e-05,
+      "loss": 0.2754,
+      "step": 1580
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.809437386569873e-05,
+      "loss": 0.2763,
+      "step": 1590
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.839685420447671e-05,
+      "loss": 0.2856,
+      "step": 1600
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.8699334543254694e-05,
+      "loss": 0.2469,
+      "step": 1610
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.900181488203267e-05,
+      "loss": 0.3344,
+      "step": 1620
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.930429522081065e-05,
+      "loss": 0.278,
+      "step": 1630
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.960677555958863e-05,
+      "loss": 0.2579,
+      "step": 1640
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.990925589836661e-05,
+      "loss": 0.2675,
+      "step": 1650
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9380952380952381,
+      "eval_loss": 0.16112102568149567,
+      "eval_runtime": 89.9897,
+      "eval_samples_per_second": 130.682,
+      "eval_steps_per_second": 4.089,
+      "step": 1653
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.9976473751428384e-05,
+      "loss": 0.2925,
+      "step": 1660
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.99428648248975e-05,
+      "loss": 0.2836,
+      "step": 1670
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.990925589836661e-05,
+      "loss": 0.2793,
+      "step": 1680
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.9875646971835723e-05,
+      "loss": 0.2755,
+      "step": 1690
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.984203804530484e-05,
+      "loss": 0.2517,
+      "step": 1700
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.980842911877395e-05,
+      "loss": 0.2751,
+      "step": 1710
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.977482019224306e-05,
+      "loss": 0.2478,
+      "step": 1720
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.9741211265712176e-05,
+      "loss": 0.2766,
+      "step": 1730
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.970760233918128e-05,
+      "loss": 0.3006,
+      "step": 1740
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.96739934126504e-05,
+      "loss": 0.2907,
+      "step": 1750
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.9640384486119516e-05,
+      "loss": 0.276,
+      "step": 1760
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.960677555958863e-05,
+      "loss": 0.2185,
+      "step": 1770
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.957316663305774e-05,
+      "loss": 0.2669,
+      "step": 1780
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.9539557706526856e-05,
+      "loss": 0.2623,
+      "step": 1790
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.950594877999597e-05,
+      "loss": 0.2575,
+      "step": 1800
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.947233985346508e-05,
+      "loss": 0.2391,
+      "step": 1810
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.9438730926934196e-05,
+      "loss": 0.216,
+      "step": 1820
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.940512200040331e-05,
+      "loss": 0.2521,
+      "step": 1830
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.937151307387242e-05,
+      "loss": 0.2695,
+      "step": 1840
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.9337904147341535e-05,
+      "loss": 0.263,
+      "step": 1850
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.930429522081065e-05,
+      "loss": 0.2412,
+      "step": 1860
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.927068629427976e-05,
+      "loss": 0.2408,
+      "step": 1870
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.9237077367748875e-05,
+      "loss": 0.2595,
+      "step": 1880
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.9203468441217995e-05,
+      "loss": 0.2625,
+      "step": 1890
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.916985951468711e-05,
+      "loss": 0.2316,
+      "step": 1900
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.913625058815622e-05,
+      "loss": 0.2282,
+      "step": 1910
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.9102641661625335e-05,
+      "loss": 0.2574,
+      "step": 1920
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.906903273509444e-05,
+      "loss": 0.2076,
+      "step": 1930
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.9035423808563555e-05,
+      "loss": 0.2334,
+      "step": 1940
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.900181488203267e-05,
+      "loss": 0.2576,
+      "step": 1950
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.896820595550178e-05,
+      "loss": 0.22,
+      "step": 1960
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.8934597028970894e-05,
+      "loss": 0.2475,
+      "step": 1970
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.890098810244001e-05,
+      "loss": 0.2429,
+      "step": 1980
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.886737917590912e-05,
+      "loss": 0.2365,
+      "step": 1990
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.8833770249378234e-05,
+      "loss": 0.2283,
+      "step": 2000
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.880016132284735e-05,
+      "loss": 0.2326,
+      "step": 2010
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.876655239631646e-05,
+      "loss": 0.2373,
+      "step": 2020
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.8732943469785574e-05,
+      "loss": 0.2868,
+      "step": 2030
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.8699334543254694e-05,
+      "loss": 0.2698,
+      "step": 2040
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.866572561672381e-05,
+      "loss": 0.2316,
+      "step": 2050
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.863211669019292e-05,
+      "loss": 0.2225,
+      "step": 2060
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.8598507763662033e-05,
+      "loss": 0.1972,
+      "step": 2070
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.856489883713115e-05,
+      "loss": 0.2362,
+      "step": 2080
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.853128991060026e-05,
+      "loss": 0.2141,
+      "step": 2090
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.849768098406937e-05,
+      "loss": 0.2187,
+      "step": 2100
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.8464072057538486e-05,
+      "loss": 0.2326,
+      "step": 2110
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.84304631310076e-05,
+      "loss": 0.2347,
+      "step": 2120
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.839685420447671e-05,
+      "loss": 0.1873,
+      "step": 2130
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.8363245277945826e-05,
+      "loss": 0.2236,
+      "step": 2140
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.832963635141494e-05,
+      "loss": 0.2093,
+      "step": 2150
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.829602742488405e-05,
+      "loss": 0.231,
+      "step": 2160
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.826241849835316e-05,
+      "loss": 0.2064,
+      "step": 2170
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.822880957182228e-05,
+      "loss": 0.2123,
+      "step": 2180
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.819520064529139e-05,
+      "loss": 0.2217,
+      "step": 2190
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.8161591718760506e-05,
+      "loss": 0.2092,
+      "step": 2200
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.9547619047619048,
+      "eval_loss": 0.11763259768486023,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 133.37,
+      "eval_steps_per_second": 4.173,
+      "step": 2204
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.812798279222962e-05,
+      "loss": 0.2711,
+      "step": 2210
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.809437386569873e-05,
+      "loss": 0.2299,
+      "step": 2220
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 4.8060764939167845e-05,
+      "loss": 0.2254,
+      "step": 2230
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.802715601263696e-05,
+      "loss": 0.2042,
+      "step": 2240
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 4.799354708610607e-05,
+      "loss": 0.204,
+      "step": 2250
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 4.7959938159575185e-05,
+      "loss": 0.2221,
+      "step": 2260
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.79263292330443e-05,
+      "loss": 0.1991,
+      "step": 2270
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 4.789272030651341e-05,
+      "loss": 0.1889,
+      "step": 2280
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 4.7859111379982525e-05,
+      "loss": 0.1964,
+      "step": 2290
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 4.782550245345164e-05,
+      "loss": 0.2147,
+      "step": 2300
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 4.779189352692075e-05,
+      "loss": 0.205,
+      "step": 2310
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 4.7758284600389865e-05,
+      "loss": 0.1962,
+      "step": 2320
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 4.7724675673858985e-05,
+      "loss": 0.2012,
+      "step": 2330
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.76910667473281e-05,
+      "loss": 0.1847,
+      "step": 2340
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 4.765745782079721e-05,
+      "loss": 0.2092,
+      "step": 2350
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.762384889426632e-05,
+      "loss": 0.2018,
+      "step": 2360
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.759023996773543e-05,
+      "loss": 0.1775,
+      "step": 2370
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.7556631041204544e-05,
+      "loss": 0.2096,
+      "step": 2380
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.752302211467366e-05,
+      "loss": 0.2224,
+      "step": 2390
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.748941318814277e-05,
+      "loss": 0.2143,
+      "step": 2400
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.7455804261611884e-05,
+      "loss": 0.1861,
+      "step": 2410
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 4.7422195335081e-05,
+      "loss": 0.1721,
+      "step": 2420
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.738858640855011e-05,
+      "loss": 0.2087,
+      "step": 2430
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.7354977482019223e-05,
+      "loss": 0.1982,
+      "step": 2440
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 4.732136855548834e-05,
+      "loss": 0.2056,
+      "step": 2450
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.728775962895745e-05,
+      "loss": 0.1869,
+      "step": 2460
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 4.725415070242657e-05,
+      "loss": 0.2016,
+      "step": 2470
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 4.722054177589568e-05,
+      "loss": 0.2045,
+      "step": 2480
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 4.7186932849364796e-05,
+      "loss": 0.1919,
+      "step": 2490
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.715332392283391e-05,
+      "loss": 0.1956,
+      "step": 2500
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 4.711971499630302e-05,
+      "loss": 0.1939,
+      "step": 2510
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.7086106069772136e-05,
+      "loss": 0.1851,
+      "step": 2520
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.705249714324125e-05,
+      "loss": 0.1792,
+      "step": 2530
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.701888821671036e-05,
+      "loss": 0.1764,
+      "step": 2540
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 4.6985279290179476e-05,
+      "loss": 0.1948,
+      "step": 2550
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 4.695167036364859e-05,
+      "loss": 0.2128,
+      "step": 2560
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.69180614371177e-05,
+      "loss": 0.1859,
+      "step": 2570
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.6884452510586816e-05,
+      "loss": 0.1803,
+      "step": 2580
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.685084358405593e-05,
+      "loss": 0.1947,
+      "step": 2590
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.6817234657525035e-05,
+      "loss": 0.1657,
+      "step": 2600
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.678362573099415e-05,
+      "loss": 0.219,
+      "step": 2610
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.675001680446327e-05,
+      "loss": 0.1789,
+      "step": 2620
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 4.671640787793238e-05,
+      "loss": 0.2259,
+      "step": 2630
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.6682798951401495e-05,
+      "loss": 0.1959,
+      "step": 2640
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 4.664919002487061e-05,
+      "loss": 0.1877,
+      "step": 2650
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 4.661558109833972e-05,
+      "loss": 0.2103,
+      "step": 2660
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 4.6581972171808835e-05,
+      "loss": 0.1796,
+      "step": 2670
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 4.654836324527795e-05,
+      "loss": 0.2199,
+      "step": 2680
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.651475431874706e-05,
+      "loss": 0.1999,
+      "step": 2690
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.6481145392216175e-05,
+      "loss": 0.2071,
+      "step": 2700
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 4.644753646568529e-05,
+      "loss": 0.1893,
+      "step": 2710
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.64139275391544e-05,
+      "loss": 0.1842,
+      "step": 2720
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.6380318612623514e-05,
+      "loss": 0.211,
+      "step": 2730
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.634670968609263e-05,
+      "loss": 0.1775,
+      "step": 2740
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 4.631310075956174e-05,
+      "loss": 0.2008,
+      "step": 2750
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9655612244897959,
+      "eval_loss": 0.088911272585392,
+      "eval_runtime": 88.6408,
+      "eval_samples_per_second": 132.67,
+      "eval_steps_per_second": 4.152,
+      "step": 2755
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.6279491833030854e-05,
+      "loss": 0.2001,
+      "step": 2760
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.6245882906499974e-05,
+      "loss": 0.1795,
+      "step": 2770
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.621227397996909e-05,
+      "loss": 0.1639,
+      "step": 2780
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.6178665053438194e-05,
+      "loss": 0.1683,
+      "step": 2790
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.614505612690731e-05,
+      "loss": 0.1578,
+      "step": 2800
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.611144720037642e-05,
+      "loss": 0.2213,
+      "step": 2810
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.6077838273845533e-05,
+      "loss": 0.1749,
+      "step": 2820
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.604422934731465e-05,
+      "loss": 0.19,
+      "step": 2830
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.601062042078376e-05,
+      "loss": 0.1916,
+      "step": 2840
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.597701149425287e-05,
+      "loss": 0.1898,
+      "step": 2850
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.5943402567721986e-05,
+      "loss": 0.1825,
+      "step": 2860
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.59097936411911e-05,
+      "loss": 0.1549,
+      "step": 2870
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.587618471466021e-05,
+      "loss": 0.1869,
+      "step": 2880
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.5842575788129326e-05,
+      "loss": 0.1831,
+      "step": 2890
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.580896686159844e-05,
+      "loss": 0.1733,
+      "step": 2900
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.577535793506756e-05,
+      "loss": 0.1761,
+      "step": 2910
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.574174900853667e-05,
+      "loss": 0.1593,
+      "step": 2920
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.5708140082005786e-05,
+      "loss": 0.1519,
+      "step": 2930
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.56745311554749e-05,
+      "loss": 0.1691,
+      "step": 2940
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.564092222894401e-05,
+      "loss": 0.18,
+      "step": 2950
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.5607313302413126e-05,
+      "loss": 0.1723,
+      "step": 2960
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.557370437588224e-05,
+      "loss": 0.1812,
+      "step": 2970
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.554009544935135e-05,
+      "loss": 0.1448,
+      "step": 2980
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5506486522820465e-05,
+      "loss": 0.1706,
+      "step": 2990
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.547287759628958e-05,
+      "loss": 0.1806,
+      "step": 3000
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.543926866975869e-05,
+      "loss": 0.1666,
+      "step": 3010
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.5405659743227805e-05,
+      "loss": 0.1636,
+      "step": 3020
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.537205081669691e-05,
+      "loss": 0.1677,
+      "step": 3030
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.5338441890166025e-05,
+      "loss": 0.1727,
+      "step": 3040
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.530483296363514e-05,
+      "loss": 0.1694,
+      "step": 3050
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.527122403710426e-05,
+      "loss": 0.1373,
+      "step": 3060
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.523761511057337e-05,
+      "loss": 0.1584,
+      "step": 3070
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.5204006184042485e-05,
+      "loss": 0.2222,
+      "step": 3080
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.51703972575116e-05,
+      "loss": 0.1698,
+      "step": 3090
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.513678833098071e-05,
+      "loss": 0.1752,
+      "step": 3100
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.5103179404449824e-05,
+      "loss": 0.1613,
+      "step": 3110
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.506957047791894e-05,
+      "loss": 0.171,
+      "step": 3120
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.503596155138805e-05,
+      "loss": 0.1553,
+      "step": 3130
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.5002352624857164e-05,
+      "loss": 0.1386,
+      "step": 3140
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.496874369832628e-05,
+      "loss": 0.1477,
+      "step": 3150
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.493513477179539e-05,
+      "loss": 0.1855,
+      "step": 3160
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.4901525845264504e-05,
+      "loss": 0.1883,
+      "step": 3170
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.486791691873362e-05,
+      "loss": 0.1567,
+      "step": 3180
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.483430799220273e-05,
+      "loss": 0.1638,
+      "step": 3190
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.480069906567185e-05,
+      "loss": 0.166,
+      "step": 3200
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.4767090139140964e-05,
+      "loss": 0.1547,
+      "step": 3210
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.473348121261007e-05,
+      "loss": 0.1818,
+      "step": 3220
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.469987228607918e-05,
+      "loss": 0.1971,
+      "step": 3230
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.4666263359548296e-05,
+      "loss": 0.1431,
+      "step": 3240
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.463265443301741e-05,
+      "loss": 0.1535,
+      "step": 3250
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.459904550648652e-05,
+      "loss": 0.1746,
+      "step": 3260
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.4565436579955636e-05,
+      "loss": 0.1668,
+      "step": 3270
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.453182765342475e-05,
+      "loss": 0.1607,
+      "step": 3280
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.449821872689386e-05,
+      "loss": 0.1688,
+      "step": 3290
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.4464609800362976e-05,
+      "loss": 0.1555,
+      "step": 3300
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.9758503401360544,
+      "eval_loss": 0.06657847762107849,
+      "eval_runtime": 90.9074,
+      "eval_samples_per_second": 129.362,
+      "eval_steps_per_second": 4.048,
+      "step": 3306
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.443100087383209e-05,
+      "loss": 0.2004,
+      "step": 3310
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.43973919473012e-05,
+      "loss": 0.1803,
+      "step": 3320
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.4363783020770316e-05,
+      "loss": 0.1602,
+      "step": 3330
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.433017409423943e-05,
+      "loss": 0.1455,
+      "step": 3340
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.429656516770855e-05,
+      "loss": 0.1604,
+      "step": 3350
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.426295624117766e-05,
+      "loss": 0.1555,
+      "step": 3360
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.4229347314646775e-05,
+      "loss": 0.1832,
+      "step": 3370
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.419573838811589e-05,
+      "loss": 0.1599,
+      "step": 3380
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.4162129461585e-05,
+      "loss": 0.1567,
+      "step": 3390
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.4128520535054115e-05,
+      "loss": 0.1451,
+      "step": 3400
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.409491160852323e-05,
+      "loss": 0.1649,
+      "step": 3410
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.406130268199234e-05,
+      "loss": 0.1749,
+      "step": 3420
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.4027693755461455e-05,
+      "loss": 0.1554,
+      "step": 3430
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.399408482893057e-05,
+      "loss": 0.1618,
+      "step": 3440
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.396047590239968e-05,
+      "loss": 0.1495,
+      "step": 3450
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.392686697586879e-05,
+      "loss": 0.1577,
+      "step": 3460
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.38932580493379e-05,
