diff --git "a/cost_to_carry_frequency_1001/checkpoint-70000/trainer_state.json" "b/cost_to_carry_frequency_1001/checkpoint-70000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/cost_to_carry_frequency_1001/checkpoint-70000/trainer_state.json"
@@ -0,0 +1,10473 @@
+{
+  "best_global_step": 65000,
+  "best_metric": 3.531397819519043,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/cost_to_carry_frequency_1001/checkpoint-40000",
+  "epoch": 20.390351899324166,
+  "eval_steps": 1000,
+  "global_step": 70000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01456536937776742,
+      "grad_norm": 1.1286877393722534,
+      "learning_rate": 0.000294,
+      "loss": 8.4423,
+      "step": 50
+    },
+    {
+      "epoch": 0.02913073875553484,
+      "grad_norm": 0.6508575081825256,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7639,
+      "step": 100
+    },
+    {
+      "epoch": 0.04369610813330226,
+      "grad_norm": 0.5259286165237427,
+      "learning_rate": 0.0005998286213931798,
+      "loss": 6.3498,
+      "step": 150
+    },
+    {
+      "epoch": 0.05826147751106968,
+      "grad_norm": 0.5010858774185181,
+      "learning_rate": 0.0005996537452637714,
+      "loss": 6.1594,
+      "step": 200
+    },
+    {
+      "epoch": 0.0728268468888371,
+      "grad_norm": 0.4704155921936035,
+      "learning_rate": 0.0005994788691343632,
+      "loss": 5.999,
+      "step": 250
+    },
+    {
+      "epoch": 0.08739221626660452,
+      "grad_norm": 0.4632050693035126,
+      "learning_rate": 0.0005993039930049548,
+      "loss": 5.8689,
+      "step": 300
+    },
+    {
+      "epoch": 0.10195758564437195,
+      "grad_norm": 0.41974759101867676,
+      "learning_rate": 0.0005991291168755465,
+      "loss": 5.7517,
+      "step": 350
+    },
+    {
+      "epoch": 0.11652295502213936,
+      "grad_norm": 0.46889957785606384,
+      "learning_rate": 0.0005989542407461382,
+      "loss": 5.6116,
+      "step": 400
+    },
+    {
+      "epoch": 0.13108832439990678,
+      "grad_norm": 0.5077112317085266,
+      "learning_rate": 0.0005987793646167297,
+      "loss": 5.5009,
+      "step": 450
+    },
+    {
+      "epoch": 0.1456536937776742,
+      "grad_norm": 0.44262439012527466,
+      "learning_rate": 0.0005986044884873214,
+      "loss": 5.4239,
+      "step": 500
+    },
+    {
+      "epoch": 0.16021906315544163,
+      "grad_norm": 0.42257460951805115,
+      "learning_rate": 0.0005984296123579131,
+      "loss": 5.3357,
+      "step": 550
+    },
+    {
+      "epoch": 0.17478443253320905,
+      "grad_norm": 0.47247999906539917,
+      "learning_rate": 0.0005982547362285047,
+      "loss": 5.2543,
+      "step": 600
+    },
+    {
+      "epoch": 0.18934980191097647,
+      "grad_norm": 0.4900086522102356,
+      "learning_rate": 0.0005980798600990964,
+      "loss": 5.2059,
+      "step": 650
+    },
+    {
+      "epoch": 0.2039151712887439,
+      "grad_norm": 0.4248013496398926,
+      "learning_rate": 0.0005979049839696881,
+      "loss": 5.1351,
+      "step": 700
+    },
+    {
+      "epoch": 0.2184805406665113,
+      "grad_norm": 0.46461302042007446,
+      "learning_rate": 0.0005977301078402798,
+      "loss": 5.0682,
+      "step": 750
+    },
+    {
+      "epoch": 0.23304591004427871,
+      "grad_norm": 0.4625150263309479,
+      "learning_rate": 0.0005975552317108715,
+      "loss": 5.0336,
+      "step": 800
+    },
+    {
+      "epoch": 0.24761127942204614,
+      "grad_norm": 0.45261090993881226,
+      "learning_rate": 0.0005973803555814631,
+      "loss": 4.9726,
+      "step": 850
+    },
+    {
+      "epoch": 0.26217664879981356,
+      "grad_norm": 0.4530528485774994,
+      "learning_rate": 0.0005972054794520547,
+      "loss": 4.929,
+      "step": 900
+    },
+    {
+      "epoch": 0.276742018177581,
+      "grad_norm": 0.42197898030281067,
+      "learning_rate": 0.0005970306033226464,
+      "loss": 4.8544,
+      "step": 950
+    },
+    {
+      "epoch": 0.2913073875553484,
+      "grad_norm": 0.4997400939464569,
+      "learning_rate": 0.0005968557271932381,
+      "loss": 4.8349,
+      "step": 1000
+    },
+    {
+      "epoch": 0.2913073875553484,
+      "eval_accuracy": 0.254550716518311,
+      "eval_loss": 4.753195285797119,
+      "eval_runtime": 182.1018,
+      "eval_samples_per_second": 91.383,
+      "eval_steps_per_second": 5.717,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30587275693311583,
+      "grad_norm": 0.43173840641975403,
+      "learning_rate": 0.0005966808510638297,
+      "loss": 4.7646,
+      "step": 1050
+    },
+    {
+      "epoch": 0.32043812631088325,
+      "grad_norm": 0.46507543325424194,
+      "learning_rate": 0.0005965059749344214,
+      "loss": 4.725,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3350034956886507,
+      "grad_norm": 0.42841699719429016,
+      "learning_rate": 0.0005963310988050131,
+      "loss": 4.703,
+      "step": 1150
+    },
+    {
+      "epoch": 0.3495688650664181,
+      "grad_norm": 0.4995660185813904,
+      "learning_rate": 0.0005961562226756047,
+      "loss": 4.6536,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3641342344441855,
+      "grad_norm": 0.5018981099128723,
+      "learning_rate": 0.0005959813465461965,
+      "loss": 4.629,
+      "step": 1250
+    },
+    {
+      "epoch": 0.37869960382195295,
+      "grad_norm": 0.4232887029647827,
+      "learning_rate": 0.000595806470416788,
+      "loss": 4.6239,
+      "step": 1300
+    },
+    {
+      "epoch": 0.39326497319972037,
+      "grad_norm": 0.5123177170753479,
+      "learning_rate": 0.0005956315942873797,
+      "loss": 4.5829,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4078303425774878,
+      "grad_norm": 0.40194171667099,
+      "learning_rate": 0.0005954567181579714,
+      "loss": 4.5675,
+      "step": 1400
+    },
+    {
+      "epoch": 0.42239571195525516,
+      "grad_norm": 0.629336953163147,
+      "learning_rate": 0.000595281842028563,
+      "loss": 4.5371,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4369610813330226,
+      "grad_norm": 0.5839784145355225,
+      "learning_rate": 0.0005951069658991547,
+      "loss": 4.5088,
+      "step": 1500
+    },
+    {
+      "epoch": 0.45152645071079,
+      "grad_norm": 0.4359549283981323,
+      "learning_rate": 0.0005949320897697464,
+      "loss": 4.5029,
+      "step": 1550
+    },
+    {
+      "epoch": 0.46609182008855743,
+      "grad_norm": 0.4207645356655121,
+      "learning_rate": 0.0005947572136403381,
+      "loss": 4.4746,
+      "step": 1600
+    },
+    {
+      "epoch": 0.48065718946632485,
+      "grad_norm": 0.4240971803665161,
+      "learning_rate": 0.0005945823375109297,
+      "loss": 4.4534,
+      "step": 1650
+    },
+    {
+      "epoch": 0.4952225588440923,
+      "grad_norm": 0.40806975960731506,
+      "learning_rate": 0.0005944074613815215,
+      "loss": 4.4362,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5097879282218597,
+      "grad_norm": 0.39968279004096985,
+      "learning_rate": 0.000594232585252113,
+      "loss": 4.419,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5243532975996271,
+      "grad_norm": 0.4166070520877838,
+      "learning_rate": 0.0005940577091227047,
+      "loss": 4.3963,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5389186669773945,
+      "grad_norm": 0.39624881744384766,
+      "learning_rate": 0.0005938828329932964,
+      "loss": 4.3801,
+      "step": 1850
+    },
+    {
+      "epoch": 0.553484036355162,
+      "grad_norm": 0.38437262177467346,
+      "learning_rate": 0.000593707956863888,
+      "loss": 4.3683,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5680494057329294,
+      "grad_norm": 0.4019283950328827,
+      "learning_rate": 0.0005935330807344797,
+      "loss": 4.3582,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5826147751106968,
+      "grad_norm": 0.4059256613254547,
+      "learning_rate": 0.0005933582046050714,
+      "loss": 4.3363,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5826147751106968,
+      "eval_accuracy": 0.2986111331607231,
+      "eval_loss": 4.290746212005615,
+      "eval_runtime": 181.2807,
+      "eval_samples_per_second": 91.797,
+      "eval_steps_per_second": 5.742,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5971801444884642,
+      "grad_norm": 0.3620131015777588,
+      "learning_rate": 0.000593183328475663,
+      "loss": 4.3338,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6117455138662317,
+      "grad_norm": 0.3989953398704529,
+      "learning_rate": 0.0005930084523462546,
+      "loss": 4.3305,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6263108832439991,
+      "grad_norm": 0.4535677134990692,
+      "learning_rate": 0.0005928335762168463,
+      "loss": 4.3148,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6408762526217665,
+      "grad_norm": 0.4322173595428467,
+      "learning_rate": 0.000592658700087438,
+      "loss": 4.2914,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6554416219995339,
+      "grad_norm": 0.3768172264099121,
+      "learning_rate": 0.0005924838239580297,
+      "loss": 4.2939,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6700069913773014,
+      "grad_norm": 0.38451075553894043,
+      "learning_rate": 0.0005923089478286214,
+      "loss": 4.2683,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6845723607550688,
+      "grad_norm": 0.3866783380508423,
+      "learning_rate": 0.000592134071699213,
+      "loss": 4.2522,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6991377301328362,
+      "grad_norm": 0.35611629486083984,
+      "learning_rate": 0.0005919591955698047,
+      "loss": 4.2531,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7137030995106036,
+      "grad_norm": 0.3990307152271271,
+      "learning_rate": 0.0005917843194403964,
+      "loss": 4.248,
+      "step": 2450
+    },
+    {
+      "epoch": 0.728268468888371,
+      "grad_norm": 0.3886300027370453,
+      "learning_rate": 0.000591609443310988,
+      "loss": 4.238,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7428338382661385,
+      "grad_norm": 0.40542346239089966,
+      "learning_rate": 0.0005914345671815796,
+      "loss": 4.2279,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7573992076439059,
+      "grad_norm": 0.3999042809009552,
+      "learning_rate": 0.0005912596910521713,
+      "loss": 4.222,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7719645770216733,
+      "grad_norm": 0.37563833594322205,
+      "learning_rate": 0.0005910848149227629,
+      "loss": 4.2156,
+      "step": 2650
+    },
+    {
+      "epoch": 0.7865299463994407,
+      "grad_norm": 0.35306981205940247,
+      "learning_rate": 0.0005909099387933547,
+      "loss": 4.1809,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8010953157772082,
+      "grad_norm": 0.37034013867378235,
+      "learning_rate": 0.0005907350626639463,
+      "loss": 4.188,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8156606851549756,
+      "grad_norm": 0.35935133695602417,
+      "learning_rate": 0.000590560186534538,
+      "loss": 4.181,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8302260545327429,
+      "grad_norm": 0.33182665705680847,
+      "learning_rate": 0.0005903853104051297,
+      "loss": 4.1834,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8447914239105103,
+      "grad_norm": 0.3539564609527588,
+      "learning_rate": 0.0005902104342757214,
+      "loss": 4.1631,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8593567932882777,
+      "grad_norm": 0.3768957555294037,
+      "learning_rate": 0.000590035558146313,
+      "loss": 4.1573,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8739221626660452,
+      "grad_norm": 0.3533536493778229,
+      "learning_rate": 0.0005898606820169046,
+      "loss": 4.1568,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8739221626660452,
+      "eval_accuracy": 0.3155079566171826,
+      "eval_loss": 4.0979695320129395,
+      "eval_runtime": 181.0102,
+      "eval_samples_per_second": 91.934,
+      "eval_steps_per_second": 5.751,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8884875320438126,
+      "grad_norm": 0.37725579738616943,
+      "learning_rate": 0.0005896858058874963,
+      "loss": 4.1293,
+      "step": 3050
+    },
+    {
+      "epoch": 0.90305290142158,
+      "grad_norm": 0.36971405148506165,
+      "learning_rate": 0.0005895109297580879,
+      "loss": 4.1306,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9176182707993474,
+      "grad_norm": 0.36352452635765076,
+      "learning_rate": 0.0005893360536286797,
+      "loss": 4.1261,
+      "step": 3150
+    },
+    {
+      "epoch": 0.9321836401771149,
+      "grad_norm": 0.3436596691608429,
+      "learning_rate": 0.0005891611774992713,
+      "loss": 4.1158,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9467490095548823,
+      "grad_norm": 0.3361954391002655,
+      "learning_rate": 0.000588986301369863,
+      "loss": 4.1148,
+      "step": 3250
+    },
+    {
+      "epoch": 0.9613143789326497,
+      "grad_norm": 0.32826173305511475,
+      "learning_rate": 0.0005888114252404547,
+      "loss": 4.1172,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9758797483104171,
+      "grad_norm": 0.3943990468978882,
+      "learning_rate": 0.0005886365491110463,
+      "loss": 4.1124,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9904451176881846,
+      "grad_norm": 0.3514852225780487,
+      "learning_rate": 0.000588461672981638,
+      "loss": 4.1085,
+      "step": 3400
+    },
+    {
+      "epoch": 1.0049522255884409,
+      "grad_norm": 0.353822261095047,
+      "learning_rate": 0.0005882867968522296,
+      "loss": 4.0743,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0195175949662083,
+      "grad_norm": 0.34162387251853943,
+      "learning_rate": 0.0005881119207228212,
+      "loss": 4.033,
+      "step": 3500
+    },
+    {
+      "epoch": 1.0340829643439757,
+      "grad_norm": 0.3392149806022644,
+      "learning_rate": 0.0005879370445934129,
+      "loss": 4.0297,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0486483337217432,
+      "grad_norm": 0.36307448148727417,
+      "learning_rate": 0.0005877621684640046,
+      "loss": 4.0216,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0632137030995106,
+      "grad_norm": 0.3409338891506195,
+      "learning_rate": 0.0005875872923345963,
+      "loss": 4.0055,
+      "step": 3650
+    },
+    {
+      "epoch": 1.077779072477278,
+      "grad_norm": 0.3846415579319,
+      "learning_rate": 0.000587412416205188,
+      "loss": 4.0206,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0923444418550454,
+      "grad_norm": 0.3451685607433319,
+      "learning_rate": 0.0005872375400757797,
+      "loss": 4.0094,
+      "step": 3750
+    },
+    {
+      "epoch": 1.1069098112328128,
+      "grad_norm": 0.3383120596408844,
+      "learning_rate": 0.0005870626639463713,
+      "loss": 3.998,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1214751806105803,
+      "grad_norm": 0.3451167941093445,
+      "learning_rate": 0.0005868877878169629,
+      "loss": 4.0044,
+      "step": 3850
+    },
+    {
+      "epoch": 1.1360405499883477,
+      "grad_norm": 0.35912787914276123,
+      "learning_rate": 0.0005867129116875546,
+      "loss": 4.0043,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1506059193661151,
+      "grad_norm": 0.35060566663742065,
+      "learning_rate": 0.0005865380355581462,
+      "loss": 3.998,
+      "step": 3950
+    },
+    {
+      "epoch": 1.1651712887438825,
+      "grad_norm": 0.33662569522857666,
+      "learning_rate": 0.0005863631594287379,
+      "loss": 3.9878,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1651712887438825,
+      "eval_accuracy": 0.3255429408255445,
+      "eval_loss": 3.9938113689422607,
+      "eval_runtime": 181.0545,
+      "eval_samples_per_second": 91.912,
+      "eval_steps_per_second": 5.75,
+      "step": 4000
+    },
+    {
+      "epoch": 1.17973665812165,
+      "grad_norm": 0.3480958342552185,
+      "learning_rate": 0.0005861882832993296,
+      "loss": 4.002,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1943020274994174,
+      "grad_norm": 0.35515499114990234,
+      "learning_rate": 0.0005860134071699212,
+      "loss": 3.9967,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2088673968771848,
+      "grad_norm": 0.3413291871547699,
+      "learning_rate": 0.000585838531040513,
+      "loss": 3.9886,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2234327662549522,
+      "grad_norm": 0.3490436375141144,
+      "learning_rate": 0.0005856636549111046,
+      "loss": 3.9804,
+      "step": 4200
+    },
+    {
+      "epoch": 1.2379981356327197,
+      "grad_norm": 0.367343932390213,
+      "learning_rate": 0.0005854887787816963,
+      "loss": 3.9875,
+      "step": 4250
+    },
+    {
+      "epoch": 1.252563505010487,
+      "grad_norm": 0.3125546872615814,
+      "learning_rate": 0.0005853139026522879,
+      "loss": 3.9789,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2671288743882545,
+      "grad_norm": 0.35544246435165405,
+      "learning_rate": 0.0005851390265228796,
+      "loss": 3.9683,
+      "step": 4350
+    },
+    {
+      "epoch": 1.281694243766022,
+      "grad_norm": 0.36066731810569763,
+      "learning_rate": 0.0005849641503934712,
+      "loss": 3.9686,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2962596131437893,
+      "grad_norm": 0.36807703971862793,
+      "learning_rate": 0.0005847892742640629,
+      "loss": 3.961,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3108249825215568,
+      "grad_norm": 0.3239951729774475,
+      "learning_rate": 0.0005846143981346546,
+      "loss": 3.9551,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3253903518993242,
+      "grad_norm": 0.334408164024353,
+      "learning_rate": 0.0005844395220052462,
+      "loss": 3.9596,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3399557212770916,
+      "grad_norm": 0.3215316832065582,
+      "learning_rate": 0.000584264645875838,
+      "loss": 3.9618,
+      "step": 4600
+    },
+    {
+      "epoch": 1.354521090654859,
+      "grad_norm": 0.33398136496543884,
+      "learning_rate": 0.0005840897697464296,
+      "loss": 3.9556,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3690864600326265,
+      "grad_norm": 0.31839656829833984,
+      "learning_rate": 0.0005839148936170212,
+      "loss": 3.9485,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3836518294103939,
+      "grad_norm": 0.338927298784256,
+      "learning_rate": 0.0005837400174876129,
+      "loss": 3.9457,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3982171987881613,
+      "grad_norm": 0.3404344916343689,
+      "learning_rate": 0.0005835651413582045,
+      "loss": 3.9378,
+      "step": 4800
+    },
+    {
+      "epoch": 1.4127825681659287,
+      "grad_norm": 0.34017398953437805,
+      "learning_rate": 0.0005833902652287962,
+      "loss": 3.9348,
+      "step": 4850
+    },
+    {
+      "epoch": 1.4273479375436962,
+      "grad_norm": 0.3449731767177582,