+      "loss": 0.1686,
+      "step": 3470
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.3859649122807014e-05,
+      "loss": 0.1512,
+      "step": 3480
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 4.3826040196276134e-05,
+      "loss": 0.142,
+      "step": 3490
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.379243126974525e-05,
+      "loss": 0.1296,
+      "step": 3500
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.375882234321436e-05,
+      "loss": 0.1749,
+      "step": 3510
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.3725213416683474e-05,
+      "loss": 0.1287,
+      "step": 3520
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 4.369160449015259e-05,
+      "loss": 0.1591,
+      "step": 3530
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 4.36579955636217e-05,
+      "loss": 0.1646,
+      "step": 3540
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.3624386637090814e-05,
+      "loss": 0.1669,
+      "step": 3550
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 4.359077771055993e-05,
+      "loss": 0.1745,
+      "step": 3560
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 4.355716878402904e-05,
+      "loss": 0.1339,
+      "step": 3570
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.3523559857498154e-05,
+      "loss": 0.1505,
+      "step": 3580
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 4.348995093096727e-05,
+      "loss": 0.1501,
+      "step": 3590
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 4.345634200443638e-05,
+      "loss": 0.1425,
+      "step": 3600
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 4.342273307790549e-05,
+      "loss": 0.1381,
+      "step": 3610
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 4.3389124151374606e-05,
+      "loss": 0.1773,
+      "step": 3620
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 4.335551522484372e-05,
+      "loss": 0.1831,
+      "step": 3630
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 4.332190629831284e-05,
+      "loss": 0.1729,
+      "step": 3640
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 4.3288297371781946e-05,
+      "loss": 0.1337,
+      "step": 3650
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 4.325468844525106e-05,
+      "loss": 0.1339,
+      "step": 3660
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 4.322107951872017e-05,
+      "loss": 0.1695,
+      "step": 3670
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.3187470592189286e-05,
+      "loss": 0.1652,
+      "step": 3680
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 4.31538616656584e-05,
+      "loss": 0.1709,
+      "step": 3690
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.312025273912751e-05,
+      "loss": 0.1514,
+      "step": 3700
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.3086643812596626e-05,
+      "loss": 0.1344,
+      "step": 3710
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 4.305303488606574e-05,
+      "loss": 0.151,
+      "step": 3720
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 4.301942595953485e-05,
+      "loss": 0.1602,
+      "step": 3730
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 4.2985817033003965e-05,
+      "loss": 0.1411,
+      "step": 3740
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.295220810647308e-05,
+      "loss": 0.1457,
+      "step": 3750
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 4.291859917994219e-05,
+      "loss": 0.1132,
+      "step": 3760
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.2884990253411305e-05,
+      "loss": 0.1373,
+      "step": 3770
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 4.285138132688042e-05,
+      "loss": 0.1297,
+      "step": 3780
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 4.281777240034954e-05,
+      "loss": 0.1425,
+      "step": 3790
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 4.278416347381865e-05,
+      "loss": 0.1357,
+      "step": 3800
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 4.2750554547287765e-05,
+      "loss": 0.1225,
+      "step": 3810
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.271694562075688e-05,
+      "loss": 0.1388,
+      "step": 3820
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 4.268333669422599e-05,
+      "loss": 0.166,
+      "step": 3830
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.2649727767695105e-05,
+      "loss": 0.1563,
+      "step": 3840
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 4.261611884116422e-05,
+      "loss": 0.1614,
+      "step": 3850
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.9778061224489796,
+      "eval_loss": 0.0575876459479332,
+      "eval_runtime": 88.2297,
+      "eval_samples_per_second": 133.288,
+      "eval_steps_per_second": 4.171,
+      "step": 3857
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.258250991463333e-05,
+      "loss": 0.1573,
+      "step": 3860
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 4.2548900988102444e-05,
+      "loss": 0.1575,
+      "step": 3870
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 4.251529206157156e-05,
+      "loss": 0.1446,
+      "step": 3880
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 4.248168313504067e-05,
+      "loss": 0.1544,
+      "step": 3890
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 4.244807420850978e-05,
+      "loss": 0.1389,
+      "step": 3900
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.241446528197889e-05,
+      "loss": 0.136,
+      "step": 3910
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 4.2380856355448004e-05,
+      "loss": 0.1322,
+      "step": 3920
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.2347247428917124e-05,
+      "loss": 0.1487,
+      "step": 3930
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.231363850238624e-05,
+      "loss": 0.1416,
+      "step": 3940
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 4.228002957585535e-05,
+      "loss": 0.1459,
+      "step": 3950
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 4.2246420649324464e-05,
+      "loss": 0.1284,
+      "step": 3960
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 4.221281172279358e-05,
+      "loss": 0.1278,
+      "step": 3970
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 4.217920279626269e-05,
+      "loss": 0.0966,
+      "step": 3980
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.21455938697318e-05,
+      "loss": 0.1293,
+      "step": 3990
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 4.2111984943200916e-05,
+      "loss": 0.1325,
+      "step": 4000
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.207837601667003e-05,
+      "loss": 0.1379,
+      "step": 4010
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 4.204476709013914e-05,
+      "loss": 0.1452,
+      "step": 4020
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.2011158163608256e-05,
+      "loss": 0.1415,
+      "step": 4030
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.197754923707737e-05,
+      "loss": 0.1389,
+      "step": 4040
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 4.194394031054648e-05,
+      "loss": 0.1375,
+      "step": 4050
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.1910331384015596e-05,
+      "loss": 0.1546,
+      "step": 4060
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.187672245748471e-05,
+      "loss": 0.1283,
+      "step": 4070
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 4.184311353095382e-05,
+      "loss": 0.0967,
+      "step": 4080
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 4.1809504604422936e-05,
+      "loss": 0.1421,
+      "step": 4090
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.177589567789205e-05,
+      "loss": 0.1371,
+      "step": 4100
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.174228675136116e-05,
+      "loss": 0.1198,
+      "step": 4110
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 4.1708677824830275e-05,
+      "loss": 0.1341,
+      "step": 4120
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.167506889829939e-05,
+      "loss": 0.1169,
+      "step": 4130
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.16414599717685e-05,
+      "loss": 0.1469,
+      "step": 4140
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.1607851045237615e-05,
+      "loss": 0.1157,
+      "step": 4150
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.157424211870673e-05,
+      "loss": 0.1422,
+      "step": 4160
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.154063319217584e-05,
+      "loss": 0.1531,
+      "step": 4170
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.1507024265644955e-05,
+      "loss": 0.1344,
+      "step": 4180
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.147341533911407e-05,
+      "loss": 0.1449,
+      "step": 4190
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.143980641258318e-05,
+      "loss": 0.1334,
+      "step": 4200
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.1406197486052295e-05,
+      "loss": 0.1244,
+      "step": 4210
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.1372588559521415e-05,
+      "loss": 0.1321,
+      "step": 4220
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.133897963299053e-05,
+      "loss": 0.1476,
+      "step": 4230
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.130537070645964e-05,
+      "loss": 0.1224,
+      "step": 4240
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.1271761779928754e-05,
+      "loss": 0.1328,
+      "step": 4250
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.123815285339787e-05,
+      "loss": 0.1091,
+      "step": 4260
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.120454392686698e-05,
+      "loss": 0.1312,
+      "step": 4270
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.1170935000336094e-05,
+      "loss": 0.1274,
+      "step": 4280
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.113732607380521e-05,
+      "loss": 0.1284,
+      "step": 4290
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.110371714727432e-05,
+      "loss": 0.125,
+      "step": 4300
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.1070108220743434e-05,
+      "loss": 0.135,
+      "step": 4310
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.103649929421255e-05,
+      "loss": 0.1476,
+      "step": 4320
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.1002890367681654e-05,
+      "loss": 0.1292,
+      "step": 4330
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.096928144115077e-05,
+      "loss": 0.1276,
+      "step": 4340
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.093567251461988e-05,
+      "loss": 0.1182,
+      "step": 4350
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.090206358808899e-05,
+      "loss": 0.144,
+      "step": 4360
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.086845466155811e-05,
+      "loss": 0.1317,
+      "step": 4370
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.0834845735027227e-05,
+      "loss": 0.124,
+      "step": 4380
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.080123680849634e-05,
+      "loss": 0.1294,
+      "step": 4390
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.076762788196545e-05,
+      "loss": 0.1518,
+      "step": 4400
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.9813775510204081,
+      "eval_loss": 0.051723662763834,
+      "eval_runtime": 87.652,
+      "eval_samples_per_second": 134.167,
+      "eval_steps_per_second": 4.198,
+      "step": 4408
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.0734018955434566e-05,
+      "loss": 0.1425,
+      "step": 4410
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 4.070041002890368e-05,
+      "loss": 0.1141,
+      "step": 4420
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.066680110237279e-05,
+      "loss": 0.141,
+      "step": 4430
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.0633192175841906e-05,
+      "loss": 0.1136,
+      "step": 4440
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 4.059958324931102e-05,
+      "loss": 0.1416,
+      "step": 4450
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.056597432278013e-05,
+      "loss": 0.1271,
+      "step": 4460
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 4.0532365396249246e-05,
+      "loss": 0.1273,
+      "step": 4470
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.049875646971836e-05,
+      "loss": 0.1239,
+      "step": 4480
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 4.046514754318747e-05,
+      "loss": 0.1007,
+      "step": 4490
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 4.0431538616656585e-05,
+      "loss": 0.1346,
+      "step": 4500
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 4.03979296901257e-05,
+      "loss": 0.1208,
+      "step": 4510
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 4.036432076359481e-05,
+      "loss": 0.1196,
+      "step": 4520
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.0330711837063925e-05,
+      "loss": 0.1316,
+      "step": 4530
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 4.029710291053304e-05,
+      "loss": 0.1376,
+      "step": 4540
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.026349398400215e-05,
+      "loss": 0.1251,
+      "step": 4550
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 4.0229885057471265e-05,
+      "loss": 0.1242,
+      "step": 4560
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 4.019627613094038e-05,
+      "loss": 0.121,
+      "step": 4570
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 4.016266720440949e-05,
+      "loss": 0.1385,
+      "step": 4580
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.0129058277878605e-05,
+      "loss": 0.1356,
+      "step": 4590
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.009544935134772e-05,
+      "loss": 0.1314,
+      "step": 4600
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.006184042481683e-05,
+      "loss": 0.1247,
+      "step": 4610
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.0028231498285944e-05,
+      "loss": 0.1288,
+      "step": 4620
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 3.999462257175506e-05,
+      "loss": 0.127,
+      "step": 4630
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 3.996101364522417e-05,
+      "loss": 0.1062,
+      "step": 4640
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 3.9927404718693284e-05,
+      "loss": 0.1134,
+      "step": 4650
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 3.9893795792162404e-05,
+      "loss": 0.137,
+      "step": 4660
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 3.986018686563152e-05,
+      "loss": 0.1019,
+      "step": 4670
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 3.982657793910063e-05,
+      "loss": 0.1251,
+      "step": 4680
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 3.9792969012569744e-05,
+      "loss": 0.1209,
+      "step": 4690
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 3.975936008603886e-05,
+      "loss": 0.118,
+      "step": 4700
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 3.972575115950797e-05,
+      "loss": 0.1138,
+      "step": 4710
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 3.9692142232977084e-05,
+      "loss": 0.1455,
+      "step": 4720
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 3.96585333064462e-05,
+      "loss": 0.1098,
+      "step": 4730
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 3.962492437991531e-05,
+      "loss": 0.1317,
+      "step": 4740
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 3.959131545338442e-05,
+      "loss": 0.1177,
+      "step": 4750
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 3.955770652685353e-05,
+      "loss": 0.1285,
+      "step": 4760
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 3.952409760032264e-05,
+      "loss": 0.152,
+      "step": 4770
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 3.9490488673791756e-05,
+      "loss": 0.1207,
+      "step": 4780
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 3.945687974726087e-05,
+      "loss": 0.1252,
+      "step": 4790
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 3.942327082072998e-05,
+      "loss": 0.1261,
+      "step": 4800
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.93896618941991e-05,
+      "loss": 0.13,
+      "step": 4810
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.9356052967668216e-05,
+      "loss": 0.1387,
+      "step": 4820
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.932244404113733e-05,
+      "loss": 0.1472,
+      "step": 4830
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.928883511460644e-05,
+      "loss": 0.1361,
+      "step": 4840
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.9255226188075556e-05,
+      "loss": 0.1073,
+      "step": 4850
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.922161726154467e-05,
+      "loss": 0.1318,
+      "step": 4860
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.918800833501378e-05,
+      "loss": 0.1179,
+      "step": 4870
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.9154399408482895e-05,
+      "loss": 0.1345,
+      "step": 4880
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.912079048195201e-05,
+      "loss": 0.1559,
+      "step": 4890
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.908718155542112e-05,
+      "loss": 0.1328,
+      "step": 4900
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.9053572628890235e-05,
+      "loss": 0.1314,
+      "step": 4910
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.901996370235935e-05,
+      "loss": 0.1252,
+      "step": 4920
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.898635477582846e-05,
+      "loss": 0.0969,
+      "step": 4930
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.8952745849297575e-05,
+      "loss": 0.125,
+      "step": 4940
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.891913692276669e-05,
+      "loss": 0.1231,
+      "step": 4950
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9812074829931973,
+      "eval_loss": 0.05281613767147064,
+      "eval_runtime": 88.537,
+      "eval_samples_per_second": 132.826,
+      "eval_steps_per_second": 4.156,
+      "step": 4959
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.88855279962358e-05,
+      "loss": 0.1307,
+      "step": 4960
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 3.8851919069704915e-05,
+      "loss": 0.1105,
+      "step": 4970
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 3.881831014317403e-05,
+      "loss": 0.1156,
+      "step": 4980
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 3.878470121664314e-05,
+      "loss": 0.1015,
+      "step": 4990
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 3.8751092290112254e-05,
+      "loss": 0.1139,
+      "step": 5000
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 3.871748336358137e-05,
+      "loss": 0.1106,
+      "step": 5010
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 3.868387443705048e-05,
+      "loss": 0.0973,
+      "step": 5020
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 3.8650265510519594e-05,
+      "loss": 0.1044,
+      "step": 5030
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 3.861665658398871e-05,
+      "loss": 0.1266,
+      "step": 5040
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 3.858304765745782e-05,
+      "loss": 0.0969,
+      "step": 5050
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 3.8549438730926934e-05,
+      "loss": 0.131,
+      "step": 5060
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 3.851582980439605e-05,
+      "loss": 0.1117,
+      "step": 5070
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 3.848222087786516e-05,
+      "loss": 0.1342,
+      "step": 5080
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 3.8448611951334274e-05,
+      "loss": 0.105,
+      "step": 5090
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 3.8415003024803394e-05,
+      "loss": 0.1004,
+      "step": 5100
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 3.838139409827251e-05,
+      "loss": 0.1168,
+      "step": 5110
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 3.834778517174162e-05,
+      "loss": 0.1099,
+      "step": 5120
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 3.831417624521073e-05,
+      "loss": 0.1086,
+      "step": 5130
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 3.8280567318679847e-05,
+      "loss": 0.1173,
+      "step": 5140
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 3.824695839214896e-05,
+      "loss": 0.1454,
+      "step": 5150
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 3.821334946561807e-05,
+      "loss": 0.1052,
+      "step": 5160
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 3.8179740539087186e-05,
+      "loss": 0.1406,
+      "step": 5170
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 3.81461316125563e-05,
+      "loss": 0.1191,
+      "step": 5180
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 3.8112522686025406e-05,
+      "loss": 0.136,
+      "step": 5190
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 3.807891375949452e-05,
+      "loss": 0.1123,
+      "step": 5200
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 3.804530483296363e-05,
+      "loss": 0.1285,
+      "step": 5210
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 3.8011695906432746e-05,
+      "loss": 0.1202,
+      "step": 5220
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 3.797808697990186e-05,
+      "loss": 0.1319,
+      "step": 5230
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 3.794447805337098e-05,
+      "loss": 0.1074,
+      "step": 5240
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 3.791086912684009e-05,
+      "loss": 0.1064,
+      "step": 5250
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 3.7877260200309205e-05,
+      "loss": 0.1085,
+      "step": 5260
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 3.784365127377832e-05,
+      "loss": 0.1077,
+      "step": 5270
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 3.781004234724743e-05,
+      "loss": 0.1251,
+      "step": 5280
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 3.7776433420716545e-05,
+      "loss": 0.1143,
+      "step": 5290
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.774282449418566e-05,
+      "loss": 0.1026,
+      "step": 5300
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.770921556765477e-05,
+      "loss": 0.1171,
+      "step": 5310
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.7675606641123885e-05,
+      "loss": 0.1243,
+      "step": 5320
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.7641997714593e-05,
+      "loss": 0.126,
+      "step": 5330
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.760838878806211e-05,
+      "loss": 0.1278,
+      "step": 5340
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 3.7574779861531225e-05,
+      "loss": 0.1031,
+      "step": 5350
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 3.754117093500034e-05,