+      "learning_rate": 0.0005832153890993879,
+      "loss": 3.9387,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4419133069214636,
+      "grad_norm": 0.31628668308258057,
+      "learning_rate": 0.0005830405129699796,
+      "loss": 3.9351,
+      "step": 4950
+    },
+    {
+      "epoch": 1.456478676299231,
+      "grad_norm": 0.3243519961833954,
+      "learning_rate": 0.0005828656368405712,
+      "loss": 3.9343,
+      "step": 5000
+    },
+    {
+      "epoch": 1.456478676299231,
+      "eval_accuracy": 0.3317180081591796,
+      "eval_loss": 3.916283130645752,
+      "eval_runtime": 180.8146,
+      "eval_samples_per_second": 92.033,
+      "eval_steps_per_second": 5.757,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4710440456769984,
+      "grad_norm": 0.31584593653678894,
+      "learning_rate": 0.0005826907607111629,
+      "loss": 3.9289,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4856094150547658,
+      "grad_norm": 0.3198137879371643,
+      "learning_rate": 0.0005825158845817546,
+      "loss": 3.9308,
+      "step": 5100
+    },
+    {
+      "epoch": 1.500174784432533,
+      "grad_norm": 0.34546130895614624,
+      "learning_rate": 0.0005823410084523462,
+      "loss": 3.9231,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5147401538103007,
+      "grad_norm": 0.3420121967792511,
+      "learning_rate": 0.0005821661323229379,
+      "loss": 3.9264,
+      "step": 5200
+    },
+    {
+      "epoch": 1.529305523188068,
+      "grad_norm": 0.3230381906032562,
+      "learning_rate": 0.0005819912561935295,
+      "loss": 3.9056,
+      "step": 5250
+    },
+    {
+      "epoch": 1.5438708925658355,
+      "grad_norm": 0.35279926657676697,
+      "learning_rate": 0.0005818163800641212,
+      "loss": 3.9261,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5584362619436027,
+      "grad_norm": 0.31480661034584045,
+      "learning_rate": 0.0005816415039347129,
+      "loss": 3.9077,
+      "step": 5350
+    },
+    {
+      "epoch": 1.5730016313213704,
+      "grad_norm": 0.3272147476673126,
+      "learning_rate": 0.0005814666278053045,
+      "loss": 3.908,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5875670006991376,
+      "grad_norm": 0.34226250648498535,
+      "learning_rate": 0.0005812917516758962,
+      "loss": 3.9047,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6021323700769052,
+      "grad_norm": 0.3341136872768402,
+      "learning_rate": 0.0005811168755464879,
+      "loss": 3.9064,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6166977394546724,
+      "grad_norm": 0.32607197761535645,
+      "learning_rate": 0.0005809419994170794,
+      "loss": 3.9164,
+      "step": 5550
+    },
+    {
+      "epoch": 1.63126310883244,
+      "grad_norm": 0.31550896167755127,
+      "learning_rate": 0.0005807671232876712,
+      "loss": 3.8887,
+      "step": 5600
+    },
+    {
+      "epoch": 1.6458284782102073,
+      "grad_norm": 0.33480650186538696,
+      "learning_rate": 0.0005805922471582628,
+      "loss": 3.9082,
+      "step": 5650
+    },
+    {
+      "epoch": 1.660393847587975,
+      "grad_norm": 0.32147085666656494,
+      "learning_rate": 0.0005804173710288545,
+      "loss": 3.8912,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6749592169657421,
+      "grad_norm": 0.30887845158576965,
+      "learning_rate": 0.0005802424948994462,
+      "loss": 3.9012,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6895245863435098,
+      "grad_norm": 0.31019526720046997,
+      "learning_rate": 0.0005800676187700379,
+      "loss": 3.8959,
+      "step": 5800
+    },
+    {
+      "epoch": 1.704089955721277,
+      "grad_norm": 0.34193840622901917,
+      "learning_rate": 0.0005798927426406295,
+      "loss": 3.8893,
+      "step": 5850
+    },
+    {
+      "epoch": 1.7186553250990446,
+      "grad_norm": 0.32311904430389404,
+      "learning_rate": 0.0005797178665112212,
+      "loss": 3.8786,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7332206944768118,
+      "grad_norm": 0.33148226141929626,
+      "learning_rate": 0.0005795429903818129,
+      "loss": 3.8832,
+      "step": 5950
+    },
+    {
+      "epoch": 1.7477860638545795,
+      "grad_norm": 0.32679638266563416,
+      "learning_rate": 0.0005793681142524044,
+      "loss": 3.8895,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7477860638545795,
+      "eval_accuracy": 0.3367914180793412,
+      "eval_loss": 3.8570826053619385,
+      "eval_runtime": 181.05,
+      "eval_samples_per_second": 91.914,
+      "eval_steps_per_second": 5.75,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7623514332323467,
+      "grad_norm": 0.31540998816490173,
+      "learning_rate": 0.0005791932381229961,
+      "loss": 3.8822,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7769168026101143,
+      "grad_norm": 0.3205857574939728,
+      "learning_rate": 0.0005790183619935878,
+      "loss": 3.8743,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7914821719878815,
+      "grad_norm": 0.3156723082065582,
+      "learning_rate": 0.0005788434858641795,
+      "loss": 3.8689,
+      "step": 6150
+    },
+    {
+      "epoch": 1.8060475413656492,
+      "grad_norm": 0.33912575244903564,
+      "learning_rate": 0.0005786686097347712,
+      "loss": 3.8712,
+      "step": 6200
+    },
+    {
+      "epoch": 1.8206129107434164,
+      "grad_norm": 0.32791027426719666,
+      "learning_rate": 0.0005784937336053628,
+      "loss": 3.8814,
+      "step": 6250
+    },
+    {
+      "epoch": 1.835178280121184,
+      "grad_norm": 0.3195091485977173,
+      "learning_rate": 0.0005783188574759545,
+      "loss": 3.8667,
+      "step": 6300
+    },
+    {
+      "epoch": 1.8497436494989512,
+      "grad_norm": 0.3110474646091461,
+      "learning_rate": 0.0005781439813465462,
+      "loss": 3.8601,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8643090188767188,
+      "grad_norm": 0.3255619406700134,
+      "learning_rate": 0.0005779691052171379,
+      "loss": 3.866,
+      "step": 6400
+    },
+    {
+      "epoch": 1.878874388254486,
+      "grad_norm": 0.34154996275901794,
+      "learning_rate": 0.0005777942290877294,
+      "loss": 3.8775,
+      "step": 6450
+    },
+    {
+      "epoch": 1.8934397576322537,
+      "grad_norm": 0.347554475069046,
+      "learning_rate": 0.0005776193529583211,
+      "loss": 3.8739,
+      "step": 6500
+    },
+    {
+      "epoch": 1.908005127010021,
+      "grad_norm": 0.328535258769989,
+      "learning_rate": 0.0005774444768289128,
+      "loss": 3.8648,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9225704963877885,
+      "grad_norm": 0.3176211416721344,
+      "learning_rate": 0.0005772696006995045,
+      "loss": 3.8481,
+      "step": 6600
+    },
+    {
+      "epoch": 1.9371358657655557,
+      "grad_norm": 0.3081546723842621,
+      "learning_rate": 0.0005770947245700962,
+      "loss": 3.8432,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9517012351433234,
+      "grad_norm": 0.31069493293762207,
+      "learning_rate": 0.0005769198484406878,
+      "loss": 3.8413,
+      "step": 6700
+    },
+    {
+      "epoch": 1.9662666045210906,
+      "grad_norm": 0.31990841031074524,
+      "learning_rate": 0.0005767449723112795,
+      "loss": 3.861,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9808319738988582,
+      "grad_norm": 0.3148524761199951,
+      "learning_rate": 0.0005765700961818712,
+      "loss": 3.8455,
+      "step": 6800
+    },
+    {
+      "epoch": 1.9953973432766254,
+      "grad_norm": 0.3133419156074524,
+      "learning_rate": 0.0005763952200524627,
+      "loss": 3.8619,
+      "step": 6850
+    },
+    {
+      "epoch": 2.0099044511768818,
+      "grad_norm": 0.30236005783081055,
+      "learning_rate": 0.0005762203439230544,
+      "loss": 3.7771,
+      "step": 6900
+    },
+    {
+      "epoch": 2.0244698205546494,
+      "grad_norm": 0.3117513358592987,
+      "learning_rate": 0.0005760454677936461,
+      "loss": 3.74,
+      "step": 6950
+    },
+    {
+      "epoch": 2.0390351899324166,
+      "grad_norm": 0.3317764103412628,
+      "learning_rate": 0.0005758705916642378,
+      "loss": 3.7541,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0390351899324166,
+      "eval_accuracy": 0.34124602768890316,
+      "eval_loss": 3.8186287879943848,
+      "eval_runtime": 181.1173,
+      "eval_samples_per_second": 91.88,
+      "eval_steps_per_second": 5.748,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0536005593101843,
+      "grad_norm": 0.32364246249198914,
+      "learning_rate": 0.0005756957155348294,
+      "loss": 3.744,
+      "step": 7050
+    },
+    {
+      "epoch": 2.0681659286879515,
+      "grad_norm": 0.3094733953475952,
+      "learning_rate": 0.0005755208394054211,
+      "loss": 3.7481,
+      "step": 7100
+    },
+    {
+      "epoch": 2.082731298065719,
+      "grad_norm": 0.31528735160827637,
+      "learning_rate": 0.0005753459632760128,
+      "loss": 3.7572,
+      "step": 7150
+    },
+    {
+      "epoch": 2.0972966674434863,
+      "grad_norm": 0.3072337210178375,
+      "learning_rate": 0.0005751710871466045,
+      "loss": 3.7534,
+      "step": 7200
+    },
+    {
+      "epoch": 2.111862036821254,
+      "grad_norm": 0.311673104763031,
+      "learning_rate": 0.0005749962110171962,
+      "loss": 3.7452,
+      "step": 7250
+    },
+    {
+      "epoch": 2.126427406199021,
+      "grad_norm": 0.3028483986854553,
+      "learning_rate": 0.0005748213348877877,
+      "loss": 3.7552,
+      "step": 7300
+    },
+    {
+      "epoch": 2.140992775576789,