+      "loss": 0.0903,
+      "step": 5360
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 3.750756200846945e-05,
+      "loss": 0.104,
+      "step": 5370
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 3.7473953081938564e-05,
+      "loss": 0.1049,
+      "step": 5380
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 3.744034415540768e-05,
+      "loss": 0.109,
+      "step": 5390
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 3.740673522887679e-05,
+      "loss": 0.1081,
+      "step": 5400
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 3.7373126302345904e-05,
+      "loss": 0.1312,
+      "step": 5410
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 3.733951737581502e-05,
+      "loss": 0.1262,
+      "step": 5420
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 3.730590844928413e-05,
+      "loss": 0.1223,
+      "step": 5430
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.7272299522753244e-05,
+      "loss": 0.0986,
+      "step": 5440
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.723869059622236e-05,
+      "loss": 0.1119,
+      "step": 5450
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 3.720508166969147e-05,
+      "loss": 0.1225,
+      "step": 5460
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 3.7171472743160584e-05,
+      "loss": 0.1179,
+      "step": 5470
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 3.71378638166297e-05,
+      "loss": 0.1293,
+      "step": 5480
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 3.710425489009881e-05,
+      "loss": 0.1294,
+      "step": 5490
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 3.707064596356792e-05,
+      "loss": 0.1164,
+      "step": 5500
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 0.1076,
+      "step": 5510
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.9850340136054422,
+      "eval_loss": 0.042637791484594345,
+      "eval_runtime": 88.191,
+      "eval_samples_per_second": 133.347,
+      "eval_steps_per_second": 4.173,
+      "step": 5510
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 3.700342811050615e-05,
+      "loss": 0.1241,
+      "step": 5520
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 3.696981918397527e-05,
+      "loss": 0.1161,
+      "step": 5530
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 3.693621025744438e-05,
+      "loss": 0.1173,
+      "step": 5540
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 3.6902601330913496e-05,
+      "loss": 0.1031,
+      "step": 5550
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 3.686899240438261e-05,
+      "loss": 0.1121,
+      "step": 5560
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 3.683538347785172e-05,
+      "loss": 0.0974,
+      "step": 5570
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 3.6801774551320836e-05,
+      "loss": 0.1095,
+      "step": 5580
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 3.676816562478995e-05,
+      "loss": 0.1056,
+      "step": 5590
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 3.673455669825906e-05,
+      "loss": 0.1198,
+      "step": 5600
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 3.6700947771728176e-05,
+      "loss": 0.118,
+      "step": 5610
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 3.666733884519728e-05,
+      "loss": 0.0991,
+      "step": 5620
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 3.6633729918666395e-05,
+      "loss": 0.1053,
+      "step": 5630
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 3.660012099213551e-05,
+      "loss": 0.1473,
+      "step": 5640
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 3.656651206560462e-05,
+      "loss": 0.0962,
+      "step": 5650
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 3.6532903139073735e-05,
+      "loss": 0.1114,
+      "step": 5660
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 3.649929421254285e-05,
+      "loss": 0.0885,
+      "step": 5670
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 3.646568528601197e-05,
+      "loss": 0.097,
+      "step": 5680
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 3.643207635948108e-05,
+      "loss": 0.1266,
+      "step": 5690
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 3.6398467432950195e-05,
+      "loss": 0.1148,
+      "step": 5700
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 3.636485850641931e-05,
+      "loss": 0.1058,
+      "step": 5710
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 3.633124957988842e-05,
+      "loss": 0.1412,
+      "step": 5720
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 3.6297640653357535e-05,
+      "loss": 0.1142,
+      "step": 5730
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 3.626403172682665e-05,
+      "loss": 0.0994,
+      "step": 5740
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 3.623042280029576e-05,
+      "loss": 0.1222,
+      "step": 5750
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 3.6196813873764874e-05,
+      "loss": 0.1147,
+      "step": 5760
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 3.616320494723399e-05,
+      "loss": 0.1083,
+      "step": 5770
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 3.61295960207031e-05,
+      "loss": 0.0861,
+      "step": 5780
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 3.6095987094172214e-05,
+      "loss": 0.1176,
+      "step": 5790
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 3.606237816764133e-05,
+      "loss": 0.1146,
+      "step": 5800
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 3.602876924111044e-05,
+      "loss": 0.1087,
+      "step": 5810
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 3.5995160314579554e-05,
+      "loss": 0.0939,
+      "step": 5820
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 3.596155138804867e-05,
+      "loss": 0.0986,
+      "step": 5830
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 3.592794246151778e-05,
+      "loss": 0.0903,
+      "step": 5840
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 3.5894333534986894e-05,
+      "loss": 0.1134,
+      "step": 5850
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 3.586072460845601e-05,
+      "loss": 0.0934,
+      "step": 5860
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 3.582711568192512e-05,
+      "loss": 0.1176,
+      "step": 5870
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 3.579350675539423e-05,
+      "loss": 0.0956,
+      "step": 5880
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 3.5759897828863347e-05,
+      "loss": 0.1081,
+      "step": 5890
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 3.572628890233246e-05,
+      "loss": 0.1129,
+      "step": 5900
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 3.569267997580157e-05,
+      "loss": 0.1191,
+      "step": 5910
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 3.5659071049270686e-05,
+      "loss": 0.0908,
+      "step": 5920
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 3.56254621227398e-05,
+      "loss": 0.0859,
+      "step": 5930
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 3.559185319620891e-05,
+      "loss": 0.1106,
+      "step": 5940
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 3.5558244269678026e-05,
+      "loss": 0.1064,
+      "step": 5950
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 3.552463534314714e-05,
+      "loss": 0.1048,
+      "step": 5960
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 3.549102641661626e-05,
+      "loss": 0.0892,
+      "step": 5970
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 3.545741749008537e-05,
+      "loss": 0.1068,
+      "step": 5980
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 3.5423808563554486e-05,
+      "loss": 0.1051,
+      "step": 5990
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 3.53901996370236e-05,
+      "loss": 0.0972,
+      "step": 6000
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 3.535659071049271e-05,
+      "loss": 0.0979,
+      "step": 6010
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 3.5322981783961826e-05,
+      "loss": 0.1069,
+      "step": 6020
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 3.528937285743094e-05,
+      "loss": 0.1021,
+      "step": 6030
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 3.525576393090005e-05,
+      "loss": 0.0965,
+      "step": 6040
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 3.522215500436916e-05,
+      "loss": 0.1088,
+      "step": 6050
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 3.518854607783827e-05,
+      "loss": 0.0953,
+      "step": 6060
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9795068027210885,
+      "eval_loss": 0.06343987584114075,
+      "eval_runtime": 86.522,
+      "eval_samples_per_second": 135.919,
+      "eval_steps_per_second": 4.253,
+      "step": 6061
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 3.5154937151307385e-05,
+      "loss": 0.1525,
+      "step": 6070
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 3.51213282247765e-05,
+      "loss": 0.1161,
+      "step": 6080
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 3.508771929824561e-05,
+      "loss": 0.107,
+      "step": 6090
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 3.5054110371714725e-05,
+      "loss": 0.0959,
+      "step": 6100
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 3.502050144518384e-05,
+      "loss": 0.0871,
+      "step": 6110
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 3.498689251865296e-05,
+      "loss": 0.1002,
+      "step": 6120
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 3.495328359212207e-05,
+      "loss": 0.1089,
+      "step": 6130
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 3.4919674665591184e-05,
+      "loss": 0.1019,
+      "step": 6140
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 3.48860657390603e-05,
+      "loss": 0.0936,
+      "step": 6150
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 3.485245681252941e-05,
+      "loss": 0.0953,
+      "step": 6160
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 3.4818847885998524e-05,
+      "loss": 0.0887,
+      "step": 6170
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 3.478523895946764e-05,
+      "loss": 0.0983,
+      "step": 6180
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 3.475163003293675e-05,
+      "loss": 0.1225,
+      "step": 6190
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 3.4718021106405864e-05,
+      "loss": 0.1196,
+      "step": 6200
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 3.468441217987498e-05,
+      "loss": 0.108,
+      "step": 6210
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 3.465080325334409e-05,
+      "loss": 0.1024,
+      "step": 6220
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 3.4617194326813204e-05,
+      "loss": 0.0882,
+      "step": 6230
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 3.458358540028232e-05,
+      "loss": 0.1142,
+      "step": 6240
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 3.454997647375143e-05,
+      "loss": 0.1056,
+      "step": 6250
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 3.451636754722054e-05,
+      "loss": 0.0868,
+      "step": 6260
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 3.4482758620689657e-05,
+      "loss": 0.0968,
+      "step": 6270
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 3.444914969415877e-05,
+      "loss": 0.0966,
+      "step": 6280
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 3.441554076762788e-05,
+      "loss": 0.0917,
+      "step": 6290
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 3.4381931841096996e-05,
+      "loss": 0.0865,
+      "step": 6300
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 3.434832291456611e-05,
+      "loss": 0.1162,
+      "step": 6310
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 3.431471398803522e-05,
+      "loss": 0.0999,
+      "step": 6320
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 3.4281105061504336e-05,
+      "loss": 0.1169,
+      "step": 6330
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 3.424749613497345e-05,
+      "loss": 0.0961,
+      "step": 6340
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 3.421388720844256e-05,
+      "loss": 0.101,
+      "step": 6350
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 3.4180278281911676e-05,
+      "loss": 0.1006,
+      "step": 6360
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 3.414666935538079e-05,
+      "loss": 0.0942,
+      "step": 6370
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 3.41130604288499e-05,
+      "loss": 0.0985,
+      "step": 6380
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 3.4079451502319016e-05,
+      "loss": 0.0986,
+      "step": 6390
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 3.404584257578813e-05,
+      "loss": 0.1069,
+      "step": 6400
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 3.401223364925725e-05,
+      "loss": 0.109,
+      "step": 6410
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 3.397862472272636e-05,
+      "loss": 0.0929,
+      "step": 6420
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 3.3945015796195475e-05,
+      "loss": 0.1104,
+      "step": 6430
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 3.391140686966459e-05,
+      "loss": 0.1126,
+      "step": 6440
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 3.38777979431337e-05,
+      "loss": 0.0899,
+      "step": 6450
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 3.3844189016602815e-05,
+      "loss": 0.0756,
+      "step": 6460
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 3.381058009007193e-05,
+      "loss": 0.0904,
+      "step": 6470
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 3.3776971163541035e-05,
+      "loss": 0.0968,
+      "step": 6480
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 3.374336223701015e-05,
+      "loss": 0.1116,
+      "step": 6490
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 3.370975331047926e-05,
+      "loss": 0.1048,
+      "step": 6500
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 3.3676144383948374e-05,
+      "loss": 0.117,
+      "step": 6510
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 3.364253545741749e-05,
+      "loss": 0.0941,
+      "step": 6520
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 3.36089265308866e-05,
+      "loss": 0.0775,
+      "step": 6530
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 3.3575317604355714e-05,
+      "loss": 0.1104,
+      "step": 6540
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 3.3541708677824834e-05,
+      "loss": 0.0935,
+      "step": 6550
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 3.350809975129395e-05,
+      "loss": 0.1071,
+      "step": 6560
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 3.347449082476306e-05,
+      "loss": 0.092,
+      "step": 6570
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 3.3440881898232174e-05,
+      "loss": 0.1038,
+      "step": 6580
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 3.340727297170129e-05,
+      "loss": 0.1006,
+      "step": 6590
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 3.33736640451704e-05,
+      "loss": 0.0806,
+      "step": 6600
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 3.3340055118639514e-05,
+      "loss": 0.1097,
+      "step": 6610
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.985969387755102,
+      "eval_loss": 0.039833247661590576,
+      "eval_runtime": 86.1101,
+      "eval_samples_per_second": 136.569,
+      "eval_steps_per_second": 4.274,
+      "step": 6612
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 3.330644619210863e-05,
+      "loss": 0.1124,
+      "step": 6620
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 3.327283726557774e-05,
+      "loss": 0.0846,
+      "step": 6630
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 3.323922833904685e-05,
+      "loss": 0.0932,
+      "step": 6640
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 3.3205619412515967e-05,
+      "loss": 0.1111,
+      "step": 6650
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 3.317201048598508e-05,
+      "loss": 0.0798,
+      "step": 6660
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 3.313840155945419e-05,
+      "loss": 0.0977,
+      "step": 6670
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 3.3104792632923306e-05,
+      "loss": 0.1022,
+      "step": 6680
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 3.307118370639242e-05,
+      "loss": 0.0833,
+      "step": 6690
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 3.303757477986153e-05,
+      "loss": 0.1004,
+      "step": 6700
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 3.3003965853330646e-05,
+      "loss": 0.1135,
+      "step": 6710
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 3.297035692679976e-05,
+      "loss": 0.0931,
+      "step": 6720
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 3.293674800026887e-05,
+      "loss": 0.0889,
+      "step": 6730
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 3.2903139073737986e-05,
+      "loss": 0.1122,
+      "step": 6740
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 3.28695301472071e-05,
+      "loss": 0.1103,
+      "step": 6750
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 3.283592122067621e-05,
+      "loss": 0.0818,
+      "step": 6760
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 3.2802312294145326e-05,
+      "loss": 0.1051,
+      "step": 6770
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 3.276870336761444e-05,
+      "loss": 0.1094,
+      "step": 6780
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 3.273509444108355e-05,
+      "loss": 0.0968,
+      "step": 6790
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 3.2701485514552665e-05,
+      "loss": 0.0776,
+      "step": 6800
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 3.266787658802178e-05,
+      "loss": 0.093,
+      "step": 6810
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 3.263426766149089e-05,
+      "loss": 0.0938,
+      "step": 6820
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 3.2600658734960005e-05,
+      "loss": 0.1,
+      "step": 6830
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 3.256704980842912e-05,
+      "loss": 0.0891,
+      "step": 6840
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 3.253344088189824e-05,
+      "loss": 0.0861,
+      "step": 6850
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 3.249983195536735e-05,
+      "loss": 0.1001,
+      "step": 6860
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 3.2466223028836465e-05,
+      "loss": 0.0826,
+      "step": 6870
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 3.243261410230558e-05,
+      "loss": 0.0914,
+      "step": 6880
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 3.239900517577469e-05,
+      "loss": 0.0991,
+      "step": 6890
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 3.2365396249243804e-05,
+      "loss": 0.088,
+      "step": 6900
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 3.233178732271291e-05,
+      "loss": 0.0804,
+      "step": 6910
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 3.2298178396182024e-05,
+      "loss": 0.0855,
+      "step": 6920
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 3.226456946965114e-05,
+      "loss": 0.1007,
+      "step": 6930
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 3.223096054312025e-05,
+      "loss": 0.0868,
+      "step": 6940
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 3.2197351616589364e-05,
+      "loss": 0.1029,
+      "step": 6950
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 3.216374269005848e-05,
+      "loss": 0.1003,
+      "step": 6960
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 3.213013376352759e-05,
+      "loss": 0.0916,
+      "step": 6970
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 3.2096524836996704e-05,
+      "loss": 0.0901,
+      "step": 6980
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 3.2062915910465824e-05,
+      "loss": 0.0776,
+      "step": 6990
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 3.202930698393494e-05,
+      "loss": 0.0831,
+      "step": 7000
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 3.199569805740405e-05,
+      "loss": 0.097,
+      "step": 7010
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 3.196208913087316e-05,
+      "loss": 0.0994,
+      "step": 7020
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 3.192848020434228e-05,
+      "loss": 0.0881,
+      "step": 7030
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 3.189487127781139e-05,
+      "loss": 0.0973,
+      "step": 7040
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 3.18612623512805e-05,
+      "loss": 0.1038,
+      "step": 7050
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 3.1827653424749616e-05,
+      "loss": 0.0963,
+      "step": 7060
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 3.179404449821873e-05,
+      "loss": 0.0753,
+      "step": 7070
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 3.176043557168784e-05,
+      "loss": 0.1137,
+      "step": 7080
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 3.1726826645156956e-05,
+      "loss": 0.1004,
+      "step": 7090
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 3.169321771862607e-05,
+      "loss": 0.0787,
+      "step": 7100
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 3.165960879209518e-05,
+      "loss": 0.0888,
+      "step": 7110
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 3.1625999865564296e-05,
+      "loss": 0.0931,
+      "step": 7120
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 3.159239093903341e-05,
+      "loss": 0.0769,
+      "step": 7130
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 3.155878201250252e-05,
+      "loss": 0.1172,
+      "step": 7140
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 3.1525173085971636e-05,
+      "loss": 0.1079,
+      "step": 7150
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 3.149156415944075e-05,
+      "loss": 0.0763,
+      "step": 7160
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9866496598639456,
+      "eval_loss": 0.034792013466358185,
+      "eval_runtime": 85.3683,
+      "eval_samples_per_second": 137.756,
+      "eval_steps_per_second": 4.311,
+      "step": 7163
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 3.145795523290986e-05,
+      "loss": 0.0968,
+      "step": 7170
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 3.1424346306378975e-05,
+      "loss": 0.0799,
+      "step": 7180
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 3.139073737984809e-05,
+      "loss": 0.0901,
+      "step": 7190
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 3.13571284533172e-05,
+      "loss": 0.0963,
+      "step": 7200
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 3.1323519526786315e-05,
+      "loss": 0.0835,
+      "step": 7210
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 3.128991060025543e-05,
+      "loss": 0.0945,
+      "step": 7220
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 3.125630167372454e-05,
+      "loss": 0.0855,
+      "step": 7230
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 3.1222692747193655e-05,
+      "loss": 0.1057,
+      "step": 7240
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 3.118908382066277e-05,
+      "loss": 0.095,
+      "step": 7250
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 3.115547489413188e-05,
+      "loss": 0.0803,
+      "step": 7260
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 3.1121865967600994e-05,
+      "loss": 0.0974,
+      "step": 7270
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 3.1088257041070114e-05,
+      "loss": 0.1054,
+      "step": 7280
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 3.105464811453923e-05,
+      "loss": 0.0947,
+      "step": 7290
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 3.102103918800834e-05,
+      "loss": 0.0841,
+      "step": 7300
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 3.0987430261477454e-05,
+      "loss": 0.0727,
+      "step": 7310
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 3.095382133494657e-05,
+      "loss": 0.0938,
+      "step": 7320
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 3.092021240841568e-05,
+      "loss": 0.0843,
+      "step": 7330
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 3.088660348188479e-05,
+      "loss": 0.0847,
+      "step": 7340
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 3.08529945553539e-05,
+      "loss": 0.097,
+      "step": 7350
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 3.0819385628823014e-05,
+      "loss": 0.073,
+      "step": 7360
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 3.078577670229213e-05,
+      "loss": 0.0816,
+      "step": 7370
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 3.075216777576124e-05,
+      "loss": 0.1048,
+      "step": 7380
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 3.071855884923035e-05,
+      "loss": 0.0818,
+      "step": 7390
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 3.068494992269947e-05,
+      "loss": 0.0861,
+      "step": 7400
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 3.065134099616858e-05,
+      "loss": 0.098,
+      "step": 7410
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 3.061773206963769e-05,
+      "loss": 0.097,
+      "step": 7420
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 3.058412314310681e-05,
+      "loss": 0.099,
+      "step": 7430
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 3.0550514216575926e-05,
+      "loss": 0.0813,
+      "step": 7440
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 3.051690529004504e-05,
+      "loss": 0.0703,
+      "step": 7450
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 3.0483296363514153e-05,
+      "loss": 0.09,
+      "step": 7460
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 3.0449687436983266e-05,
+      "loss": 0.1,
+      "step": 7470
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 3.041607851045238e-05,
+      "loss": 0.0844,
+      "step": 7480
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 3.0382469583921493e-05,
+      "loss": 0.0985,
+      "step": 7490
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 3.0348860657390606e-05,
+      "loss": 0.0917,
+      "step": 7500
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 3.0315251730859716e-05,
+      "loss": 0.0905,
+      "step": 7510
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 3.028164280432883e-05,
+      "loss": 0.104,
+      "step": 7520
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 3.0248033877797942e-05,
+      "loss": 0.1051,
+      "step": 7530
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 3.0214424951267055e-05,
+      "loss": 0.0711,
+      "step": 7540
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 3.018081602473617e-05,
+      "loss": 0.0658,
+      "step": 7550
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 3.0147207098205282e-05,
+      "loss": 0.086,
+      "step": 7560
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 3.0113598171674402e-05,
+      "loss": 0.0957,
+      "step": 7570
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 3.0079989245143515e-05,
+      "loss": 0.1112,
+      "step": 7580
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 3.004638031861263e-05,
+      "loss": 0.0761,
+      "step": 7590
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 3.0012771392081738e-05,
+      "loss": 0.1093,
+      "step": 7600
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.997916246555085e-05,
+      "loss": 0.099,
+      "step": 7610
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.9945553539019965e-05,
+      "loss": 0.0899,
+      "step": 7620
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.9911944612489078e-05,
+      "loss": 0.101,
+      "step": 7630
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.987833568595819e-05,
+      "loss": 0.0875,
+      "step": 7640
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.9844726759427304e-05,
+      "loss": 0.0684,
+      "step": 7650
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.9811117832896418e-05,
+      "loss": 0.079,
+      "step": 7660
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.977750890636553e-05,
+      "loss": 0.0883,
+      "step": 7670
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.9743899979834644e-05,
+      "loss": 0.079,
+      "step": 7680
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.9710291053303757e-05,
+      "loss": 0.0961,
+      "step": 7690
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.967668212677287e-05,
+      "loss": 0.0901,
+      "step": 7700
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.9643073200241984e-05,
+      "loss": 0.0895,
+      "step": 7710
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.9884353741496599,
+      "eval_loss": 0.03410044312477112,
+      "eval_runtime": 88.1283,
+      "eval_samples_per_second": 133.442,
+      "eval_steps_per_second": 4.176,
+      "step": 7714
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.96094642737111e-05,
+      "loss": 0.0972,
+      "step": 7720
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.9575855347180214e-05,
+      "loss": 0.0918,
+      "step": 7730
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.9542246420649327e-05,
+      "loss": 0.0791,
+      "step": 7740
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.950863749411844e-05,
+      "loss": 0.0766,
+      "step": 7750
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.9475028567587554e-05,
+      "loss": 0.0704,
+      "step": 7760
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.9441419641056667e-05,
+      "loss": 0.0926,
+      "step": 7770
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.940781071452578e-05,
+      "loss": 0.0775,
+      "step": 7780
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.9374201787994893e-05,
+      "loss": 0.0713,
+      "step": 7790
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.9340592861464007e-05,
+      "loss": 0.0739,
+      "step": 7800
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.930698393493312e-05,
+      "loss": 0.0748,
+      "step": 7810
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.9273375008402233e-05,
+      "loss": 0.0659,
+      "step": 7820
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.9239766081871346e-05,
+      "loss": 0.096,
+      "step": 7830
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.920615715534046e-05,
+      "loss": 0.0814,
+      "step": 7840
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.917254822880957e-05,
+      "loss": 0.0614,
+      "step": 7850
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.9138939302278683e-05,
+      "loss": 0.088,
+      "step": 7860
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.9105330375747803e-05,
+      "loss": 0.0876,
+      "step": 7870
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.9071721449216916e-05,
+      "loss": 0.0821,
+      "step": 7880
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.903811252268603e-05,
+      "loss": 0.0911,
+      "step": 7890
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.9004503596155142e-05,
+      "loss": 0.0846,
+      "step": 7900
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.8970894669624256e-05,
+      "loss": 0.0782,
+      "step": 7910
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.893728574309337e-05,
+      "loss": 0.1068,
+      "step": 7920
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.8903676816562482e-05,
+      "loss": 0.0989,
+      "step": 7930
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.8870067890031592e-05,
+      "loss": 0.077,
+      "step": 7940
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.8836458963500705e-05,
+      "loss": 0.0857,
+      "step": 7950
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.880285003696982e-05,
+      "loss": 0.0883,
+      "step": 7960
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.876924111043893e-05,
+      "loss": 0.0841,
+      "step": 7970
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.8735632183908045e-05,
+      "loss": 0.0777,
+      "step": 7980
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.8702023257377158e-05,
+      "loss": 0.0862,
+      "step": 7990
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.866841433084627e-05,
+      "loss": 0.0851,
+      "step": 8000
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.863480540431539e-05,
+      "loss": 0.1016,
+      "step": 8010
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.8601196477784505e-05,
+      "loss": 0.0648,
+      "step": 8020
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.8567587551253615e-05,
+      "loss": 0.073,
+      "step": 8030
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.8533978624722728e-05,
+      "loss": 0.0886,
+      "step": 8040
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.850036969819184e-05,
+      "loss": 0.0982,
+      "step": 8050
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.8466760771660954e-05,
+      "loss": 0.0733,
+      "step": 8060
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.8433151845130067e-05,
+      "loss": 0.0904,
+      "step": 8070
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.839954291859918e-05,
+      "loss": 0.0766,
+      "step": 8080
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.8365933992068294e-05,
+      "loss": 0.0739,
+      "step": 8090
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.8332325065537407e-05,
+      "loss": 0.0793,
+      "step": 8100
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.829871613900652e-05,
+      "loss": 0.0833,
+      "step": 8110
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.8265107212475634e-05,
+      "loss": 0.0671,
+      "step": 8120
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.8231498285944747e-05,
+      "loss": 0.0827,
+      "step": 8130
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.819788935941386e-05,
+      "loss": 0.104,
+      "step": 8140
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.8164280432882973e-05,
+      "loss": 0.0865,
+      "step": 8150
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.813067150635209e-05,
+      "loss": 0.0742,
+      "step": 8160
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.8097062579821203e-05,
+      "loss": 0.0939,
+      "step": 8170
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.8063453653290317e-05,
+      "loss": 0.0829,
+      "step": 8180
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.802984472675943e-05,
+      "loss": 0.0772,
+      "step": 8190
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.7996235800228543e-05,
+      "loss": 0.0803,
+      "step": 8200
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.7962626873697656e-05,
+      "loss": 0.0712,
+      "step": 8210
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.792901794716677e-05,
+      "loss": 0.0878,
+      "step": 8220
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.7895409020635883e-05,
+      "loss": 0.075,
+      "step": 8230
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.7861800094104996e-05,
+      "loss": 0.0682,
+      "step": 8240
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.782819116757411e-05,
+      "loss": 0.0792,
+      "step": 8250
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.7794582241043222e-05,
+      "loss": 0.06,
+      "step": 8260
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.988265306122449,
+      "eval_loss": 0.038060512393713,
+      "eval_runtime": 86.3796,
+      "eval_samples_per_second": 136.143,
+      "eval_steps_per_second": 4.26,
+      "step": 8265
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.7760973314512336e-05,
+      "loss": 0.0624,
+      "step": 8270
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.7727364387981446e-05,
+      "loss": 0.0864,
+      "step": 8280
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.769375546145056e-05,
+      "loss": 0.0712,
+      "step": 8290
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.766014653491968e-05,
+      "loss": 0.0754,
+      "step": 8300
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.7626537608388792e-05,
+      "loss": 0.0823,
+      "step": 8310
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2.7592928681857905e-05,
+      "loss": 0.0785,
+      "step": 8320
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 2.755931975532702e-05,
+      "loss": 0.0727,
+      "step": 8330
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 2.7525710828796132e-05,
+      "loss": 0.0837,
+      "step": 8340
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2.7492101902265245e-05,
+      "loss": 0.0697,
+      "step": 8350
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 2.745849297573436e-05,
+      "loss": 0.0755,
+      "step": 8360
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 2.7424884049203468e-05,
+      "loss": 0.0685,
+      "step": 8370
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2.739127512267258e-05,
+      "loss": 0.0878,
+      "step": 8380
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 2.7357666196141695e-05,
+      "loss": 0.0872,
+      "step": 8390
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 2.7324057269610808e-05,
+      "loss": 0.0913,
+      "step": 8400
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.729044834307992e-05,
+      "loss": 0.1044,
+      "step": 8410
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 2.7256839416549034e-05,
+      "loss": 0.086,
+      "step": 8420
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 2.7223230490018148e-05,
+      "loss": 0.082,
+      "step": 8430
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 2.718962156348726e-05,
+      "loss": 0.0765,
+      "step": 8440
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 2.715601263695638e-05,
+      "loss": 0.0909,
+      "step": 8450
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 2.7122403710425494e-05,
+      "loss": 0.0809,
+      "step": 8460
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 2.7088794783894604e-05,
+      "loss": 0.0871,
+      "step": 8470
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 2.7055185857363717e-05,
+      "loss": 0.0731,
+      "step": 8480
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2.702157693083283e-05,
+      "loss": 0.0829,
+      "step": 8490
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 2.6987968004301944e-05,
+      "loss": 0.0766,
+      "step": 8500
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 2.6954359077771057e-05,
+      "loss": 0.0706,
+      "step": 8510
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2.692075015124017e-05,
+      "loss": 0.0839,
+      "step": 8520
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 2.6887141224709283e-05,
+      "loss": 0.0949,
+      "step": 8530
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 2.6853532298178397e-05,
+      "loss": 0.0709,
+      "step": 8540
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 2.681992337164751e-05,
+      "loss": 0.0803,
+      "step": 8550
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 2.6786314445116623e-05,
+      "loss": 0.0794,
+      "step": 8560
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 2.6752705518585736e-05,
+      "loss": 0.0794,
+      "step": 8570
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 2.671909659205485e-05,
+      "loss": 0.0976,
+      "step": 8580
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 2.6685487665523966e-05,
+      "loss": 0.076,
+      "step": 8590
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 2.665187873899308e-05,
+      "loss": 0.0807,
+      "step": 8600
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 2.6618269812462193e-05,
+      "loss": 0.065,
+      "step": 8610
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 2.6584660885931306e-05,
+      "loss": 0.0742,
+      "step": 8620
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 2.655105195940042e-05,
+      "loss": 0.0813,
+      "step": 8630
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 2.6517443032869533e-05,
+      "loss": 0.082,
+      "step": 8640
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 2.6483834106338646e-05,
+      "loss": 0.0603,
+      "step": 8650
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2.645022517980776e-05,
+      "loss": 0.0797,
+      "step": 8660
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 2.6416616253276872e-05,
+      "loss": 0.0813,
+      "step": 8670
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 2.6383007326745985e-05,
+      "loss": 0.0753,
+      "step": 8680
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2.63493984002151e-05,
+      "loss": 0.0837,
+      "step": 8690
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 2.6315789473684212e-05,
+      "loss": 0.0803,
+      "step": 8700
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 2.6282180547153322e-05,
+      "loss": 0.0798,
+      "step": 8710
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 2.6248571620622435e-05,
+      "loss": 0.0784,
+      "step": 8720
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 2.621496269409155e-05,
+      "loss": 0.0709,
+      "step": 8730
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 2.618135376756067e-05,
+      "loss": 0.0597,
+      "step": 8740
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 2.614774484102978e-05,
+      "loss": 0.076,
+      "step": 8750
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 2.6114135914498895e-05,
+      "loss": 0.0918,
+      "step": 8760
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2.6080526987968008e-05,
+      "loss": 0.0762,
+      "step": 8770
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 2.604691806143712e-05,
+      "loss": 0.0874,
+      "step": 8780
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 2.6013309134906235e-05,
+      "loss": 0.0906,
+      "step": 8790
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 2.5979700208375344e-05,
+      "loss": 0.0797,
+      "step": 8800
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 2.5946091281844458e-05,
+      "loss": 0.0767,
+      "step": 8810
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.9875,
+      "eval_loss": 0.038219138979911804,
+      "eval_runtime": 89.4197,
+      "eval_samples_per_second": 131.515,
+      "eval_steps_per_second": 4.115,
+      "step": 8816
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 2.591248235531357e-05,
+      "loss": 0.0849,
+      "step": 8820
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2.5878873428782684e-05,
+      "loss": 0.0787,
+      "step": 8830
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 2.5845264502251797e-05,
+      "loss": 0.0676,
+      "step": 8840
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 2.581165557572091e-05,
+      "loss": 0.0793,
+      "step": 8850
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2.5778046649190024e-05,
+      "loss": 0.0829,
+      "step": 8860
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 2.5744437722659137e-05,
+      "loss": 0.0768,
+      "step": 8870
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 2.571082879612825e-05,
+      "loss": 0.0725,
+      "step": 8880
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2.567721986959737e-05,
+      "loss": 0.0915,
+      "step": 8890
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 2.564361094306648e-05,
+      "loss": 0.0825,
+      "step": 8900
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 2.5610002016535593e-05,
+      "loss": 0.0989,
+      "step": 8910
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 2.5576393090004707e-05,
+      "loss": 0.087,
+      "step": 8920
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 2.554278416347382e-05,
+      "loss": 0.0834,
+      "step": 8930
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 2.5509175236942933e-05,
+      "loss": 0.0783,
+      "step": 8940
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2.5475566310412046e-05,
+      "loss": 0.082,
+      "step": 8950
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 2.544195738388116e-05,
+      "loss": 0.0713,
+      "step": 8960
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 2.5408348457350273e-05,
+      "loss": 0.076,
+      "step": 8970
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 2.5374739530819386e-05,
+      "loss": 0.0873,
+      "step": 8980
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 2.53411306042885e-05,
+      "loss": 0.0718,
+      "step": 8990
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 2.5307521677757613e-05,
+      "loss": 0.0876,
+      "step": 9000
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 2.5273912751226726e-05,
+      "loss": 0.0772,
+      "step": 9010
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 2.524030382469584e-05,
+      "loss": 0.0717,
+      "step": 9020
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2.5206694898164956e-05,
+      "loss": 0.0849,
+      "step": 9030
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 2.517308597163407e-05,
+      "loss": 0.0668,
+      "step": 9040
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 2.5139477045103182e-05,
+      "loss": 0.0794,
+      "step": 9050
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2.5105868118572295e-05,
+      "loss": 0.0676,
+      "step": 9060
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 2.507225919204141e-05,
+      "loss": 0.086,
+      "step": 9070
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 2.5038650265510522e-05,
+      "loss": 0.0791,
+      "step": 9080
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 2.5005041338979635e-05,
+      "loss": 0.0718,
+      "step": 9090
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 2.497143241244875e-05,
+      "loss": 0.0885,
+      "step": 9100
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 2.4937823485917862e-05,
+      "loss": 0.065,