+      "grad_norm": 0.31722408533096313,
+      "learning_rate": 0.0005746464587583794,
+      "loss": 3.7551,
+      "step": 7350
+    },
+    {
+      "epoch": 2.155558144954556,
+      "grad_norm": 0.305298388004303,
+      "learning_rate": 0.0005744715826289711,
+      "loss": 3.7554,
+      "step": 7400
+    },
+    {
+      "epoch": 2.1701235143323236,
+      "grad_norm": 0.3021174371242523,
+      "learning_rate": 0.0005742967064995627,
+      "loss": 3.7625,
+      "step": 7450
+    },
+    {
+      "epoch": 2.184688883710091,
+      "grad_norm": 0.3206949830055237,
+      "learning_rate": 0.0005741218303701544,
+      "loss": 3.7539,
+      "step": 7500
+    },
+    {
+      "epoch": 2.1992542530878585,
+      "grad_norm": 0.31087061762809753,
+      "learning_rate": 0.0005739469542407461,
+      "loss": 3.7532,
+      "step": 7550
+    },
+    {
+      "epoch": 2.2138196224656257,
+      "grad_norm": 0.34177103638648987,
+      "learning_rate": 0.0005737720781113378,
+      "loss": 3.7596,
+      "step": 7600
+    },
+    {
+      "epoch": 2.2283849918433933,
+      "grad_norm": 0.3387856185436249,
+      "learning_rate": 0.0005735972019819295,
+      "loss": 3.7607,
+      "step": 7650
+    },
+    {
+      "epoch": 2.2429503612211605,
+      "grad_norm": 0.32056301832199097,
+      "learning_rate": 0.000573422325852521,
+      "loss": 3.7518,
+      "step": 7700
+    },
+    {
+      "epoch": 2.257515730598928,
+      "grad_norm": 0.3297360837459564,
+      "learning_rate": 0.0005732474497231127,
+      "loss": 3.7713,
+      "step": 7750
+    },
+    {
+      "epoch": 2.2720810999766954,
+      "grad_norm": 0.319235235452652,
+      "learning_rate": 0.0005730725735937044,
+      "loss": 3.7558,
+      "step": 7800
+    },
+    {
+      "epoch": 2.286646469354463,
+      "grad_norm": 0.3082825839519501,
+      "learning_rate": 0.0005728976974642961,
+      "loss": 3.7732,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3012118387322302,
+      "grad_norm": 0.3217436969280243,
+      "learning_rate": 0.0005727228213348877,
+      "loss": 3.7594,
+      "step": 7900
+    },
+    {
+      "epoch": 2.3157772081099974,
+      "grad_norm": 0.31779220700263977,
+      "learning_rate": 0.0005725479452054794,
+      "loss": 3.7608,
+      "step": 7950
+    },
+    {
+      "epoch": 2.330342577487765,
+      "grad_norm": 0.3270404040813446,
+      "learning_rate": 0.0005723730690760711,
+      "loss": 3.7621,
+      "step": 8000
+    },
+    {
+      "epoch": 2.330342577487765,
+      "eval_accuracy": 0.3436997085100095,
+      "eval_loss": 3.786313772201538,
+      "eval_runtime": 181.0475,
+      "eval_samples_per_second": 91.915,
+      "eval_steps_per_second": 5.75,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3449079468655327,
+      "grad_norm": 0.3205696642398834,
+      "learning_rate": 0.0005721981929466627,
+      "loss": 3.7601,
+      "step": 8050
+    },
+    {
+      "epoch": 2.3594733162433,
+      "grad_norm": 0.3155924379825592,
+      "learning_rate": 0.0005720233168172545,
+      "loss": 3.7452,
+      "step": 8100
+    },
+    {
+      "epoch": 2.374038685621067,
+      "grad_norm": 0.316879540681839,
+      "learning_rate": 0.000571848440687846,
+      "loss": 3.7518,
+      "step": 8150
+    },
+    {
+      "epoch": 2.3886040549988348,
+      "grad_norm": 0.31144627928733826,
+      "learning_rate": 0.0005716735645584377,
+      "loss": 3.7424,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4031694243766024,
+      "grad_norm": 0.3123636245727539,
+      "learning_rate": 0.0005714986884290294,
+      "loss": 3.7572,
+      "step": 8250
+    },
+    {
+      "epoch": 2.4177347937543696,
+      "grad_norm": 0.29581794142723083,
+      "learning_rate": 0.000571323812299621,
+      "loss": 3.7512,
+      "step": 8300
+    },
+    {
+      "epoch": 2.432300163132137,
+      "grad_norm": 0.3172610104084015,
+      "learning_rate": 0.0005711489361702127,
+      "loss": 3.751,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4468655325099045,
+      "grad_norm": 0.31691786646842957,
+      "learning_rate": 0.0005709740600408044,
+      "loss": 3.7504,
+      "step": 8400
+    },
+    {
+      "epoch": 2.461430901887672,
+      "grad_norm": 0.3049279451370239,
+      "learning_rate": 0.0005707991839113961,
+      "loss": 3.7418,
+      "step": 8450
+    },
+    {
+      "epoch": 2.4759962712654393,
+      "grad_norm": 0.2990638315677643,
+      "learning_rate": 0.0005706243077819877,
+      "loss": 3.7492,
+      "step": 8500
+    },
+    {
+      "epoch": 2.4905616406432065,
+      "grad_norm": 0.31371548771858215,
+      "learning_rate": 0.0005704494316525793,
+      "loss": 3.7476,
+      "step": 8550
+    },
+    {
+      "epoch": 2.505127010020974,
+      "grad_norm": 0.3349527418613434,
+      "learning_rate": 0.000570274555523171,
+      "loss": 3.7443,
+      "step": 8600
+    },
+    {
+      "epoch": 2.519692379398742,
+      "grad_norm": 0.3164826035499573,
+      "learning_rate": 0.0005700996793937627,
+      "loss": 3.7447,
+      "step": 8650
+    },
+    {
+      "epoch": 2.534257748776509,
+      "grad_norm": 0.30736449360847473,
+      "learning_rate": 0.0005699248032643544,
+      "loss": 3.7428,
+      "step": 8700
+    },
+    {
+      "epoch": 2.548823118154276,
+      "grad_norm": 0.3016897737979889,
+      "learning_rate": 0.000569749927134946,
+      "loss": 3.7504,
+      "step": 8750
+    },
+    {
+      "epoch": 2.563388487532044,
+      "grad_norm": 0.31572335958480835,
+      "learning_rate": 0.0005695750510055377,
+      "loss": 3.7594,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5779538569098115,
+      "grad_norm": 0.3104640543460846,
+      "learning_rate": 0.0005694001748761294,
+      "loss": 3.7405,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5925192262875787,
+      "grad_norm": 0.3017583191394806,
+      "learning_rate": 0.000569225298746721,
+      "loss": 3.7371,
+      "step": 8900
+    },
+    {
+      "epoch": 2.607084595665346,
+      "grad_norm": 0.3301398456096649,
+      "learning_rate": 0.0005690504226173127,
+      "loss": 3.7445,
+      "step": 8950
+    },
+    {
+      "epoch": 2.6216499650431135,
+      "grad_norm": 0.2939837872982025,
+      "learning_rate": 0.0005688755464879043,
+      "loss": 3.7407,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6216499650431135,
+      "eval_accuracy": 0.3467777167444518,
+      "eval_loss": 3.754408836364746,
+      "eval_runtime": 181.0383,
+      "eval_samples_per_second": 91.92,
+      "eval_steps_per_second": 5.75,
+      "step": 9000
+    },
+    {
+      "epoch": 2.636215334420881,
+      "grad_norm": 0.32832562923431396,
+      "learning_rate": 0.000568700670358496,
+      "loss": 3.737,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6507807037986484,
+      "grad_norm": 0.3066832423210144,
+      "learning_rate": 0.0005685257942290877,
+      "loss": 3.7403,
+      "step": 9100
+    },
+    {
+      "epoch": 2.6653460731764156,
+      "grad_norm": 0.3145512044429779,
+      "learning_rate": 0.0005683509180996793,
+      "loss": 3.7382,
+      "step": 9150
+    },
+    {
+      "epoch": 2.6799114425541832,
+      "grad_norm": 0.31166455149650574,
+      "learning_rate": 0.000568176041970271,
+      "loss": 3.7282,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6944768119319504,
+      "grad_norm": 0.34423258900642395,
+      "learning_rate": 0.0005680011658408627,
+      "loss": 3.7228,
+      "step": 9250
+    },
+    {
+      "epoch": 2.709042181309718,
+      "grad_norm": 0.3051367700099945,
+      "learning_rate": 0.0005678262897114544,
+      "loss": 3.745,
+      "step": 9300
+    },
+    {
+      "epoch": 2.7236075506874853,
+      "grad_norm": 0.308801531791687,
+      "learning_rate": 0.000567651413582046,
+      "loss": 3.7226,
+      "step": 9350
+    },
+    {
+      "epoch": 2.738172920065253,
+      "grad_norm": 0.3089958727359772,
+      "learning_rate": 0.0005674765374526377,
+      "loss": 3.7308,
+      "step": 9400
+    },
+    {
+      "epoch": 2.75273828944302,
+      "grad_norm": 0.3324858844280243,
+      "learning_rate": 0.0005673016613232293,
+      "loss": 3.7492,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7673036588207878,
+      "grad_norm": 0.30547377467155457,
+      "learning_rate": 0.0005671267851938209,
+      "loss": 3.7329,
+      "step": 9500
+    },
+    {
+      "epoch": 2.781869028198555,
+      "grad_norm": 0.310508131980896,
+      "learning_rate": 0.0005669519090644127,
+      "loss": 3.7208,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7964343975763226,
+      "grad_norm": 0.3160189390182495,
+      "learning_rate": 0.0005667770329350043,
+      "loss": 3.724,
+      "step": 9600
+    },
+    {
+      "epoch": 2.81099976695409,
+      "grad_norm": 0.30078431963920593,
+      "learning_rate": 0.000566602156805596,
+      "loss": 3.7418,
+      "step": 9650
+    },
+    {
+      "epoch": 2.8255651363318575,
+      "grad_norm": 0.32854485511779785,
+      "learning_rate": 0.0005664272806761877,
+      "loss": 3.7191,
+      "step": 9700
+    },
+    {
+      "epoch": 2.8401305057096247,
+      "grad_norm": 0.29520857334136963,
+      "learning_rate": 0.0005662524045467793,
+      "loss": 3.729,
+      "step": 9750
+    },
+    {
+      "epoch": 2.8546958750873923,
+      "grad_norm": 0.29994896054267883,
+      "learning_rate": 0.000566077528417371,
+      "loss": 3.7392,
+      "step": 9800
+    },
+    {
+      "epoch": 2.8692612444651595,
+      "grad_norm": 0.3165612518787384,