+      "step": 9110
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 2.4904214559386975e-05,
+      "loss": 0.09,
+      "step": 9120
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 2.4870605632856088e-05,
+      "loss": 0.0879,
+      "step": 9130
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 2.48369967063252e-05,
+      "loss": 0.0962,
+      "step": 9140
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 2.4803387779794315e-05,
+      "loss": 0.0818,
+      "step": 9150
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 2.4769778853263428e-05,
+      "loss": 0.0813,
+      "step": 9160
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 2.473616992673254e-05,
+      "loss": 0.074,
+      "step": 9170
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 2.4702561000201654e-05,
+      "loss": 0.0719,
+      "step": 9180
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 2.4668952073670768e-05,
+      "loss": 0.0709,
+      "step": 9190
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2.463534314713988e-05,
+      "loss": 0.0755,
+      "step": 9200
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.4601734220608998e-05,
+      "loss": 0.0744,
+      "step": 9210
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 2.456812529407811e-05,
+      "loss": 0.086,
+      "step": 9220
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 2.453451636754722e-05,
+      "loss": 0.0799,
+      "step": 9230
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 2.4500907441016334e-05,
+      "loss": 0.06,
+      "step": 9240
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 2.4467298514485447e-05,
+      "loss": 0.0776,
+      "step": 9250
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 2.443368958795456e-05,
+      "loss": 0.1042,
+      "step": 9260
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 2.4400080661423674e-05,
+      "loss": 0.0727,
+      "step": 9270
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 2.4366471734892787e-05,
+      "loss": 0.0663,
+      "step": 9280
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 2.4332862808361903e-05,
+      "loss": 0.0727,
+      "step": 9290
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 2.4299253881831017e-05,
+      "loss": 0.0735,
+      "step": 9300
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 2.426564495530013e-05,
+      "loss": 0.0776,
+      "step": 9310
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 2.4232036028769243e-05,
+      "loss": 0.0856,
+      "step": 9320
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 2.4198427102238356e-05,
+      "loss": 0.0784,
+      "step": 9330
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 2.416481817570747e-05,
+      "loss": 0.0803,
+      "step": 9340
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 2.413120924917658e-05,
+      "loss": 0.0816,
+      "step": 9350
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 2.4097600322645696e-05,
+      "loss": 0.0868,
+      "step": 9360
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.9897959183673469,
+      "eval_loss": 0.030931392684578896,
+      "eval_runtime": 93.3354,
+      "eval_samples_per_second": 125.997,
+      "eval_steps_per_second": 3.943,
+      "step": 9367
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 2.406399139611481e-05,
+      "loss": 0.0912,
+      "step": 9370
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 2.4030382469583923e-05,
+      "loss": 0.0875,
+      "step": 9380
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 2.3996773543053036e-05,
+      "loss": 0.0967,
+      "step": 9390
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 2.396316461652215e-05,
+      "loss": 0.0694,
+      "step": 9400
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 2.3929555689991262e-05,
+      "loss": 0.079,
+      "step": 9410
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 2.3895946763460376e-05,
+      "loss": 0.0861,
+      "step": 9420
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 2.3862337836929492e-05,
+      "loss": 0.0589,
+      "step": 9430
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 2.3828728910398606e-05,
+      "loss": 0.0766,
+      "step": 9440
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 2.3795119983867715e-05,
+      "loss": 0.0865,
+      "step": 9450
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 2.376151105733683e-05,
+      "loss": 0.0678,
+      "step": 9460
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 2.3727902130805942e-05,
+      "loss": 0.0917,
+      "step": 9470
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 2.3694293204275055e-05,
+      "loss": 0.0663,
+      "step": 9480
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 2.366068427774417e-05,
+      "loss": 0.0658,
+      "step": 9490
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 2.3627075351213285e-05,
+      "loss": 0.0682,
+      "step": 9500
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 2.3593466424682398e-05,
+      "loss": 0.0742,
+      "step": 9510
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 2.355985749815151e-05,
+      "loss": 0.0777,
+      "step": 9520
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 2.3526248571620625e-05,
+      "loss": 0.0692,
+      "step": 9530
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 2.3492639645089738e-05,
+      "loss": 0.0635,
+      "step": 9540
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 2.345903071855885e-05,
+      "loss": 0.072,
+      "step": 9550
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 2.3425421792027964e-05,
+      "loss": 0.0757,
+      "step": 9560
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 2.3391812865497074e-05,
+      "loss": 0.0674,
+      "step": 9570
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 2.335820393896619e-05,
+      "loss": 0.0863,
+      "step": 9580
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 2.3324595012435304e-05,
+      "loss": 0.0557,
+      "step": 9590
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 2.3290986085904417e-05,
+      "loss": 0.0717,
+      "step": 9600
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 2.325737715937353e-05,
+      "loss": 0.0839,
+      "step": 9610
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 2.3223768232842644e-05,
+      "loss": 0.1014,
+      "step": 9620
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 2.3190159306311757e-05,
+      "loss": 0.0713,
+      "step": 9630
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 2.315655037978087e-05,
+      "loss": 0.0706,
+      "step": 9640
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 2.3122941453249987e-05,
+      "loss": 0.0863,
+      "step": 9650
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 2.3089332526719097e-05,
+      "loss": 0.0724,
+      "step": 9660
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 2.305572360018821e-05,
+      "loss": 0.0779,
+      "step": 9670
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 2.3022114673657323e-05,
+      "loss": 0.074,
+      "step": 9680
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 2.2988505747126437e-05,
+      "loss": 0.0873,
+      "step": 9690
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 2.295489682059555e-05,
+      "loss": 0.0611,
+      "step": 9700
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 2.2921287894064663e-05,
+      "loss": 0.0722,
+      "step": 9710
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 2.288767896753378e-05,
+      "loss": 0.0635,
+      "step": 9720
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.2854070041002893e-05,
+      "loss": 0.0717,
+      "step": 9730
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2.2820461114472006e-05,
+      "loss": 0.0706,
+      "step": 9740
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 2.278685218794112e-05,
+      "loss": 0.0835,
+      "step": 9750
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 2.2753243261410233e-05,
+      "loss": 0.0607,
+      "step": 9760
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2.2719634334879346e-05,
+      "loss": 0.0719,
+      "step": 9770
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 2.2686025408348456e-05,
+      "loss": 0.08,
+      "step": 9780
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 2.265241648181757e-05,
+      "loss": 0.0632,
+      "step": 9790
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 2.2618807555286686e-05,
+      "loss": 0.0801,
+      "step": 9800
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 2.25851986287558e-05,
+      "loss": 0.0633,
+      "step": 9810
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.2551589702224912e-05,
+      "loss": 0.0684,
+      "step": 9820
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.2517980775694025e-05,
+      "loss": 0.0931,
+      "step": 9830
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 2.248437184916314e-05,
+      "loss": 0.0636,
+      "step": 9840
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 2.2450762922632252e-05,
+      "loss": 0.0578,
+      "step": 9850
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 2.2417153996101365e-05,
+      "loss": 0.0681,
+      "step": 9860
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 2.2383545069570482e-05,
+      "loss": 0.0815,
+      "step": 9870
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2.234993614303959e-05,
+      "loss": 0.0702,
+      "step": 9880
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 2.2316327216508705e-05,
+      "loss": 0.0705,
+      "step": 9890
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 2.2282718289977818e-05,
+      "loss": 0.077,
+      "step": 9900
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 2.224910936344693e-05,
+      "loss": 0.091,
+      "step": 9910
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.9885204081632653,
+      "eval_loss": 0.033893804997205734,
+      "eval_runtime": 94.7988,
+      "eval_samples_per_second": 124.052,
+      "eval_steps_per_second": 3.882,
+      "step": 9918
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 2.2215500436916045e-05,
+      "loss": 0.0747,
+      "step": 9920
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 2.2181891510385158e-05,
+      "loss": 0.0653,
+      "step": 9930
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 2.2148282583854274e-05,
+      "loss": 0.0706,
+      "step": 9940
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 2.2114673657323388e-05,
+      "loss": 0.0881,
+      "step": 9950
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 2.20810647307925e-05,
+      "loss": 0.0606,
+      "step": 9960
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2.2047455804261614e-05,
+      "loss": 0.0647,
+      "step": 9970
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.2013846877730727e-05,
+      "loss": 0.0759,
+      "step": 9980
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 2.198023795119984e-05,
+      "loss": 0.083,
+      "step": 9990
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.194662902466895e-05,
+      "loss": 0.0767,
+      "step": 10000
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 2.1913020098138067e-05,
+      "loss": 0.0786,
+      "step": 10010
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.187941117160718e-05,
+      "loss": 0.0529,
+      "step": 10020
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.1845802245076294e-05,
+      "loss": 0.0665,
+      "step": 10030
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 2.1812193318545407e-05,
+      "loss": 0.0803,
+      "step": 10040
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.177858439201452e-05,
+      "loss": 0.0678,
+      "step": 10050
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 2.1744975465483633e-05,
+      "loss": 0.0722,
+      "step": 10060
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.1711366538952747e-05,
+      "loss": 0.0861,
+      "step": 10070
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.167775761242186e-05,
+      "loss": 0.0818,
+      "step": 10080
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 2.1644148685890973e-05,
+      "loss": 0.0609,
+      "step": 10090
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.1610539759360086e-05,
+      "loss": 0.0746,
+      "step": 10100
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.15769308328292e-05,
+      "loss": 0.0693,
+      "step": 10110
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.1543321906298313e-05,
+      "loss": 0.0697,
+      "step": 10120
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.1509712979767426e-05,
+      "loss": 0.0666,
+      "step": 10130
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2.147610405323654e-05,
+      "loss": 0.0764,
+      "step": 10140
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.1442495126705653e-05,
+      "loss": 0.0668,
+      "step": 10150
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.140888620017477e-05,
+      "loss": 0.064,
+      "step": 10160
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.1375277273643882e-05,
+      "loss": 0.0864,
+      "step": 10170
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.1341668347112996e-05,
+      "loss": 0.0639,
+      "step": 10180
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.130805942058211e-05,
+      "loss": 0.0714,
+      "step": 10190
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.1274450494051222e-05,
+      "loss": 0.0626,
+      "step": 10200
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.1240841567520335e-05,
+      "loss": 0.0695,
+      "step": 10210
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.1207232640989445e-05,
+      "loss": 0.0717,
+      "step": 10220
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.1173623714458562e-05,
+      "loss": 0.0627,
+      "step": 10230
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.1140014787927675e-05,
+      "loss": 0.0688,
+      "step": 10240
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.110640586139679e-05,
+      "loss": 0.0819,
+      "step": 10250
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.10727969348659e-05,
+      "loss": 0.0769,
+      "step": 10260
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 2.1039188008335015e-05,
+      "loss": 0.0736,
+      "step": 10270
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.1005579081804128e-05,
+      "loss": 0.0615,
+      "step": 10280
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.097197015527324e-05,
+      "loss": 0.0745,
+      "step": 10290
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.0938361228742355e-05,
+      "loss": 0.0651,
+      "step": 10300
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2.0904752302211468e-05,
+      "loss": 0.084,
+      "step": 10310
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 2.087114337568058e-05,
+      "loss": 0.0601,
+      "step": 10320
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 2.0837534449149694e-05,
+      "loss": 0.0639,
+      "step": 10330
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 2.0803925522618808e-05,
+      "loss": 0.071,
+      "step": 10340
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 2.077031659608792e-05,
+      "loss": 0.0666,
+      "step": 10350
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.0736707669557034e-05,
+      "loss": 0.0597,
+      "step": 10360
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.0703098743026147e-05,
+      "loss": 0.0785,
+      "step": 10370
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.0669489816495264e-05,
+      "loss": 0.0633,
+      "step": 10380
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.0635880889964377e-05,
+      "loss": 0.0696,
+      "step": 10390
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.060227196343349e-05,
+      "loss": 0.0779,
+      "step": 10400
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.0568663036902604e-05,
+      "loss": 0.0736,
+      "step": 10410
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.0535054110371717e-05,
+      "loss": 0.0493,
+      "step": 10420
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.0501445183840827e-05,
+      "loss": 0.068,
+      "step": 10430
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.046783625730994e-05,
+      "loss": 0.0605,
+      "step": 10440
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.0434227330779057e-05,
+      "loss": 0.0839,
+      "step": 10450
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.040061840424817e-05,
+      "loss": 0.0817,
+      "step": 10460
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.9913265306122448,
+      "eval_loss": 0.024345681071281433,
+      "eval_runtime": 94.3004,
+      "eval_samples_per_second": 124.708,
+      "eval_steps_per_second": 3.902,
+      "step": 10469
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.0367009477717283e-05,
+      "loss": 0.0596,
+      "step": 10470
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.0333400551186396e-05,
+      "loss": 0.0677,
+      "step": 10480
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.029979162465551e-05,
+      "loss": 0.0586,
+      "step": 10490
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.0266182698124623e-05,
+      "loss": 0.0604,
+      "step": 10500
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.0232573771593736e-05,
+      "loss": 0.0762,
+      "step": 10510
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.019896484506285e-05,
+      "loss": 0.0718,
+      "step": 10520
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.0165355918531963e-05,
+      "loss": 0.0782,
+      "step": 10530
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.0131746992001076e-05,
+      "loss": 0.0796,
+      "step": 10540
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.009813806547019e-05,
+      "loss": 0.0571,
+      "step": 10550
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0064529138939302e-05,
+      "loss": 0.0824,
+      "step": 10560
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0030920212408416e-05,
+      "loss": 0.0629,
+      "step": 10570
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.999731128587753e-05,
+      "loss": 0.0743,
+      "step": 10580
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9963702359346642e-05,
+      "loss": 0.0584,
+      "step": 10590
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.993009343281576e-05,
+      "loss": 0.0795,
+      "step": 10600
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.9896484506284872e-05,
+      "loss": 0.0591,
+      "step": 10610
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.9862875579753985e-05,
+      "loss": 0.054,
+      "step": 10620
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.98292666532231e-05,
+      "loss": 0.0861,
+      "step": 10630
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.979565772669221e-05,
+      "loss": 0.0676,
+      "step": 10640
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.976204880016132e-05,
+      "loss": 0.0638,
+      "step": 10650
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.9728439873630435e-05,
+      "loss": 0.0749,
+      "step": 10660
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.969483094709955e-05,
+      "loss": 0.0758,
+      "step": 10670
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.9661222020568665e-05,
+      "loss": 0.0876,
+      "step": 10680
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.9627613094037778e-05,
+      "loss": 0.0623,
+      "step": 10690
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.959400416750689e-05,
+      "loss": 0.0703,
+      "step": 10700
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.9560395240976004e-05,
+      "loss": 0.0613,
+      "step": 10710
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.9526786314445118e-05,
+      "loss": 0.0503,
+      "step": 10720
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.949317738791423e-05,
+      "loss": 0.0578,
+      "step": 10730
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.9459568461383344e-05,
+      "loss": 0.0764,
+      "step": 10740
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.9425959534852457e-05,
+      "loss": 0.0758,
+      "step": 10750
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.939235060832157e-05,
+      "loss": 0.0513,
+      "step": 10760
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.9358741681790684e-05,
+      "loss": 0.0644,
+      "step": 10770
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.9325132755259797e-05,
+      "loss": 0.0676,
+      "step": 10780
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.929152382872891e-05,
+      "loss": 0.0714,
+      "step": 10790
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.9257914902198024e-05,
+      "loss": 0.0673,
+      "step": 10800
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.9224305975667137e-05,
+      "loss": 0.0716,
+      "step": 10810
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.9190697049136253e-05,
+      "loss": 0.0619,
+      "step": 10820
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.9157088122605367e-05,
+      "loss": 0.0715,
+      "step": 10830
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.912347919607448e-05,
+      "loss": 0.0817,
+      "step": 10840
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.9089870269543593e-05,
+      "loss": 0.044,
+      "step": 10850
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.9056261343012703e-05,
+      "loss": 0.055,
+      "step": 10860
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.9022652416481816e-05,
+      "loss": 0.047,
+      "step": 10870
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.898904348995093e-05,
+      "loss": 0.0662,
+      "step": 10880
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.8955434563420046e-05,
+      "loss": 0.0526,
+      "step": 10890
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.892182563688916e-05,
+      "loss": 0.0808,
+      "step": 10900
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8888216710358273e-05,
+      "loss": 0.07,
+      "step": 10910
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.8854607783827386e-05,
+      "loss": 0.0509,
+      "step": 10920
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.88209988572965e-05,
+      "loss": 0.0651,
+      "step": 10930
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.8787389930765612e-05,
+      "loss": 0.0626,
+      "step": 10940
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.8753781004234726e-05,
+      "loss": 0.0642,
+      "step": 10950
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.872017207770384e-05,
+      "loss": 0.0629,
+      "step": 10960
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.8686563151172952e-05,
+      "loss": 0.068,
+      "step": 10970
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8652954224642065e-05,
+      "loss": 0.0559,
+      "step": 10980
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.861934529811118e-05,
+      "loss": 0.0757,
+      "step": 10990