+      "learning_rate": 0.0005659026522879626,
+      "loss": 3.7333,
+      "step": 9850
+    },
+    {
+      "epoch": 2.883826613842927,
+      "grad_norm": 0.30563628673553467,
+      "learning_rate": 0.0005657277761585543,
+      "loss": 3.7296,
+      "step": 9900
+    },
+    {
+      "epoch": 2.8983919832206944,
+      "grad_norm": 0.30479544401168823,
+      "learning_rate": 0.0005655529000291459,
+      "loss": 3.7187,
+      "step": 9950
+    },
+    {
+      "epoch": 2.912957352598462,
+      "grad_norm": 0.3024917244911194,
+      "learning_rate": 0.0005653780238997376,
+      "loss": 3.7164,
+      "step": 10000
+    },
+    {
+      "epoch": 2.912957352598462,
+      "eval_accuracy": 0.34957125558487273,
+      "eval_loss": 3.7293238639831543,
+      "eval_runtime": 181.0388,
+      "eval_samples_per_second": 91.92,
+      "eval_steps_per_second": 5.75,
+      "step": 10000
+    },
+    {
+      "epoch": 2.927522721976229,
+      "grad_norm": 0.30794304609298706,
+      "learning_rate": 0.0005652031477703293,
+      "loss": 3.7329,
+      "step": 10050
+    },
+    {
+      "epoch": 2.942088091353997,
+      "grad_norm": 0.30510345101356506,
+      "learning_rate": 0.000565028271640921,
+      "loss": 3.7128,
+      "step": 10100
+    },
+    {
+      "epoch": 2.956653460731764,
+      "grad_norm": 0.2942156493663788,
+      "learning_rate": 0.0005648533955115127,
+      "loss": 3.7253,
+      "step": 10150
+    },
+    {
+      "epoch": 2.9712188301095317,
+      "grad_norm": 0.3010208308696747,
+      "learning_rate": 0.0005646785193821043,
+      "loss": 3.7275,
+      "step": 10200
+    },
+    {
+      "epoch": 2.985784199487299,
+      "grad_norm": 0.29207906126976013,
+      "learning_rate": 0.000564503643252696,
+      "loss": 3.7014,
+      "step": 10250
+    },
+    {
+      "epoch": 3.0002913073875552,
+      "grad_norm": 0.3005887269973755,
+      "learning_rate": 0.0005643287671232876,
+      "loss": 3.7275,
+      "step": 10300
+    },
+    {
+      "epoch": 3.014856676765323,
+      "grad_norm": 0.3155250549316406,
+      "learning_rate": 0.0005641538909938792,
+      "loss": 3.6032,
+      "step": 10350
+    },
+    {
+      "epoch": 3.02942204614309,
+      "grad_norm": 0.32792675495147705,
+      "learning_rate": 0.0005639790148644709,
+      "loss": 3.6166,
+      "step": 10400
+    },
+    {
+      "epoch": 3.0439874155208577,
+      "grad_norm": 0.32439231872558594,
+      "learning_rate": 0.0005638041387350626,
+      "loss": 3.6214,
+      "step": 10450
+    },
+    {
+      "epoch": 3.058552784898625,
+      "grad_norm": 0.3086802661418915,
+      "learning_rate": 0.0005636292626056543,
+      "loss": 3.6229,
+      "step": 10500
+    },
+    {
+      "epoch": 3.0731181542763926,
+      "grad_norm": 0.30797359347343445,
+      "learning_rate": 0.000563454386476246,
+      "loss": 3.6147,
+      "step": 10550
+    },
+    {
+      "epoch": 3.0876835236541598,
+      "grad_norm": 0.3180497884750366,
+      "learning_rate": 0.0005632795103468376,
+      "loss": 3.6364,
+      "step": 10600
+    },
+    {
+      "epoch": 3.1022488930319274,
+      "grad_norm": 0.31789934635162354,
+      "learning_rate": 0.0005631046342174293,
+      "loss": 3.6297,
+      "step": 10650
+    },
+    {
+      "epoch": 3.1168142624096946,
+      "grad_norm": 0.29927411675453186,
+      "learning_rate": 0.000562929758088021,
+      "loss": 3.6296,
+      "step": 10700
+    },
+    {
+      "epoch": 3.1313796317874623,
+      "grad_norm": 0.31321072578430176,
+      "learning_rate": 0.0005627548819586126,
+      "loss": 3.6228,
+      "step": 10750
+    },
+    {
+      "epoch": 3.1459450011652295,
+      "grad_norm": 0.31705328822135925,
+      "learning_rate": 0.0005625800058292042,
+      "loss": 3.6362,
+      "step": 10800
+    },
+    {
+      "epoch": 3.160510370542997,
+      "grad_norm": 0.31445786356925964,
+      "learning_rate": 0.0005624051296997959,
+      "loss": 3.6349,
+      "step": 10850
+    },
+    {
+      "epoch": 3.1750757399207643,
+      "grad_norm": 0.31521472334861755,
+      "learning_rate": 0.0005622302535703876,
+      "loss": 3.6345,
+      "step": 10900
+    },
+    {
+      "epoch": 3.189641109298532,
+      "grad_norm": 0.3396650552749634,
+      "learning_rate": 0.0005620553774409792,
+      "loss": 3.6506,
+      "step": 10950
+    },
+    {
+      "epoch": 3.204206478676299,
+      "grad_norm": 0.32105159759521484,
+      "learning_rate": 0.000561880501311571,
+      "loss": 3.6345,
+      "step": 11000
+    },
+    {
+      "epoch": 3.204206478676299,
+      "eval_accuracy": 0.3514612895248115,
+      "eval_loss": 3.7179360389709473,
+      "eval_runtime": 181.4428,
+      "eval_samples_per_second": 91.715,
+      "eval_steps_per_second": 5.737,
+      "step": 11000
+    },
+    {
+      "epoch": 3.218771848054067,
+      "grad_norm": 0.3084251582622528,
+      "learning_rate": 0.0005617056251821626,
+      "loss": 3.6305,
+      "step": 11050
+    },
+    {
+      "epoch": 3.233337217431834,
+      "grad_norm": 0.3207090198993683,
+      "learning_rate": 0.0005615307490527543,
+      "loss": 3.6489,
+      "step": 11100
+    },
+    {
+      "epoch": 3.2479025868096016,
+      "grad_norm": 0.316230833530426,
+      "learning_rate": 0.000561355872923346,
+      "loss": 3.6243,
+      "step": 11150
+    },
+    {
+      "epoch": 3.262467956187369,
+      "grad_norm": 0.3016257584095001,
+      "learning_rate": 0.0005611809967939375,
+      "loss": 3.6362,
+      "step": 11200
+    },
+    {
+      "epoch": 3.2770333255651365,
+      "grad_norm": 0.3337656259536743,
+      "learning_rate": 0.0005610061206645292,
+      "loss": 3.6496,
+      "step": 11250
+    },
+    {
+      "epoch": 3.2915986949429037,
+      "grad_norm": 0.3145868182182312,
+      "learning_rate": 0.0005608312445351209,
+      "loss": 3.6391,
+      "step": 11300
+    },
+    {
+      "epoch": 3.3061640643206713,
+      "grad_norm": 0.3270737826824188,
+      "learning_rate": 0.0005606563684057126,
+      "loss": 3.6309,
+      "step": 11350
+    },
+    {
+      "epoch": 3.3207294336984385,
+      "grad_norm": 0.31686511635780334,
+      "learning_rate": 0.0005604814922763042,
+      "loss": 3.638,
+      "step": 11400
+    },
+    {
+      "epoch": 3.335294803076206,
+      "grad_norm": 0.30375340580940247,
+      "learning_rate": 0.0005603066161468959,
+      "loss": 3.6385,
+      "step": 11450
+    },
+    {
+      "epoch": 3.3498601724539734,
+      "grad_norm": 0.3264949321746826,
+      "learning_rate": 0.0005601317400174876,
+      "loss": 3.6382,
+      "step": 11500
+    },
+    {
+      "epoch": 3.364425541831741,
+      "grad_norm": 0.32871800661087036,
+      "learning_rate": 0.0005599568638880793,
+      "loss": 3.6486,
+      "step": 11550
+    },
+    {
+      "epoch": 3.3789909112095082,
+      "grad_norm": 0.3065817654132843,
+      "learning_rate": 0.0005597819877586709,
+      "loss": 3.6278,
+      "step": 11600
+    },
+    {
+      "epoch": 3.393556280587276,
+      "grad_norm": 0.3440123200416565,
+      "learning_rate": 0.0005596071116292625,
+      "loss": 3.651,
+      "step": 11650
+    },
+    {
+      "epoch": 3.408121649965043,
+      "grad_norm": 0.3192084729671478,
+      "learning_rate": 0.0005594322354998542,
+      "loss": 3.6438,
+      "step": 11700
+    },
+    {
+      "epoch": 3.4226870193428107,
+      "grad_norm": 0.31226348876953125,
+      "learning_rate": 0.0005592573593704459,
+      "loss": 3.648,
+      "step": 11750
+    },
+    {
+      "epoch": 3.437252388720578,
+      "grad_norm": 0.2955343723297119,
+      "learning_rate": 0.0005590824832410375,
+      "loss": 3.6414,
+      "step": 11800
+    },
+    {
+      "epoch": 3.4518177580983456,
+      "grad_norm": 0.3033134341239929,
+      "learning_rate": 0.0005589076071116292,
+      "loss": 3.6433,
+      "step": 11850
+    },
+    {
+      "epoch": 3.4663831274761128,
+      "grad_norm": 0.3129415214061737,
+      "learning_rate": 0.0005587327309822209,
+      "loss": 3.6434,
+      "step": 11900
+    },
+    {
+      "epoch": 3.4809484968538804,
+      "grad_norm": 0.31332719326019287,
+      "learning_rate": 0.0005585578548528126,
+      "loss": 3.642,
+      "step": 11950
+    },
+    {
+      "epoch": 3.4955138662316476,
+      "grad_norm": 0.3182447850704193,
+      "learning_rate": 0.0005583829787234043,
+      "loss": 3.6454,
+      "step": 12000
+    },
+    {
+      "epoch": 3.4955138662316476,
+      "eval_accuracy": 0.35339601067836246,
+      "eval_loss": 3.6975326538085938,
+      "eval_runtime": 181.039,
+      "eval_samples_per_second": 91.919,
+      "eval_steps_per_second": 5.75,
+      "step": 12000
+    },
+    {
+      "epoch": 3.510079235609415,
+      "grad_norm": 0.30527937412261963,
+      "learning_rate": 0.0005582081025939958,
+      "loss": 3.6415,
+      "step": 12050
+    },
+    {
+      "epoch": 3.5246446049871825,
+      "grad_norm": 0.3121168315410614,
+      "learning_rate": 0.0005580332264645875,
+      "loss": 3.6291,
+      "step": 12100
+    },
+    {
+      "epoch": 3.53920997436495,
+      "grad_norm": 0.32664182782173157,
+      "learning_rate": 0.0005578583503351792,
+      "loss": 3.6479,
+      "step": 12150
+    },
+    {
+      "epoch": 3.5537753437427173,
+      "grad_norm": 0.34277892112731934,
+      "learning_rate": 0.0005576834742057709,
+      "loss": 3.6362,
+      "step": 12200
+    },
+    {
+      "epoch": 3.5683407131204845,
+      "grad_norm": 0.3118998408317566,
+      "learning_rate": 0.0005575085980763625,