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.8585736371580292e-05,
+      "loss": 0.0474,
+      "step": 11000
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.8552127445049405e-05,
+      "loss": 0.0637,
+      "step": 11010
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.8518518518518518e-05,
+      "loss": 0.0641,
+      "step": 11020
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.9906462585034014,
+      "eval_loss": 0.028639109805226326,
+      "eval_runtime": 93.0053,
+      "eval_samples_per_second": 126.444,
+      "eval_steps_per_second": 3.957,
+      "step": 11020
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.8484909591987635e-05,
+      "loss": 0.1102,
+      "step": 11030
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.8451300665456748e-05,
+      "loss": 0.0515,
+      "step": 11040
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.841769173892586e-05,
+      "loss": 0.0574,
+      "step": 11050
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.8384082812394975e-05,
+      "loss": 0.0778,
+      "step": 11060
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.8350473885864088e-05,
+      "loss": 0.0593,
+      "step": 11070
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.8316864959333198e-05,
+      "loss": 0.0805,
+      "step": 11080
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.828325603280231e-05,
+      "loss": 0.0552,
+      "step": 11090
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.8249647106271424e-05,
+      "loss": 0.0628,
+      "step": 11100
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.821603817974054e-05,
+      "loss": 0.0729,
+      "step": 11110
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.8182429253209654e-05,
+      "loss": 0.0604,
+      "step": 11120
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.8148820326678767e-05,
+      "loss": 0.0785,
+      "step": 11130
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.811521140014788e-05,
+      "loss": 0.0626,
+      "step": 11140
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.8081602473616994e-05,
+      "loss": 0.0844,
+      "step": 11150
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.8047993547086107e-05,
+      "loss": 0.0616,
+      "step": 11160
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.801438462055522e-05,
+      "loss": 0.0695,
+      "step": 11170
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.7980775694024334e-05,
+      "loss": 0.057,
+      "step": 11180
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.7947166767493447e-05,
+      "loss": 0.0715,
+      "step": 11190
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.791355784096256e-05,
+      "loss": 0.063,
+      "step": 11200
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.7879948914431673e-05,
+      "loss": 0.082,
+      "step": 11210
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.7846339987900787e-05,
+      "loss": 0.0751,
+      "step": 11220
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.78127310613699e-05,
+      "loss": 0.0669,
+      "step": 11230
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.7779122134839013e-05,
+      "loss": 0.08,
+      "step": 11240
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.774551320830813e-05,
+      "loss": 0.0879,
+      "step": 11250
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.7711904281777243e-05,
+      "loss": 0.0531,
+      "step": 11260
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.7678295355246356e-05,
+      "loss": 0.0548,
+      "step": 11270
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.764468642871547e-05,
+      "loss": 0.0625,
+      "step": 11280
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.761107750218458e-05,
+      "loss": 0.0682,
+      "step": 11290
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.7577468575653692e-05,
+      "loss": 0.0648,
+      "step": 11300
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.7543859649122806e-05,
+      "loss": 0.0621,
+      "step": 11310
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.751025072259192e-05,
+      "loss": 0.0669,
+      "step": 11320
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.7476641796061036e-05,
+      "loss": 0.0696,
+      "step": 11330
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.744303286953015e-05,
+      "loss": 0.0558,
+      "step": 11340
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.7409423942999262e-05,
+      "loss": 0.0674,
+      "step": 11350
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.7375815016468375e-05,
+      "loss": 0.0601,
+      "step": 11360
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.734220608993749e-05,
+      "loss": 0.061,
+      "step": 11370
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.7308597163406602e-05,
+      "loss": 0.07,
+      "step": 11380
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.7274988236875715e-05,
+      "loss": 0.0709,
+      "step": 11390
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.7241379310344828e-05,
+      "loss": 0.0685,
+      "step": 11400
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.720777038381394e-05,
+      "loss": 0.0609,
+      "step": 11410
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.7174161457283055e-05,
+      "loss": 0.053,
+      "step": 11420
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.7140552530752168e-05,
+      "loss": 0.0776,
+      "step": 11430
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.710694360422128e-05,
+      "loss": 0.0643,
+      "step": 11440
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.7073334677690395e-05,
+      "loss": 0.0674,
+      "step": 11450
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.7039725751159508e-05,
+      "loss": 0.0736,
+      "step": 11460
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.7006116824628624e-05,
+      "loss": 0.051,
+      "step": 11470
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.6972507898097738e-05,
+      "loss": 0.0589,
+      "step": 11480
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.693889897156685e-05,
+      "loss": 0.0505,
+      "step": 11490
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.6905290045035964e-05,
+      "loss": 0.0825,
+      "step": 11500
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.6871681118505074e-05,
+      "loss": 0.0765,
+      "step": 11510
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.6838072191974187e-05,
+      "loss": 0.0667,
+      "step": 11520
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.68044632654433e-05,
+      "loss": 0.0554,
+      "step": 11530
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.6770854338912417e-05,
+      "loss": 0.0698,
+      "step": 11540
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.673724541238153e-05,
+      "loss": 0.0499,
+      "step": 11550
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.6703636485850644e-05,
+      "loss": 0.078,
+      "step": 11560
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.6670027559319757e-05,
+      "loss": 0.0703,
+      "step": 11570
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.9905612244897959,
+      "eval_loss": 0.031435031443834305,
+      "eval_runtime": 89.4132,
+      "eval_samples_per_second": 131.524,
+      "eval_steps_per_second": 4.116,
+      "step": 11571
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.663641863278887e-05,
+      "loss": 0.0546,
+      "step": 11580
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.6602809706257983e-05,
+      "loss": 0.0598,
+      "step": 11590
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.6569200779727097e-05,
+      "loss": 0.0656,
+      "step": 11600
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.653559185319621e-05,
+      "loss": 0.0749,
+      "step": 11610
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.6501982926665323e-05,
+      "loss": 0.0553,
+      "step": 11620
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.6468374000134436e-05,
+      "loss": 0.0651,
+      "step": 11630
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.643476507360355e-05,
+      "loss": 0.0651,
+      "step": 11640
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.6401156147072663e-05,
+      "loss": 0.051,
+      "step": 11650
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.6367547220541776e-05,
+      "loss": 0.0522,
+      "step": 11660
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.633393829401089e-05,
+      "loss": 0.0564,
+      "step": 11670
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.6300329367480002e-05,
+      "loss": 0.0452,
+      "step": 11680
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.626672044094912e-05,
+      "loss": 0.0592,
+      "step": 11690
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.6233111514418232e-05,
+      "loss": 0.0559,
+      "step": 11700
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.6199502587887346e-05,
+      "loss": 0.0665,
+      "step": 11710
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.6165893661356455e-05,
+      "loss": 0.0611,
+      "step": 11720
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.613228473482557e-05,
+      "loss": 0.0698,
+      "step": 11730
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.6098675808294682e-05,
+      "loss": 0.0698,
+      "step": 11740
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.6065066881763795e-05,
+      "loss": 0.0512,
+      "step": 11750
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.6031457955232912e-05,
+      "loss": 0.0646,
+      "step": 11760
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.5997849028702025e-05,
+      "loss": 0.0707,
+      "step": 11770
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.596424010217114e-05,
+      "loss": 0.062,
+      "step": 11780
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.593063117564025e-05,
+      "loss": 0.0624,
+      "step": 11790
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.5897022249109365e-05,
+      "loss": 0.0541,
+      "step": 11800
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.5863413322578478e-05,
+      "loss": 0.0603,
+      "step": 11810
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.582980439604759e-05,
+      "loss": 0.041,
+      "step": 11820
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.5796195469516705e-05,
+      "loss": 0.0476,
+      "step": 11830
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.5762586542985818e-05,
+      "loss": 0.0559,
+      "step": 11840
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.572897761645493e-05,
+      "loss": 0.0472,
+      "step": 11850
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.5695368689924044e-05,
+      "loss": 0.0557,
+      "step": 11860
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.5661759763393158e-05,
+      "loss": 0.0561,
+      "step": 11870
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.562815083686227e-05,
+      "loss": 0.06,
+      "step": 11880
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.5594541910331384e-05,
+      "loss": 0.0583,
+      "step": 11890
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.5560932983800497e-05,
+      "loss": 0.0719,
+      "step": 11900
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.5527324057269614e-05,
+      "loss": 0.0631,
+      "step": 11910
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.5493715130738727e-05,
+      "loss": 0.0783,
+      "step": 11920
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.546010620420784e-05,
+      "loss": 0.0584,
+      "step": 11930
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.542649727767695e-05,
+      "loss": 0.0533,
+      "step": 11940
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.5392888351146063e-05,
+      "loss": 0.0607,
+      "step": 11950
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.5359279424615177e-05,
+      "loss": 0.0729,
+      "step": 11960
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.532567049808429e-05,
+      "loss": 0.0619,
+      "step": 11970
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.5292061571553407e-05,
+      "loss": 0.0633,
+      "step": 11980
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.525845264502252e-05,
+      "loss": 0.0523,
+      "step": 11990
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.5224843718491633e-05,
+      "loss": 0.0634,
+      "step": 12000
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.5191234791960746e-05,
+      "loss": 0.0624,
+      "step": 12010
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.5157625865429858e-05,
+      "loss": 0.0606,
+      "step": 12020
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.5124016938898971e-05,
+      "loss": 0.06,
+      "step": 12030
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.5090408012368084e-05,
+      "loss": 0.0563,
+      "step": 12040
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.5056799085837201e-05,
+      "loss": 0.0725,
+      "step": 12050
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.5023190159306314e-05,
+      "loss": 0.0551,
+      "step": 12060
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.4989581232775426e-05,
+      "loss": 0.0665,
+      "step": 12070
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.4955972306244539e-05,
+      "loss": 0.0552,
+      "step": 12080
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.4922363379713652e-05,
+      "loss": 0.0585,
+      "step": 12090
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.4888754453182765e-05,
+      "loss": 0.0433,
+      "step": 12100
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.4855145526651879e-05,
+      "loss": 0.0699,
+      "step": 12110
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.4821536600120992e-05,
+      "loss": 0.0642,
+      "step": 12120
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.9913265306122448,
+      "eval_loss": 0.026108374819159508,
+      "eval_runtime": 89.8581,
+      "eval_samples_per_second": 130.873,
+      "eval_steps_per_second": 4.095,
+      "step": 12122
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.4787927673590107e-05,
+      "loss": 0.0492,
+      "step": 12130
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.475431874705922e-05,
+      "loss": 0.0549,
+      "step": 12140
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.4720709820528333e-05,
+      "loss": 0.073,
+      "step": 12150
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.4687100893997447e-05,
+      "loss": 0.0549,
+      "step": 12160
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.465349196746656e-05,
+      "loss": 0.0548,
+      "step": 12170
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.4619883040935673e-05,
+      "loss": 0.0586,
+      "step": 12180
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.4586274114404785e-05,
+      "loss": 0.0655,
+      "step": 12190
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.4552665187873901e-05,
+      "loss": 0.0431,
+      "step": 12200
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.4519056261343015e-05,
+      "loss": 0.0469,
+      "step": 12210
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.4485447334812128e-05,
+      "loss": 0.0619,
+      "step": 12220
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.4451838408281241e-05,
+      "loss": 0.0617,
+      "step": 12230
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.4418229481750353e-05,
+      "loss": 0.0627,
+      "step": 12240
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.4384620555219466e-05,
+      "loss": 0.0783,
+      "step": 12250
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.4351011628688579e-05,
+      "loss": 0.0535,
+      "step": 12260
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.4317402702157696e-05,
+      "loss": 0.0459,
+      "step": 12270
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.4283793775626807e-05,
+      "loss": 0.0643,
+      "step": 12280
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.425018484909592e-05,
+      "loss": 0.0682,
+      "step": 12290
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.4216575922565034e-05,
+      "loss": 0.0481,
+      "step": 12300
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.4182966996034147e-05,
+      "loss": 0.0554,
+      "step": 12310
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.414935806950326e-05,
+      "loss": 0.0577,
+      "step": 12320
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.4115749142972373e-05,
+      "loss": 0.0579,
+      "step": 12330
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.4082140216441487e-05,
+      "loss": 0.0616,
+      "step": 12340
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.4048531289910602e-05,
+      "loss": 0.0447,
+      "step": 12350
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.4014922363379715e-05,
+      "loss": 0.0595,
+      "step": 12360
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.3981313436848828e-05,
+      "loss": 0.0526,
+      "step": 12370
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.3947704510317941e-05,
+      "loss": 0.0565,
+      "step": 12380
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.3914095583787055e-05,
+      "loss": 0.0706,
+      "step": 12390
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.3880486657256168e-05,
+      "loss": 0.038,
+      "step": 12400
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.384687773072528e-05,
+      "loss": 0.0733,
+      "step": 12410
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.3813268804194396e-05,
+      "loss": 0.0435,
+      "step": 12420
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 1.377965987766351e-05,
+      "loss": 0.0634,
+      "step": 12430
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 1.3746050951132623e-05,
+      "loss": 0.0487,
+      "step": 12440
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 1.3712442024601734e-05,
+      "loss": 0.0474,
+      "step": 12450
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 1.3678833098070847e-05,
+      "loss": 0.0516,
+      "step": 12460
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 1.364522417153996e-05,
+      "loss": 0.0638,
+      "step": 12470
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 1.3611615245009074e-05,
+      "loss": 0.0544,
+      "step": 12480
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 1.357800631847819e-05,
+      "loss": 0.0543,
+      "step": 12490
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 1.3544397391947302e-05,
+      "loss": 0.0732,
+      "step": 12500
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 1.3510788465416415e-05,
+      "loss": 0.0489,
+      "step": 12510
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 1.3477179538885528e-05,
+      "loss": 0.049,
+      "step": 12520
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 1.3443570612354642e-05,
+      "loss": 0.0542,
+      "step": 12530
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.3409961685823755e-05,
+      "loss": 0.0441,
+      "step": 12540
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 1.3376352759292868e-05,
+      "loss": 0.0571,
+      "step": 12550
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 1.3342743832761983e-05,
+      "loss": 0.0571,
+      "step": 12560
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 1.3309134906231096e-05,
+      "loss": 0.0463,
+      "step": 12570
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 1.327552597970021e-05,
+      "loss": 0.0619,
+      "step": 12580
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 1.3241917053169323e-05,
+      "loss": 0.0636,
+      "step": 12590
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 1.3208308126638436e-05,
+      "loss": 0.0562,
+      "step": 12600
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 1.317469920010755e-05,
+      "loss": 0.0626,
+      "step": 12610
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 1.3141090273576661e-05,
+      "loss": 0.0593,
+      "step": 12620
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 1.3107481347045774e-05,
+      "loss": 0.0601,
+      "step": 12630
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 1.307387242051489e-05,
+      "loss": 0.0728,
+      "step": 12640
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 1.3040263493984004e-05,
+      "loss": 0.0654,
+      "step": 12650
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 1.3006654567453117e-05,
+      "loss": 0.0471,
+      "step": 12660
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 1.2973045640922229e-05,
+      "loss": 0.0695,
+      "step": 12670
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.9920068027210884,
+      "eval_loss": 0.025952747091650963,
+      "eval_runtime": 89.5891,
+      "eval_samples_per_second": 131.266,
+      "eval_steps_per_second": 4.108,
+      "step": 12673
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 1.2939436714391342e-05,
+      "loss": 0.0492,
+      "step": 12680
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 1.2905827787860455e-05,
+      "loss": 0.0608,
+      "step": 12690
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 1.2872218861329569e-05,
+      "loss": 0.0597,
+      "step": 12700
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 1.2838609934798685e-05,
+      "loss": 0.064,
+      "step": 12710
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 1.2805001008267797e-05,
+      "loss": 0.0429,
+      "step": 12720
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 1.277139208173691e-05,
+      "loss": 0.065,
+      "step": 12730
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 1.2737783155206023e-05,
+      "loss": 0.0354,
+      "step": 12740
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 1.2704174228675136e-05,
+      "loss": 0.0745,
+      "step": 12750
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 1.267056530214425e-05,
+      "loss": 0.0426,
+      "step": 12760
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 1.2636956375613363e-05,
+      "loss": 0.0457,
+      "step": 12770
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 1.2603347449082478e-05,
+      "loss": 0.046,
+      "step": 12780
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 1.2569738522551591e-05,
+      "loss": 0.065,
+      "step": 12790
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 1.2536129596020704e-05,
+      "loss": 0.0453,
+      "step": 12800
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 1.2502520669489818e-05,
+      "loss": 0.0518,
+      "step": 12810
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 1.2468911742958931e-05,
+      "loss": 0.0702,
+      "step": 12820
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 1.2435302816428044e-05,
+      "loss": 0.0612,
+      "step": 12830
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 1.2401693889897157e-05,
+      "loss": 0.0695,
+      "step": 12840
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 1.236808496336627e-05,
+      "loss": 0.0632,
+      "step": 12850