+      "loss": 3.6316,
+      "step": 12250
+    },
+    {
+      "epoch": 3.582906082498252,
+      "grad_norm": 0.28787028789520264,
+      "learning_rate": 0.0005573337219469542,
+      "loss": 3.6378,
+      "step": 12300
+    },
+    {
+      "epoch": 3.59747145187602,
+      "grad_norm": 0.3117337226867676,
+      "learning_rate": 0.0005571588458175459,
+      "loss": 3.6465,
+      "step": 12350
+    },
+    {
+      "epoch": 3.612036821253787,
+      "grad_norm": 0.30298158526420593,
+      "learning_rate": 0.0005569839696881374,
+      "loss": 3.6424,
+      "step": 12400
+    },
+    {
+      "epoch": 3.626602190631554,
+      "grad_norm": 0.2996442914009094,
+      "learning_rate": 0.0005568090935587292,
+      "loss": 3.6427,
+      "step": 12450
+    },
+    {
+      "epoch": 3.641167560009322,
+      "grad_norm": 0.3013766407966614,
+      "learning_rate": 0.0005566342174293208,
+      "loss": 3.6436,
+      "step": 12500
+    },
+    {
+      "epoch": 3.6557329293870895,
+      "grad_norm": 0.32076093554496765,
+      "learning_rate": 0.0005564593412999125,
+      "loss": 3.6448,
+      "step": 12550
+    },
+    {
+      "epoch": 3.6702982987648567,
+      "grad_norm": 0.30450958013534546,
+      "learning_rate": 0.0005562844651705042,
+      "loss": 3.6553,
+      "step": 12600
+    },
+    {
+      "epoch": 3.684863668142624,
+      "grad_norm": 0.3103180527687073,
+      "learning_rate": 0.0005561095890410958,
+      "loss": 3.6385,
+      "step": 12650
+    },
+    {
+      "epoch": 3.6994290375203915,
+      "grad_norm": 0.3013014495372772,
+      "learning_rate": 0.0005559347129116875,
+      "loss": 3.6512,
+      "step": 12700
+    },
+    {
+      "epoch": 3.713994406898159,
+      "grad_norm": 0.3091047406196594,
+      "learning_rate": 0.0005557598367822792,
+      "loss": 3.6415,
+      "step": 12750
+    },
+    {
+      "epoch": 3.7285597762759264,
+      "grad_norm": 0.3013926148414612,
+      "learning_rate": 0.0005555849606528709,
+      "loss": 3.6494,
+      "step": 12800
+    },
+    {
+      "epoch": 3.7431251456536936,
+      "grad_norm": 0.30557090044021606,
+      "learning_rate": 0.0005554100845234624,
+      "loss": 3.6497,
+      "step": 12850
+    },
+    {
+      "epoch": 3.7576905150314612,
+      "grad_norm": 0.33003371953964233,
+      "learning_rate": 0.0005552352083940541,
+      "loss": 3.6412,
+      "step": 12900
+    },
+    {
+      "epoch": 3.772255884409229,
+      "grad_norm": 0.3131480813026428,
+      "learning_rate": 0.0005550603322646458,
+      "loss": 3.6409,
+      "step": 12950
+    },
+    {
+      "epoch": 3.786821253786996,
+      "grad_norm": 0.30647027492523193,
+      "learning_rate": 0.0005548854561352375,
+      "loss": 3.6355,
+      "step": 13000
+    },
+    {
+      "epoch": 3.786821253786996,
+      "eval_accuracy": 0.3546811208635075,
+      "eval_loss": 3.6775307655334473,
+      "eval_runtime": 180.9195,
+      "eval_samples_per_second": 91.98,
+      "eval_steps_per_second": 5.754,
+      "step": 13000
+    },
+    {
+      "epoch": 3.8013866231647633,
+      "grad_norm": 0.30315953493118286,
+      "learning_rate": 0.0005547105800058292,
+      "loss": 3.6419,
+      "step": 13050
+    },
+    {
+      "epoch": 3.815951992542531,
+      "grad_norm": 0.30151739716529846,
+      "learning_rate": 0.0005545357038764208,
+      "loss": 3.6455,
+      "step": 13100
+    },
+    {
+      "epoch": 3.8305173619202986,
+      "grad_norm": 0.2949901521205902,
+      "learning_rate": 0.0005543608277470125,
+      "loss": 3.6441,
+      "step": 13150
+    },
+    {
+      "epoch": 3.8450827312980658,
+      "grad_norm": 0.28778183460235596,
+      "learning_rate": 0.0005541859516176042,
+      "loss": 3.6412,
+      "step": 13200
+    },
+    {
+      "epoch": 3.859648100675833,
+      "grad_norm": 0.30516332387924194,
+      "learning_rate": 0.0005540110754881958,
+      "loss": 3.6375,
+      "step": 13250
+    },
+    {
+      "epoch": 3.8742134700536006,
+      "grad_norm": 0.3119845986366272,
+      "learning_rate": 0.0005538361993587874,
+      "loss": 3.6438,
+      "step": 13300
+    },
+    {
+      "epoch": 3.888778839431368,
+      "grad_norm": 0.2999977171421051,
+      "learning_rate": 0.0005536613232293791,
+      "loss": 3.6498,
+      "step": 13350
+    },
+    {
+      "epoch": 3.9033442088091355,
+      "grad_norm": 0.32405146956443787,
+      "learning_rate": 0.0005534864470999708,
+      "loss": 3.6397,
+      "step": 13400
+    },
+    {
+      "epoch": 3.9179095781869027,
+      "grad_norm": 0.31608811020851135,
+      "learning_rate": 0.0005533115709705625,
+      "loss": 3.6373,
+      "step": 13450
+    },
+    {
+      "epoch": 3.9324749475646703,
+      "grad_norm": 0.2984311282634735,
+      "learning_rate": 0.0005531366948411541,
+      "loss": 3.6524,
+      "step": 13500
+    },
+    {
+      "epoch": 3.9470403169424375,
+      "grad_norm": 0.3015180230140686,
+      "learning_rate": 0.0005529618187117458,
+      "loss": 3.6248,
+      "step": 13550
+    },
+    {
+      "epoch": 3.961605686320205,
+      "grad_norm": 0.3151557445526123,
+      "learning_rate": 0.0005527869425823375,
+      "loss": 3.6338,
+      "step": 13600
+    },
+    {
+      "epoch": 3.9761710556979724,
+      "grad_norm": 0.30130767822265625,
+      "learning_rate": 0.0005526120664529292,
+      "loss": 3.6416,
+      "step": 13650
+    },
+    {
+      "epoch": 3.99073642507574,
+      "grad_norm": 0.2960662841796875,
+      "learning_rate": 0.0005524371903235207,
+      "loss": 3.6358,
+      "step": 13700
+    },
+    {
+      "epoch": 4.005243532975996,
+      "grad_norm": 0.31960538029670715,
+      "learning_rate": 0.0005522623141941124,
+      "loss": 3.6021,
+      "step": 13750
+    },
+    {
+      "epoch": 4.0198089023537635,
+      "grad_norm": 0.31510815024375916,
+      "learning_rate": 0.0005520874380647041,
+      "loss": 3.5299,
+      "step": 13800
+    },
+    {
+      "epoch": 4.034374271731531,
+      "grad_norm": 0.31091028451919556,
+      "learning_rate": 0.0005519125619352957,
+      "loss": 3.5372,
+      "step": 13850
+    },
+    {
+      "epoch": 4.048939641109299,
+      "grad_norm": 0.32494664192199707,
+      "learning_rate": 0.0005517376858058875,
+      "loss": 3.5389,
+      "step": 13900
+    },
+    {
+      "epoch": 4.063505010487066,
+      "grad_norm": 0.32617998123168945,
+      "learning_rate": 0.0005515628096764791,
+      "loss": 3.5355,
+      "step": 13950
+    },
+    {
+      "epoch": 4.078070379864833,
+      "grad_norm": 0.325418621301651,
+      "learning_rate": 0.0005513879335470708,
+      "loss": 3.5475,
+      "step": 14000
+    },
+    {
+      "epoch": 4.078070379864833,
+      "eval_accuracy": 0.3561653243450883,
+      "eval_loss": 3.6729841232299805,
+      "eval_runtime": 181.0076,
+      "eval_samples_per_second": 91.935,
+      "eval_steps_per_second": 5.751,
+      "step": 14000
+    },
+    {
+      "epoch": 4.092635749242601,
+      "grad_norm": 0.31870585680007935,
+      "learning_rate": 0.0005512130574176625,
+      "loss": 3.5357,
+      "step": 14050
+    },
+    {
+      "epoch": 4.1072011186203685,
+      "grad_norm": 0.30783963203430176,
+      "learning_rate": 0.000551038181288254,
+      "loss": 3.5586,
+      "step": 14100
+    },
+    {
+      "epoch": 4.121766487998135,
+      "grad_norm": 0.31743520498275757,
+      "learning_rate": 0.0005508633051588457,
+      "loss": 3.5467,
+      "step": 14150
+    },
+    {
+      "epoch": 4.136331857375903,
+      "grad_norm": 0.30032074451446533,
+      "learning_rate": 0.0005506884290294374,
+      "loss": 3.5504,
+      "step": 14200
+    },
+    {
+      "epoch": 4.150897226753671,
+      "grad_norm": 0.30552420020103455,
+      "learning_rate": 0.0005505135529000291,
+      "loss": 3.5497,
+      "step": 14250
+    },
+    {
+      "epoch": 4.165462596131438,
+      "grad_norm": 0.3102892339229584,
+      "learning_rate": 0.0005503386767706207,
+      "loss": 3.5618,
+      "step": 14300
+    },
+    {
+      "epoch": 4.180027965509205,
+      "grad_norm": 0.32704657316207886,
+      "learning_rate": 0.0005501638006412124,
+      "loss": 3.5567,
+      "step": 14350
+    },
+    {
+      "epoch": 4.194593334886973,
+      "grad_norm": 0.30027180910110474,
+      "learning_rate": 0.0005499889245118041,
+      "loss": 3.5536,
+      "step": 14400
+    },
+    {
+      "epoch": 4.20915870426474,
+      "grad_norm": 0.3081790804862976,
+      "learning_rate": 0.0005498140483823958,
+      "loss": 3.5608,
+      "step": 14450
+    },
+    {
+      "epoch": 4.223724073642508,
+      "grad_norm": 0.30942654609680176,
+      "learning_rate": 0.0005496391722529875,
+      "loss": 3.5534,
+      "step": 14500
+    },
+    {
+      "epoch": 4.238289443020275,
+      "grad_norm": 0.3231661915779114,
+      "learning_rate": 0.000549464296123579,
+      "loss": 3.5601,
+      "step": 14550
+    },
+    {
+      "epoch": 4.252854812398042,
+      "grad_norm": 0.32545921206474304,
+      "learning_rate": 0.0005492894199941707,
+      "loss": 3.5609,
+      "step": 14600
+    },
+    {
+      "epoch": 4.26742018177581,
+      "grad_norm": 0.32660067081451416,
+      "learning_rate": 0.0005491145438647624,
+      "loss": 3.5699,
+      "step": 14650
+    },
+    {
+      "epoch": 4.281985551153578,
+      "grad_norm": 0.299081027507782,
+      "learning_rate": 0.000548939667735354,
+      "loss": 3.5718,
+      "step": 14700
+    },
+    {
+      "epoch": 4.296550920531344,
+      "grad_norm": 0.3111097514629364,