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 1.2334476036835384e-05,
+      "loss": 0.0531,
+      "step": 12860
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 1.2300867110304499e-05,
+      "loss": 0.0427,
+      "step": 12870
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 1.226725818377361e-05,
+      "loss": 0.0758,
+      "step": 12880
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 1.2233649257242724e-05,
+      "loss": 0.0539,
+      "step": 12890
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 1.2200040330711837e-05,
+      "loss": 0.0455,
+      "step": 12900
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 1.2166431404180952e-05,
+      "loss": 0.048,
+      "step": 12910
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 1.2132822477650065e-05,
+      "loss": 0.0817,
+      "step": 12920
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 1.2099213551119178e-05,
+      "loss": 0.0718,
+      "step": 12930
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 1.206560462458829e-05,
+      "loss": 0.067,
+      "step": 12940
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 1.2031995698057405e-05,
+      "loss": 0.0496,
+      "step": 12950
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 1.1998386771526518e-05,
+      "loss": 0.0685,
+      "step": 12960
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 1.1964777844995631e-05,
+      "loss": 0.0497,
+      "step": 12970
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 1.1931168918464746e-05,
+      "loss": 0.062,
+      "step": 12980
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 1.1897559991933858e-05,
+      "loss": 0.0556,
+      "step": 12990
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 1.1863951065402971e-05,
+      "loss": 0.044,
+      "step": 13000
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 1.1830342138872084e-05,
+      "loss": 0.0687,
+      "step": 13010
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 1.1796733212341199e-05,
+      "loss": 0.0577,
+      "step": 13020
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 1.1763124285810312e-05,
+      "loss": 0.0624,
+      "step": 13030
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 1.1729515359279426e-05,
+      "loss": 0.0558,
+      "step": 13040
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 1.1695906432748537e-05,
+      "loss": 0.0523,
+      "step": 13050
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 1.1662297506217652e-05,
+      "loss": 0.0483,
+      "step": 13060
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 1.1628688579686765e-05,
+      "loss": 0.041,
+      "step": 13070
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 1.1595079653155879e-05,
+      "loss": 0.0839,
+      "step": 13080
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 1.1561470726624994e-05,
+      "loss": 0.06,
+      "step": 13090
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 1.1527861800094105e-05,
+      "loss": 0.0437,
+      "step": 13100
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 1.1494252873563218e-05,
+      "loss": 0.0602,
+      "step": 13110
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 1.1460643947032332e-05,
+      "loss": 0.0591,
+      "step": 13120
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 1.1427035020501446e-05,
+      "loss": 0.051,
+      "step": 13130
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 1.139342609397056e-05,
+      "loss": 0.0444,
+      "step": 13140
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 1.1359817167439673e-05,
+      "loss": 0.0602,
+      "step": 13150
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 1.1326208240908785e-05,
+      "loss": 0.0778,
+      "step": 13160
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 1.12925993143779e-05,
+      "loss": 0.0694,
+      "step": 13170
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 1.1258990387847013e-05,
+      "loss": 0.0532,
+      "step": 13180
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 1.1225381461316126e-05,
+      "loss": 0.0745,
+      "step": 13190
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 1.1191772534785241e-05,
+      "loss": 0.0419,
+      "step": 13200
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 1.1158163608254352e-05,
+      "loss": 0.0497,
+      "step": 13210
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 1.1124554681723466e-05,
+      "loss": 0.0664,
+      "step": 13220
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.9927721088435374,
+      "eval_loss": 0.02406076155602932,
+      "eval_runtime": 88.7181,
+      "eval_samples_per_second": 132.555,
+      "eval_steps_per_second": 4.148,
+      "step": 13224
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 1.1090945755192579e-05,
+      "loss": 0.0527,
+      "step": 13230
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 1.1057336828661694e-05,
+      "loss": 0.0557,
+      "step": 13240
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 1.1023727902130807e-05,
+      "loss": 0.0513,
+      "step": 13250
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 1.099011897559992e-05,
+      "loss": 0.0484,
+      "step": 13260
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 1.0956510049069034e-05,
+      "loss": 0.0577,
+      "step": 13270
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 1.0922901122538147e-05,
+      "loss": 0.0664,
+      "step": 13280
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 1.088929219600726e-05,
+      "loss": 0.0708,
+      "step": 13290
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 1.0855683269476373e-05,
+      "loss": 0.0533,
+      "step": 13300
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 1.0822074342945487e-05,
+      "loss": 0.0544,
+      "step": 13310
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 1.07884654164146e-05,
+      "loss": 0.045,
+      "step": 13320
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 1.0754856489883713e-05,
+      "loss": 0.0492,
+      "step": 13330
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 1.0721247563352826e-05,
+      "loss": 0.0662,
+      "step": 13340
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 1.0687638636821941e-05,
+      "loss": 0.0539,
+      "step": 13350
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 1.0654029710291054e-05,
+      "loss": 0.0501,
+      "step": 13360
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 1.0620420783760168e-05,
+      "loss": 0.0508,
+      "step": 13370
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 1.0586811857229281e-05,
+      "loss": 0.0608,
+      "step": 13380
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 1.0553202930698394e-05,
+      "loss": 0.0582,
+      "step": 13390
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 1.0519594004167507e-05,
+      "loss": 0.0359,
+      "step": 13400
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 1.048598507763662e-05,
+      "loss": 0.0517,
+      "step": 13410
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 1.0452376151105734e-05,
+      "loss": 0.0589,
+      "step": 13420
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 1.0418767224574847e-05,
+      "loss": 0.0515,
+      "step": 13430
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 1.038515829804396e-05,
+      "loss": 0.0454,
+      "step": 13440
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 1.0351549371513074e-05,
+      "loss": 0.0442,
+      "step": 13450
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 1.0317940444982189e-05,
+      "loss": 0.0446,
+      "step": 13460
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 1.0284331518451302e-05,
+      "loss": 0.0595,
+      "step": 13470
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 1.0250722591920413e-05,
+      "loss": 0.0363,
+      "step": 13480
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 1.0217113665389528e-05,
+      "loss": 0.0671,
+      "step": 13490
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 1.0183504738858642e-05,
+      "loss": 0.0583,
+      "step": 13500
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 1.0149895812327755e-05,
+      "loss": 0.0438,
+      "step": 13510
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 1.0116286885796868e-05,
+      "loss": 0.0637,
+      "step": 13520
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 1.0082677959265981e-05,
+      "loss": 0.0582,
+      "step": 13530
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 1.0049069032735095e-05,
+      "loss": 0.0533,
+      "step": 13540
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 1.0015460106204208e-05,
+      "loss": 0.0463,
+      "step": 13550
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 9.981851179673321e-06,
+      "loss": 0.0602,
+      "step": 13560
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 9.948242253142436e-06,
+      "loss": 0.0478,
+      "step": 13570
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 9.91463332661155e-06,
+      "loss": 0.0352,
+      "step": 13580
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 9.88102440008066e-06,
+      "loss": 0.0661,
+      "step": 13590
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 9.847415473549776e-06,
+      "loss": 0.0471,
+      "step": 13600
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 9.813806547018889e-06,
+      "loss": 0.0621,
+      "step": 13610
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 9.780197620488002e-06,
+      "loss": 0.0494,
+      "step": 13620
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 9.746588693957115e-06,
+      "loss": 0.0666,
+      "step": 13630
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 9.712979767426229e-06,
+      "loss": 0.0627,
+      "step": 13640
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 9.679370840895342e-06,
+      "loss": 0.066,
+      "step": 13650
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 9.645761914364455e-06,
+      "loss": 0.0556,
+      "step": 13660
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 9.612152987833568e-06,
+      "loss": 0.0446,
+      "step": 13670
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 9.578544061302683e-06,
+      "loss": 0.0548,
+      "step": 13680
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 9.544935134771797e-06,
+      "loss": 0.0534,
+      "step": 13690
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 9.511326208240908e-06,
+      "loss": 0.053,
+      "step": 13700
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 9.477717281710023e-06,
+      "loss": 0.0375,
+      "step": 13710
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 9.444108355179136e-06,
+      "loss": 0.0436,
+      "step": 13720
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 9.41049942864825e-06,
+      "loss": 0.0546,
+      "step": 13730
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 9.376890502117363e-06,
+      "loss": 0.0477,
+      "step": 13740
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 9.343281575586476e-06,
+      "loss": 0.0479,
+      "step": 13750
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 9.30967264905559e-06,
+      "loss": 0.0571,
+      "step": 13760
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 9.276063722524703e-06,
+      "loss": 0.0552,
+      "step": 13770
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.9927721088435374,
+      "eval_loss": 0.025815660133957863,
+      "eval_runtime": 87.3188,
+      "eval_samples_per_second": 134.679,
+      "eval_steps_per_second": 4.214,
+      "step": 13775
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 9.242454795993817e-06,
+      "loss": 0.0509,
+      "step": 13780
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 9.20884586946293e-06,
+      "loss": 0.0643,
+      "step": 13790
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 9.175236942932044e-06,
+      "loss": 0.066,
+      "step": 13800
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 9.141628016401155e-06,
+      "loss": 0.0571,
+      "step": 13810
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 9.10801908987027e-06,
+      "loss": 0.0605,
+      "step": 13820
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 9.074410163339384e-06,
+      "loss": 0.059,
+      "step": 13830
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 9.040801236808497e-06,
+      "loss": 0.0467,
+      "step": 13840
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 9.00719231027761e-06,
+      "loss": 0.0615,
+      "step": 13850
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 8.973583383746723e-06,
+      "loss": 0.0514,
+      "step": 13860
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 8.939974457215837e-06,
+      "loss": 0.05,
+      "step": 13870
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 8.90636553068495e-06,
+      "loss": 0.0513,
+      "step": 13880
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 8.872756604154065e-06,
+      "loss": 0.0489,
+      "step": 13890
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 8.839147677623178e-06,
+      "loss": 0.059,
+      "step": 13900
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 8.80553875109229e-06,
+      "loss": 0.0495,
+      "step": 13910
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 8.771929824561403e-06,
+      "loss": 0.0453,
+      "step": 13920
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 8.738320898030518e-06,
+      "loss": 0.0647,
+      "step": 13930
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 8.704711971499631e-06,
+      "loss": 0.0528,
+      "step": 13940
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 8.671103044968744e-06,
+      "loss": 0.0389,
+      "step": 13950
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 8.637494118437858e-06,
+      "loss": 0.048,
+      "step": 13960
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 8.60388519190697e-06,
+      "loss": 0.0681,
+      "step": 13970
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 8.570276265376084e-06,
+      "loss": 0.0643,
+      "step": 13980
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 8.536667338845197e-06,
+      "loss": 0.0494,
+      "step": 13990
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 8.503058412314312e-06,
+      "loss": 0.0475,
+      "step": 14000
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 8.469449485783425e-06,
+      "loss": 0.0429,
+      "step": 14010
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 8.435840559252537e-06,
+      "loss": 0.064,
+      "step": 14020
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 8.40223163272165e-06,
+      "loss": 0.0601,
+      "step": 14030
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 8.368622706190765e-06,
+      "loss": 0.0587,
+      "step": 14040
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 8.335013779659878e-06,
+      "loss": 0.0439,
+      "step": 14050
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 8.301404853128992e-06,
+      "loss": 0.0594,
+      "step": 14060
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 8.267795926598105e-06,
+      "loss": 0.0501,
+      "step": 14070
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 8.234187000067218e-06,
+      "loss": 0.053,
+      "step": 14080
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 8.200578073536331e-06,
+      "loss": 0.044,
+      "step": 14090
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 8.166969147005445e-06,
+      "loss": 0.0459,
+      "step": 14100
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 8.13336022047456e-06,
+      "loss": 0.0575,
+      "step": 14110
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 8.099751293943673e-06,
+      "loss": 0.06,
+      "step": 14120
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 8.066142367412784e-06,
+      "loss": 0.0628,
+      "step": 14130
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 8.032533440881898e-06,
+      "loss": 0.0639,
+      "step": 14140
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 7.998924514351013e-06,
+      "loss": 0.0468,
+      "step": 14150
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 7.965315587820126e-06,
+      "loss": 0.0425,
+      "step": 14160
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 7.931706661289239e-06,
+      "loss": 0.0461,
+      "step": 14170
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 7.898097734758352e-06,
+      "loss": 0.0411,
+      "step": 14180
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 7.864488808227466e-06,
+      "loss": 0.052,
+      "step": 14190
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 7.830879881696579e-06,
+      "loss": 0.0648,
+      "step": 14200
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 7.797270955165692e-06,
+      "loss": 0.0479,
+      "step": 14210
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 7.763662028634807e-06,
+      "loss": 0.0527,
+      "step": 14220
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 7.73005310210392e-06,
+      "loss": 0.0468,
+      "step": 14230
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 7.696444175573032e-06,
+      "loss": 0.0595,
+      "step": 14240
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 7.662835249042145e-06,
+      "loss": 0.0449,
+      "step": 14250
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 7.62922632251126e-06,
+      "loss": 0.062,
+      "step": 14260
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 7.595617395980373e-06,
+      "loss": 0.0532,
+      "step": 14270
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 7.5620084694494855e-06,
+      "loss": 0.0462,
+      "step": 14280
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 7.5283995429186005e-06,
+      "loss": 0.0548,
+      "step": 14290
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 7.494790616387713e-06,
+      "loss": 0.052,
+      "step": 14300
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 7.461181689856826e-06,
+      "loss": 0.0478,
+      "step": 14310
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 7.427572763325939e-06,
+      "loss": 0.056,
+      "step": 14320
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.9938775510204082,
+      "eval_loss": 0.023027613759040833,
+      "eval_runtime": 89.8862,
+      "eval_samples_per_second": 130.832,
+      "eval_steps_per_second": 4.094,
+      "step": 14326
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 7.3939638367950535e-06,
+      "loss": 0.0689,
+      "step": 14330
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 7.360354910264167e-06,
+      "loss": 0.0544,
+      "step": 14340
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 7.32674598373328e-06,
+      "loss": 0.0369,
+      "step": 14350
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 7.293137057202392e-06,
+      "loss": 0.0516,
+      "step": 14360
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 7.259528130671507e-06,
+      "loss": 0.0402,
+      "step": 14370
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 7.2259192041406205e-06,
+      "loss": 0.0433,
+      "step": 14380
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 7.192310277609733e-06,
+      "loss": 0.0574,
+      "step": 14390
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 7.158701351078848e-06,
+      "loss": 0.0565,
+      "step": 14400
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 7.12509242454796e-06,
+      "loss": 0.0436,
+      "step": 14410
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 7.0914834980170735e-06,
+      "loss": 0.0537,
+      "step": 14420
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 7.057874571486187e-06,
+      "loss": 0.0491,
+      "step": 14430
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 7.024265644955301e-06,
+      "loss": 0.0467,
+      "step": 14440
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 6.990656718424414e-06,
+      "loss": 0.0553,
+      "step": 14450
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 6.957047791893527e-06,
+      "loss": 0.0454,
+      "step": 14460
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 6.92343886536264e-06,
+      "loss": 0.066,
+      "step": 14470
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 6.889829938831755e-06,
+      "loss": 0.0462,
+      "step": 14480
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 6.856221012300867e-06,
+      "loss": 0.0383,
+      "step": 14490
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 6.82261208576998e-06,
+      "loss": 0.062,
+      "step": 14500
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 6.789003159239095e-06,
+      "loss": 0.0584,
+      "step": 14510
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 6.755394232708208e-06,
+      "loss": 0.0591,
+      "step": 14520
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 6.721785306177321e-06,
+      "loss": 0.0581,
+      "step": 14530
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 6.688176379646434e-06,
+      "loss": 0.061,
+      "step": 14540
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 6.654567453115548e-06,
+      "loss": 0.0389,
+      "step": 14550
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 6.6209585265846614e-06,
+      "loss": 0.0427,
+      "step": 14560
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 6.587349600053775e-06,
+      "loss": 0.0563,
+      "step": 14570
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 6.553740673522887e-06,
+      "loss": 0.0466,
+      "step": 14580
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 6.520131746992002e-06,
+      "loss": 0.0532,
+      "step": 14590
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 6.486522820461114e-06,
+      "loss": 0.0518,
+      "step": 14600
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 6.452913893930228e-06,
+      "loss": 0.0619,
+      "step": 14610
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 6.419304967399343e-06,
+      "loss": 0.0384,
+      "step": 14620
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 6.385696040868455e-06,
+      "loss": 0.0455,
+      "step": 14630
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 6.352087114337568e-06,
+      "loss": 0.0473,
+      "step": 14640
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 6.3184781878066815e-06,
+      "loss": 0.0501,
+      "step": 14650
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 6.2848692612757956e-06,
+      "loss": 0.0517,
+      "step": 14660
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 6.251260334744909e-06,
+      "loss": 0.0446,
+      "step": 14670
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 6.217651408214022e-06,
+      "loss": 0.077,
+      "step": 14680
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 6.184042481683135e-06,
+      "loss": 0.0425,
+      "step": 14690
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 6.150433555152249e-06,
+      "loss": 0.0399,
+      "step": 14700
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 6.116824628621362e-06,
+      "loss": 0.0551,
+      "step": 14710
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 6.083215702090476e-06,