+      "learning_rate": 0.0005487647916059457,
+      "loss": 3.5505,
+      "step": 14750
+    },
+    {
+      "epoch": 4.311116289909112,
+      "grad_norm": 0.31659451127052307,
+      "learning_rate": 0.0005485899154765374,
+      "loss": 3.5575,
+      "step": 14800
+    },
+    {
+      "epoch": 4.32568165928688,
+      "grad_norm": 0.305717408657074,
+      "learning_rate": 0.0005484150393471291,
+      "loss": 3.5839,
+      "step": 14850
+    },
+    {
+      "epoch": 4.340247028664647,
+      "grad_norm": 0.3144332766532898,
+      "learning_rate": 0.0005482401632177208,
+      "loss": 3.5662,
+      "step": 14900
+    },
+    {
+      "epoch": 4.354812398042414,
+      "grad_norm": 0.34151017665863037,
+      "learning_rate": 0.0005480652870883124,
+      "loss": 3.5686,
+      "step": 14950
+    },
+    {
+      "epoch": 4.369377767420182,
+      "grad_norm": 0.2984902560710907,
+      "learning_rate": 0.000547890410958904,
+      "loss": 3.5685,
+      "step": 15000
+    },
+    {
+      "epoch": 4.369377767420182,
+      "eval_accuracy": 0.3569452338205533,
+      "eval_loss": 3.661975622177124,
+      "eval_runtime": 181.0485,
+      "eval_samples_per_second": 91.915,
+      "eval_steps_per_second": 5.75,
+      "step": 15000
+    },
+    {
+      "epoch": 4.383943136797949,
+      "grad_norm": 0.303886353969574,
+      "learning_rate": 0.0005477155348294957,
+      "loss": 3.5641,
+      "step": 15050
+    },
+    {
+      "epoch": 4.398508506175717,
+      "grad_norm": 0.29367125034332275,
+      "learning_rate": 0.0005475406587000874,
+      "loss": 3.5626,
+      "step": 15100
+    },
+    {
+      "epoch": 4.413073875553484,
+      "grad_norm": 0.3227385878562927,
+      "learning_rate": 0.000547365782570679,
+      "loss": 3.5753,
+      "step": 15150
+    },
+    {
+      "epoch": 4.427639244931251,
+      "grad_norm": 0.31565627455711365,
+      "learning_rate": 0.0005471909064412707,
+      "loss": 3.5752,
+      "step": 15200
+    },
+    {
+      "epoch": 4.442204614309019,
+      "grad_norm": 0.30896371603012085,
+      "learning_rate": 0.0005470160303118624,
+      "loss": 3.5645,
+      "step": 15250
+    },
+    {
+      "epoch": 4.456769983686787,
+      "grad_norm": 0.332472026348114,
+      "learning_rate": 0.000546841154182454,
+      "loss": 3.5725,
+      "step": 15300
+    },
+    {
+      "epoch": 4.471335353064553,
+      "grad_norm": 0.30785223841667175,
+      "learning_rate": 0.0005466662780530458,
+      "loss": 3.5719,
+      "step": 15350
+    },
+    {
+      "epoch": 4.485900722442321,
+      "grad_norm": 0.316549152135849,
+      "learning_rate": 0.0005464914019236374,
+      "loss": 3.5717,
+      "step": 15400
+    },
+    {
+      "epoch": 4.500466091820089,
+      "grad_norm": 0.3038575351238251,
+      "learning_rate": 0.000546316525794229,
+      "loss": 3.583,
+      "step": 15450
+    },
+    {
+      "epoch": 4.515031461197856,
+      "grad_norm": 0.31019827723503113,
+      "learning_rate": 0.0005461416496648207,
+      "loss": 3.5699,
+      "step": 15500
+    },
+    {
+      "epoch": 4.529596830575623,
+      "grad_norm": 0.29632434248924255,
+      "learning_rate": 0.0005459667735354123,
+      "loss": 3.5858,
+      "step": 15550
+    },
+    {
+      "epoch": 4.544162199953391,
+      "grad_norm": 0.31503385305404663,
+      "learning_rate": 0.000545791897406004,
+      "loss": 3.5735,
+      "step": 15600
+    },
+    {
+      "epoch": 4.558727569331158,
+      "grad_norm": 0.3121156692504883,
+      "learning_rate": 0.0005456170212765957,
+      "loss": 3.5692,
+      "step": 15650
+    },
+    {
+      "epoch": 4.573292938708926,
+      "grad_norm": 0.31484994292259216,
+      "learning_rate": 0.0005454421451471874,
+      "loss": 3.5725,
+      "step": 15700
+    },
+    {
+      "epoch": 4.587858308086693,
+      "grad_norm": 0.3038138449192047,
+      "learning_rate": 0.000545267269017779,
+      "loss": 3.5573,
+      "step": 15750
+    },
+    {
+      "epoch": 4.6024236774644605,
+      "grad_norm": 0.30718371272087097,
+      "learning_rate": 0.0005450923928883708,
+      "loss": 3.5645,
+      "step": 15800
+    },
+    {
+      "epoch": 4.616989046842228,
+      "grad_norm": 0.30330055952072144,
+      "learning_rate": 0.0005449175167589623,
+      "loss": 3.5691,
+      "step": 15850
+    },
+    {
+      "epoch": 4.631554416219995,
+      "grad_norm": 0.3052199184894562,
+      "learning_rate": 0.000544742640629554,
+      "loss": 3.5812,
+      "step": 15900
+    },
+    {
+      "epoch": 4.6461197855977625,
+      "grad_norm": 0.32241594791412354,
+      "learning_rate": 0.0005445677645001457,
+      "loss": 3.5724,
+      "step": 15950
+    },
+    {
+      "epoch": 4.66068515497553,
+      "grad_norm": 0.33788686990737915,
+      "learning_rate": 0.0005443928883707373,
+      "loss": 3.5778,
+      "step": 16000
+    },
+    {
+      "epoch": 4.66068515497553,
+      "eval_accuracy": 0.3581711922466273,
+      "eval_loss": 3.6478781700134277,
+      "eval_runtime": 181.0629,
+      "eval_samples_per_second": 91.907,
+      "eval_steps_per_second": 5.749,
+      "step": 16000
+    },
+    {
+      "epoch": 4.675250524353298,
+      "grad_norm": 0.3098940849304199,
+      "learning_rate": 0.000544218012241329,
+      "loss": 3.5717,
+      "step": 16050
+    },
+    {
+      "epoch": 4.689815893731065,
+      "grad_norm": 0.3171354830265045,
+      "learning_rate": 0.0005440431361119207,
+      "loss": 3.5803,
+      "step": 16100
+    },
+    {
+      "epoch": 4.704381263108832,
+      "grad_norm": 0.31793129444122314,
+      "learning_rate": 0.0005438682599825123,
+      "loss": 3.5774,
+      "step": 16150
+    },
+    {
+      "epoch": 4.7189466324866,
+      "grad_norm": 0.30565646290779114,
+      "learning_rate": 0.000543693383853104,
+      "loss": 3.5956,
+      "step": 16200
+    },
+    {
+      "epoch": 4.7335120018643675,
+      "grad_norm": 0.31578201055526733,
+      "learning_rate": 0.0005435185077236957,
+      "loss": 3.5785,
+      "step": 16250
+    },
+    {
+      "epoch": 4.748077371242134,
+      "grad_norm": 0.3025978207588196,
+      "learning_rate": 0.0005433436315942873,
+      "loss": 3.5678,
+      "step": 16300
+    },
+    {
+      "epoch": 4.762642740619902,
+      "grad_norm": 0.30548781156539917,
+      "learning_rate": 0.000543168755464879,
+      "loss": 3.5747,
+      "step": 16350
+    },
+    {
+      "epoch": 4.7772081099976695,
+      "grad_norm": 0.3177635967731476,
+      "learning_rate": 0.0005429938793354706,
+      "loss": 3.5717,
+      "step": 16400
+    },
+    {
+      "epoch": 4.791773479375437,
+      "grad_norm": 0.30422160029411316,
+      "learning_rate": 0.0005428190032060623,
+      "loss": 3.5696,
+      "step": 16450
+    },
+    {
+      "epoch": 4.806338848753205,
+      "grad_norm": 0.28537696599960327,
+      "learning_rate": 0.000542644127076654,
+      "loss": 3.5694,
+      "step": 16500
+    },
+    {
+      "epoch": 4.820904218130972,
+      "grad_norm": 0.3017064034938812,
+      "learning_rate": 0.0005424692509472457,
+      "loss": 3.581,
+      "step": 16550
+    },
+    {
+      "epoch": 4.835469587508739,
+      "grad_norm": 0.2944038212299347,
+      "learning_rate": 0.0005422943748178373,
+      "loss": 3.5788,
+      "step": 16600
+    },
+    {
+      "epoch": 4.850034956886507,
+      "grad_norm": 0.2914831042289734,
+      "learning_rate": 0.000542119498688429,
+      "loss": 3.58,
+      "step": 16650
+    },
+    {
+      "epoch": 4.864600326264274,
+      "grad_norm": 0.3166540563106537,
+      "learning_rate": 0.0005419446225590207,
+      "loss": 3.5807,
+      "step": 16700
+    },
+    {
+      "epoch": 4.879165695642041,
+      "grad_norm": 0.3148166835308075,
+      "learning_rate": 0.0005417697464296122,
+      "loss": 3.5777,
+      "step": 16750
+    },
+    {
+      "epoch": 4.893731065019809,
+      "grad_norm": 0.32924884557724,
+      "learning_rate": 0.000541594870300204,
+      "loss": 3.5819,
+      "step": 16800
+    },
+    {
+      "epoch": 4.908296434397577,
+      "grad_norm": 0.29883652925491333,
+      "learning_rate": 0.0005414199941707956,
+      "loss": 3.5849,
+      "step": 16850
+    },
+    {
+      "epoch": 4.922861803775344,
+      "grad_norm": 0.346866637468338,
+      "learning_rate": 0.0005412451180413873,
+      "loss": 3.5688,
+      "step": 16900
+    },
+    {
+      "epoch": 4.937427173153111,
+      "grad_norm": 0.2987111806869507,
+      "learning_rate": 0.000541070241911979,
+      "loss": 3.5742,
+      "step": 16950
+    },
+    {
+      "epoch": 4.951992542530879,
+      "grad_norm": 0.316847026348114,
+      "learning_rate": 0.0005408953657825706,
+      "loss": 3.5844,
+      "step": 17000
+    },
+    {
+      "epoch": 4.951992542530879,
+      "eval_accuracy": 0.3594956977385095,
+      "eval_loss": 3.631971597671509,
+      "eval_runtime": 180.9523,
+      "eval_samples_per_second": 91.963,
+      "eval_steps_per_second": 5.753,
+      "step": 17000
+    },
+    {
+      "epoch": 4.966557911908646,
+      "grad_norm": 0.2994118928909302,
+      "learning_rate": 0.0005407204896531623,
+      "loss": 3.567,
+      "step": 17050
+    },
+    {
+      "epoch": 4.981123281286413,
+      "grad_norm": 0.32432621717453003,
+      "learning_rate": 0.000540545613523754,
+      "loss": 3.5801,
+      "step": 17100
+    },
+    {
+      "epoch": 4.995688650664181,
+      "grad_norm": 0.30162033438682556,
+      "learning_rate": 0.0005403707373943456,
+      "loss": 3.5888,
+      "step": 17150
+    },
+    {