+      "loss": 0.0499,
+      "step": 14720
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 6.049606775559589e-06,
+      "loss": 0.0518,
+      "step": 14730
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 6.015997849028702e-06,
+      "loss": 0.036,
+      "step": 14740
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 5.982388922497816e-06,
+      "loss": 0.0414,
+      "step": 14750
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 5.948779995966929e-06,
+      "loss": 0.0557,
+      "step": 14760
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 5.915171069436042e-06,
+      "loss": 0.0374,
+      "step": 14770
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 5.881562142905156e-06,
+      "loss": 0.0527,
+      "step": 14780
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 5.8479532163742686e-06,
+      "loss": 0.0378,
+      "step": 14790
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 5.814344289843383e-06,
+      "loss": 0.051,
+      "step": 14800
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 5.780735363312497e-06,
+      "loss": 0.0338,
+      "step": 14810
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 5.747126436781609e-06,
+      "loss": 0.0416,
+      "step": 14820
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 5.713517510250723e-06,
+      "loss": 0.0451,
+      "step": 14830
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 5.6799085837198365e-06,
+      "loss": 0.0407,
+      "step": 14840
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 5.64629965718895e-06,
+      "loss": 0.0534,
+      "step": 14850
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 5.612690730658063e-06,
+      "loss": 0.0526,
+      "step": 14860
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 5.579081804127176e-06,
+      "loss": 0.0488,
+      "step": 14870
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.9936224489795918,
+      "eval_loss": 0.022101709619164467,
+      "eval_runtime": 90.9922,
+      "eval_samples_per_second": 129.242,
+      "eval_steps_per_second": 4.044,
+      "step": 14877
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 5.5454728775962895e-06,
+      "loss": 0.039,
+      "step": 14880
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 5.5118639510654035e-06,
+      "loss": 0.0386,
+      "step": 14890
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 5.478255024534517e-06,
+      "loss": 0.0458,
+      "step": 14900
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 5.44464609800363e-06,
+      "loss": 0.035,
+      "step": 14910
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 5.411037171472743e-06,
+      "loss": 0.0527,
+      "step": 14920
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 5.3774282449418565e-06,
+      "loss": 0.0541,
+      "step": 14930
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 5.343819318410971e-06,
+      "loss": 0.0303,
+      "step": 14940
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 5.310210391880084e-06,
+      "loss": 0.0581,
+      "step": 14950
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 5.276601465349197e-06,
+      "loss": 0.0561,
+      "step": 14960
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 5.24299253881831e-06,
+      "loss": 0.0424,
+      "step": 14970
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 5.209383612287424e-06,
+      "loss": 0.0407,
+      "step": 14980
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 5.175774685756537e-06,
+      "loss": 0.0548,
+      "step": 14990
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 5.142165759225651e-06,
+      "loss": 0.0525,
+      "step": 15000
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 5.108556832694764e-06,
+      "loss": 0.0591,
+      "step": 15010
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 5.074947906163877e-06,
+      "loss": 0.0329,
+      "step": 15020
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 5.041338979632991e-06,
+      "loss": 0.0435,
+      "step": 15030
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 5.007730053102104e-06,
+      "loss": 0.0541,
+      "step": 15040
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 4.974121126571218e-06,
+      "loss": 0.0484,
+      "step": 15050
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 4.94051220004033e-06,
+      "loss": 0.0495,
+      "step": 15060
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 4.9069032735094445e-06,
+      "loss": 0.0349,
+      "step": 15070
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 4.873294346978558e-06,
+      "loss": 0.0581,
+      "step": 15080
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 4.839685420447671e-06,
+      "loss": 0.0478,
+      "step": 15090
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 4.806076493916784e-06,
+      "loss": 0.0517,
+      "step": 15100
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 4.772467567385898e-06,
+      "loss": 0.0429,
+      "step": 15110
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 4.7388586408550115e-06,
+      "loss": 0.049,
+      "step": 15120
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 4.705249714324125e-06,
+      "loss": 0.0526,
+      "step": 15130
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 4.671640787793238e-06,
+      "loss": 0.0486,
+      "step": 15140
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 4.638031861262351e-06,
+      "loss": 0.0594,
+      "step": 15150
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 4.604422934731465e-06,
+      "loss": 0.0512,
+      "step": 15160
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 4.570814008200578e-06,
+      "loss": 0.0489,
+      "step": 15170
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 4.537205081669692e-06,
+      "loss": 0.0374,
+      "step": 15180
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 4.503596155138805e-06,
+      "loss": 0.0612,
+      "step": 15190
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 4.469987228607918e-06,
+      "loss": 0.0444,
+      "step": 15200
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 4.436378302077032e-06,
+      "loss": 0.0473,
+      "step": 15210
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 4.402769375546145e-06,
+      "loss": 0.0415,
+      "step": 15220
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 4.369160449015259e-06,
+      "loss": 0.0537,
+      "step": 15230
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 4.335551522484372e-06,
+      "loss": 0.0507,
+      "step": 15240
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 4.301942595953485e-06,
+      "loss": 0.043,
+      "step": 15250
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 4.268333669422599e-06,
+      "loss": 0.0498,
+      "step": 15260
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 4.234724742891713e-06,
+      "loss": 0.0538,
+      "step": 15270
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 4.201115816360825e-06,
+      "loss": 0.0453,
+      "step": 15280
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 4.167506889829939e-06,
+      "loss": 0.0474,
+      "step": 15290
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 4.1338979632990524e-06,
+      "loss": 0.0531,
+      "step": 15300
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 4.100289036768166e-06,
+      "loss": 0.0615,
+      "step": 15310
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 4.06668011023728e-06,
+      "loss": 0.0353,
+      "step": 15320
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 4.033071183706392e-06,
+      "loss": 0.0507,
+      "step": 15330
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.999462257175506e-06,
+      "loss": 0.0588,
+      "step": 15340
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.9658533306446195e-06,
+      "loss": 0.0368,
+      "step": 15350
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.932244404113733e-06,
+      "loss": 0.056,
+      "step": 15360
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.898635477582846e-06,
+      "loss": 0.0437,
+      "step": 15370
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.86502655105196e-06,
+      "loss": 0.0507,
+      "step": 15380
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.8314176245210725e-06,
+      "loss": 0.0474,
+      "step": 15390
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.7978086979901866e-06,
+      "loss": 0.0398,
+      "step": 15400
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.7641997714593002e-06,
+      "loss": 0.0356,
+      "step": 15410
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.730590844928413e-06,
+      "loss": 0.0389,
+      "step": 15420
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.9930272108843538,
+      "eval_loss": 0.022541489452123642,
+      "eval_runtime": 88.4176,
+      "eval_samples_per_second": 133.005,
+      "eval_steps_per_second": 4.162,
+      "step": 15428
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.6969819183975267e-06,
+      "loss": 0.0464,
+      "step": 15430
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.66337299186664e-06,
+      "loss": 0.0473,
+      "step": 15440
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.6297640653357536e-06,
+      "loss": 0.0477,
+      "step": 15450
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.5961551388048665e-06,
+      "loss": 0.0398,
+      "step": 15460
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.56254621227398e-06,
+      "loss": 0.0496,
+      "step": 15470
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.5289372857430934e-06,
+      "loss": 0.0526,
+      "step": 15480
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.495328359212207e-06,
+      "loss": 0.0594,
+      "step": 15490
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.46171943268132e-06,
+      "loss": 0.051,
+      "step": 15500
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.4281105061504335e-06,
+      "loss": 0.0407,
+      "step": 15510
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.3945015796195476e-06,
+      "loss": 0.0619,
+      "step": 15520
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.3608926530886604e-06,
+      "loss": 0.0603,
+      "step": 15530
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.327283726557774e-06,
+      "loss": 0.0355,
+      "step": 15540
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.2936748000268873e-06,
+      "loss": 0.0551,
+      "step": 15550
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.260065873496001e-06,
+      "loss": 0.0416,
+      "step": 15560
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.226456946965114e-06,
+      "loss": 0.0377,
+      "step": 15570
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.1928480204342275e-06,
+      "loss": 0.052,
+      "step": 15580
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.1592390939033407e-06,
+      "loss": 0.0318,
+      "step": 15590
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.1256301673724544e-06,
+      "loss": 0.0362,
+      "step": 15600
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.0920212408415676e-06,
+      "loss": 0.0575,
+      "step": 15610
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.058412314310681e-06,
+      "loss": 0.0479,
+      "step": 15620
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.0248033877797946e-06,
+      "loss": 0.0511,
+      "step": 15630
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 2.991194461248908e-06,
+      "loss": 0.0333,
+      "step": 15640
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2.957585534718021e-06,
+      "loss": 0.0544,
+      "step": 15650
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 2.9239766081871343e-06,
+      "loss": 0.0534,
+      "step": 15660
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 2.8903676816562484e-06,
+      "loss": 0.0393,
+      "step": 15670
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 2.8567587551253616e-06,
+      "loss": 0.0439,
+      "step": 15680
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 2.823149828594475e-06,
+      "loss": 0.0446,
+      "step": 15690
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.789540902063588e-06,
+      "loss": 0.0533,
+      "step": 15700
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 2.7559319755327018e-06,
+      "loss": 0.0483,
+      "step": 15710
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 2.722323049001815e-06,
+      "loss": 0.0476,
+      "step": 15720
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 2.6887141224709283e-06,
+      "loss": 0.0437,
+      "step": 15730
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 2.655105195940042e-06,
+      "loss": 0.0382,
+      "step": 15740
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 2.621496269409155e-06,
+      "loss": 0.0372,
+      "step": 15750
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 2.5878873428782684e-06,
+      "loss": 0.0371,
+      "step": 15760
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 2.554278416347382e-06,
+      "loss": 0.0401,
+      "step": 15770
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 2.5206694898164953e-06,
+      "loss": 0.0407,
+      "step": 15780
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 2.487060563285609e-06,
+      "loss": 0.0405,
+      "step": 15790
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 2.4534516367547222e-06,
+      "loss": 0.0456,
+      "step": 15800
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 2.4198427102238355e-06,
+      "loss": 0.0369,
+      "step": 15810
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 2.386233783692949e-06,
+      "loss": 0.0388,
+      "step": 15820
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 2.3526248571620624e-06,
+      "loss": 0.0497,
+      "step": 15830
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 2.3190159306311756e-06,
+      "loss": 0.0486,
+      "step": 15840
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 2.285407004100289e-06,
+      "loss": 0.0631,
+      "step": 15850
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 2.2517980775694025e-06,
+      "loss": 0.0443,
+      "step": 15860
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 2.218189151038516e-06,
+      "loss": 0.0378,
+      "step": 15870
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 2.1845802245076294e-06,
+      "loss": 0.0498,
+      "step": 15880
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 2.1509712979767427e-06,
+      "loss": 0.0307,
+      "step": 15890
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 2.1173623714458564e-06,
+      "loss": 0.0438,
+      "step": 15900
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 2.0837534449149696e-06,
+      "loss": 0.0412,
+      "step": 15910
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 2.050144518384083e-06,
+      "loss": 0.043,
+      "step": 15920
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 2.016535591853196e-06,
+      "loss": 0.0612,
+      "step": 15930
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.9829266653223098e-06,
+      "loss": 0.0455,
+      "step": 15940
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.949317738791423e-06,
+      "loss": 0.0336,
+      "step": 15950
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.9157088122605362e-06,
+      "loss": 0.0362,
+      "step": 15960
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.8820998857296501e-06,
+      "loss": 0.0402,
+      "step": 15970
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.9940476190476191,
+      "eval_loss": 0.023117393255233765,
+      "eval_runtime": 88.4632,
+      "eval_samples_per_second": 132.937,
+      "eval_steps_per_second": 4.16,
+      "step": 15979
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.8484909591987634e-06,
+      "loss": 0.0416,
+      "step": 15980
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.8148820326678768e-06,
+      "loss": 0.0464,
+      "step": 15990
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.78127310613699e-06,
+      "loss": 0.0369,
+      "step": 16000
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.7476641796061035e-06,
+      "loss": 0.0481,
+      "step": 16010
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.7140552530752168e-06,
+      "loss": 0.0453,
+      "step": 16020
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.6804463265443302e-06,
+      "loss": 0.0461,
+      "step": 16030
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.6468374000134437e-06,
+      "loss": 0.0424,
+      "step": 16040
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.613228473482557e-06,
+      "loss": 0.0502,
+      "step": 16050
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.5796195469516704e-06,
+      "loss": 0.0456,
+      "step": 16060
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.5460106204207838e-06,
+      "loss": 0.0469,
+      "step": 16070
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.5124016938898973e-06,
+      "loss": 0.0493,
+      "step": 16080
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.4787927673590105e-06,
+      "loss": 0.0339,
+      "step": 16090
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.4451838408281242e-06,
+      "loss": 0.0557,
+      "step": 16100
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.4115749142972374e-06,
+      "loss": 0.0577,
+      "step": 16110
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 1.3779659877663509e-06,
+      "loss": 0.0446,
+      "step": 16120
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 1.3443570612354641e-06,
+      "loss": 0.0286,
+      "step": 16130
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 1.3107481347045776e-06,
+      "loss": 0.0425,
+      "step": 16140
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 1.277139208173691e-06,
+      "loss": 0.0375,
+      "step": 16150
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 1.2435302816428045e-06,
+      "loss": 0.0365,
+      "step": 16160
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 1.2099213551119177e-06,
+      "loss": 0.0473,
+      "step": 16170
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 1.1763124285810312e-06,
+      "loss": 0.0644,
+      "step": 16180
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 1.1427035020501444e-06,
+      "loss": 0.0723,
+      "step": 16190
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 1.109094575519258e-06,
+      "loss": 0.0454,
+      "step": 16200
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 1.0754856489883713e-06,
+      "loss": 0.0434,
+      "step": 16210
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 1.0418767224574848e-06,
+      "loss": 0.0451,
+      "step": 16220
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 1.008267795926598e-06,
+      "loss": 0.0519,
+      "step": 16230
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 9.746588693957115e-07,
+      "loss": 0.0468,
+      "step": 16240
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 9.410499428648251e-07,
+      "loss": 0.0564,
+      "step": 16250
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 9.074410163339384e-07,
+      "loss": 0.0418,
+      "step": 16260
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 8.738320898030518e-07,
+      "loss": 0.0435,
+      "step": 16270
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 8.402231632721651e-07,
+      "loss": 0.0455,
+      "step": 16280
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 8.066142367412785e-07,
+      "loss": 0.0476,
+      "step": 16290
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 7.730053102103919e-07,
+      "loss": 0.0476,
+      "step": 16300
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 7.393963836795053e-07,
+      "loss": 0.0464,
+      "step": 16310
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 7.057874571486187e-07,
+      "loss": 0.0513,
+      "step": 16320
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 6.721785306177321e-07,
+      "loss": 0.0629,
+      "step": 16330
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 6.385696040868455e-07,
+      "loss": 0.0365,
+      "step": 16340
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 6.049606775559589e-07,
+      "loss": 0.0534,
+      "step": 16350
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 5.713517510250722e-07,
+      "loss": 0.0515,
+      "step": 16360
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 5.377428244941857e-07,
+      "loss": 0.0473,
+      "step": 16370
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 5.04133897963299e-07,
+      "loss": 0.0462,
+      "step": 16380
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 4.7052497143241253e-07,
+      "loss": 0.0604,
+      "step": 16390
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 4.369160449015259e-07,
+      "loss": 0.036,
+      "step": 16400
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 4.0330711837063923e-07,
+      "loss": 0.0367,
+      "step": 16410
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.6969819183975263e-07,
+      "loss": 0.0295,
+      "step": 16420
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.3608926530886603e-07,
+      "loss": 0.0486,
+      "step": 16430
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.0248033877797943e-07,
+      "loss": 0.0431,
+      "step": 16440
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 2.6887141224709284e-07,
+      "loss": 0.0352,
+      "step": 16450
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 2.3526248571620627e-07,
+      "loss": 0.0551,
+      "step": 16460
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 2.0165355918531961e-07,
+      "loss": 0.046,
+      "step": 16470
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.6804463265443302e-07,
+      "loss": 0.0506,
+      "step": 16480
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 1.3443570612354642e-07,
+      "loss": 0.0467,
+      "step": 16490
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 1.0082677959265981e-07,
+      "loss": 0.0482,
+      "step": 16500
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 6.721785306177321e-08,
+      "loss": 0.0411,
+      "step": 16510
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.3608926530886605e-08,
+      "loss": 0.0411,
+      "step": 16520
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 0.0,
+      "loss": 0.0424,
+      "step": 16530
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.9938775510204082,
+      "eval_loss": 0.02112790010869503,
+      "eval_runtime": 89.1886,
+      "eval_samples_per_second": 131.855,
+      "eval_steps_per_second": 4.126,
+      "step": 16530
+    },
+    {
+      "epoch": 30.0,
+      "step": 16530,
+      "total_flos": 1.640377515244546e+20,
+      "train_loss": 0.13347519217154796,
+      "train_runtime": 35315.2862,
+      "train_samples_per_second": 59.941,
+      "train_steps_per_second": 0.468
+    }
+  ],
+  "max_steps": 16530,
+  "num_train_epochs": 30,
+  "total_flos": 1.640377515244546e+20,
+  "trial_name": null,
+  "trial_params": null
+}