+      "epoch": 5.010195758564437,
+      "grad_norm": 0.33187609910964966,
+      "learning_rate": 0.0005401958612649372,
+      "loss": 3.5053,
+      "step": 17200
+    },
+    {
+      "epoch": 5.024761127942204,
+      "grad_norm": 0.3020380139350891,
+      "learning_rate": 0.000540020985135529,
+      "loss": 3.4737,
+      "step": 17250
+    },
+    {
+      "epoch": 5.039326497319972,
+      "grad_norm": 0.32110413908958435,
+      "learning_rate": 0.0005398461090061206,
+      "loss": 3.4768,
+      "step": 17300
+    },
+    {
+      "epoch": 5.0538918666977395,
+      "grad_norm": 0.29977384209632874,
+      "learning_rate": 0.0005396712328767123,
+      "loss": 3.4863,
+      "step": 17350
+    },
+    {
+      "epoch": 5.068457236075507,
+      "grad_norm": 0.30590590834617615,
+      "learning_rate": 0.000539496356747304,
+      "loss": 3.4695,
+      "step": 17400
+    },
+    {
+      "epoch": 5.083022605453274,
+      "grad_norm": 0.31782832741737366,
+      "learning_rate": 0.0005393214806178956,
+      "loss": 3.4834,
+      "step": 17450
+    },
+    {
+      "epoch": 5.0975879748310415,
+      "grad_norm": 0.3071076571941376,
+      "learning_rate": 0.0005391466044884873,
+      "loss": 3.4811,
+      "step": 17500
+    },
+    {
+      "epoch": 5.112153344208809,
+      "grad_norm": 0.30710092186927795,
+      "learning_rate": 0.000538971728359079,
+      "loss": 3.4857,
+      "step": 17550
+    },
+    {
+      "epoch": 5.126718713586577,
+      "grad_norm": 0.3052123486995697,
+      "learning_rate": 0.0005387968522296705,
+      "loss": 3.493,
+      "step": 17600
+    },
+    {
+      "epoch": 5.141284082964344,
+      "grad_norm": 0.31805649399757385,
+      "learning_rate": 0.0005386219761002622,
+      "loss": 3.4955,
+      "step": 17650
+    },
+    {
+      "epoch": 5.155849452342111,
+      "grad_norm": 0.3294351398944855,
+      "learning_rate": 0.0005384470999708539,
+      "loss": 3.4967,
+      "step": 17700
+    },
+    {
+      "epoch": 5.170414821719879,
+      "grad_norm": 0.3294854760169983,
+      "learning_rate": 0.0005382722238414456,
+      "loss": 3.4918,
+      "step": 17750
+    },
+    {
+      "epoch": 5.1849801910976465,
+      "grad_norm": 0.3198247253894806,
+      "learning_rate": 0.0005380973477120373,
+      "loss": 3.499,
+      "step": 17800
+    },
+    {
+      "epoch": 5.199545560475413,
+      "grad_norm": 0.32969656586647034,
+      "learning_rate": 0.000537922471582629,
+      "loss": 3.497,
+      "step": 17850
+    },
+    {
+      "epoch": 5.214110929853181,
+      "grad_norm": 0.3068716526031494,
+      "learning_rate": 0.0005377475954532206,
+      "loss": 3.4973,
+      "step": 17900
+    },
+    {
+      "epoch": 5.228676299230949,
+      "grad_norm": 0.31869903206825256,
+      "learning_rate": 0.0005375727193238123,
+      "loss": 3.5006,
+      "step": 17950
+    },
+    {
+      "epoch": 5.243241668608716,
+      "grad_norm": 0.3365494906902313,
+      "learning_rate": 0.000537397843194404,
+      "loss": 3.5025,
+      "step": 18000
+    },
+    {
+      "epoch": 5.243241668608716,
+      "eval_accuracy": 0.35994151149325737,
+      "eval_loss": 3.6374263763427734,
+      "eval_runtime": 181.0956,
+      "eval_samples_per_second": 91.891,
+      "eval_steps_per_second": 5.748,
+      "step": 18000
+    },
+    {
+      "epoch": 5.257807037986483,
+      "grad_norm": 0.3093796968460083,
+      "learning_rate": 0.0005372229670649955,
+      "loss": 3.5083,
+      "step": 18050
+    },
+    {
+      "epoch": 5.272372407364251,
+      "grad_norm": 0.3284550607204437,
+      "learning_rate": 0.0005370480909355872,
+      "loss": 3.5058,
+      "step": 18100
+    },
+    {
+      "epoch": 5.286937776742018,
+      "grad_norm": 0.2993088662624359,
+      "learning_rate": 0.0005368732148061789,
+      "loss": 3.5142,
+      "step": 18150
+    },
+    {
+      "epoch": 5.301503146119786,
+      "grad_norm": 0.31368738412857056,
+      "learning_rate": 0.0005366983386767705,
+      "loss": 3.5082,
+      "step": 18200
+    },
+    {
+      "epoch": 5.316068515497553,
+      "grad_norm": 0.31878387928009033,
+      "learning_rate": 0.0005365234625473623,
+      "loss": 3.5152,
+      "step": 18250
+    },
+    {
+      "epoch": 5.33063388487532,
+      "grad_norm": 0.3237106502056122,
+      "learning_rate": 0.0005363485864179539,
+      "loss": 3.5046,
+      "step": 18300
+    },
+    {
+      "epoch": 5.345199254253088,
+      "grad_norm": 0.31487080454826355,
+      "learning_rate": 0.0005361737102885456,
+      "loss": 3.51,
+      "step": 18350
+    },
+    {
+      "epoch": 5.359764623630856,
+      "grad_norm": 0.31325918436050415,
+      "learning_rate": 0.0005359988341591373,
+      "loss": 3.5025,
+      "step": 18400
+    },
+    {
+      "epoch": 5.374329993008622,
+      "grad_norm": 0.3141399621963501,
+      "learning_rate": 0.000535823958029729,
+      "loss": 3.5097,
+      "step": 18450
+    },
+    {
+      "epoch": 5.38889536238639,
+      "grad_norm": 0.3144991993904114,
+      "learning_rate": 0.0005356490819003205,
+      "loss": 3.528,
+      "step": 18500
+    },
+    {
+      "epoch": 5.403460731764158,
+      "grad_norm": 0.34297963976860046,
+      "learning_rate": 0.0005354742057709122,
+      "loss": 3.5246,
+      "step": 18550
+    },
+    {
+      "epoch": 5.418026101141925,
+      "grad_norm": 0.3127841353416443,
+      "learning_rate": 0.0005352993296415039,
+      "loss": 3.5066,
+      "step": 18600
+    },
+    {
+      "epoch": 5.432591470519692,
+      "grad_norm": 0.3127472400665283,
+      "learning_rate": 0.0005351244535120955,
+      "loss": 3.5295,
+      "step": 18650
+    },
+    {
+      "epoch": 5.44715683989746,
+      "grad_norm": 0.30751973390579224,
+      "learning_rate": 0.0005349495773826873,
+      "loss": 3.5081,
+      "step": 18700
+    },
+    {
+      "epoch": 5.461722209275227,
+      "grad_norm": 0.3045814037322998,
+      "learning_rate": 0.0005347747012532789,
+      "loss": 3.5063,
+      "step": 18750
+    },
+    {
+      "epoch": 5.476287578652995,
+      "grad_norm": 0.3257751166820526,
+      "learning_rate": 0.0005345998251238706,
+      "loss": 3.5257,
+      "step": 18800
+    },
+    {
+      "epoch": 5.490852948030762,
+      "grad_norm": 0.3065994679927826,
+      "learning_rate": 0.0005344249489944623,
+      "loss": 3.5187,
+      "step": 18850
+    },
+    {
+      "epoch": 5.505418317408529,
+      "grad_norm": 0.31210583448410034,
+      "learning_rate": 0.0005342500728650538,
+      "loss": 3.5167,
+      "step": 18900
+    },
+    {
+      "epoch": 5.519983686786297,
+      "grad_norm": 0.32865795493125916,
+      "learning_rate": 0.0005340751967356455,
+      "loss": 3.5198,
+      "step": 18950
+    },
+    {
+      "epoch": 5.534549056164065,
+      "grad_norm": 0.34612202644348145,
+      "learning_rate": 0.0005339003206062372,
+      "loss": 3.5089,
+      "step": 19000
+    },
+    {
+      "epoch": 5.534549056164065,
+      "eval_accuracy": 0.36097202215874286,
+      "eval_loss": 3.6249496936798096,
+      "eval_runtime": 181.005,
+      "eval_samples_per_second": 91.937,
+      "eval_steps_per_second": 5.751,
+      "step": 19000
+    },
+    {
+      "epoch": 5.549114425541831,
+      "grad_norm": 0.3225538730621338,
+      "learning_rate": 0.0005337254444768288,
+      "loss": 3.534,
+      "step": 19050
+    },
+    {
+      "epoch": 5.563679794919599,
+      "grad_norm": 0.2896052896976471,
+      "learning_rate": 0.0005335505683474205,
+      "loss": 3.5225,
+      "step": 19100
+    },
+    {
+      "epoch": 5.578245164297367,
+      "grad_norm": 0.3070700466632843,
+      "learning_rate": 0.0005333756922180122,
+      "loss": 3.5112,
+      "step": 19150
+    },
+    {
+      "epoch": 5.592810533675134,
+      "grad_norm": 0.3060336709022522,
+      "learning_rate": 0.0005332008160886039,
+      "loss": 3.5306,
+      "step": 19200
+    },
+    {
+      "epoch": 5.607375903052901,
+      "grad_norm": 0.29444536566734314,
+      "learning_rate": 0.0005330259399591956,
+      "loss": 3.5225,
+      "step": 19250
+    },
+    {
+      "epoch": 5.621941272430669,
+      "grad_norm": 0.3106236755847931,
+      "learning_rate": 0.0005328510638297873,
+      "loss": 3.5148,
+      "step": 19300
+    },
+    {
+      "epoch": 5.636506641808436,
+      "grad_norm": 0.33646246790885925,
+      "learning_rate": 0.0005326761877003788,
+      "loss": 3.5167,
+      "step": 19350
+    },
+    {
+      "epoch": 5.651072011186204,
+      "grad_norm": 0.3295743465423584,
+      "learning_rate": 0.0005325013115709705,
+      "loss": 3.5335,
+      "step": 19400
+    },
+    {
+      "epoch": 5.665637380563971,
+      "grad_norm": 0.30073535442352295,
+      "learning_rate": 0.0005323264354415622,
+      "loss": 3.5081,
+      "step": 19450
+    },
+    {
+      "epoch": 5.6802027499417385,
+      "grad_norm": 0.3157036304473877,
+      "learning_rate": 0.0005321515593121538,
+      "loss": 3.5297,
+      "step": 19500
+    },
+    {
+      "epoch": 5.694768119319506,
+      "grad_norm": 0.31594860553741455,
+      "learning_rate": 0.0005319766831827455,
+      "loss": 3.5218,
+      "step": 19550
+    },
+    {
+      "epoch": 5.709333488697274,
+      "grad_norm": 0.3158119320869446,
+      "learning_rate": 0.0005318018070533372,
+      "loss": 3.5214,
+      "step": 19600
+    },
+    {
+      "epoch": 5.7238988580750405,
+      "grad_norm": 0.31483352184295654,
+      "learning_rate": 0.0005316269309239288,
+      "loss": 3.5336,
+      "step": 19650
+    },
+    {
+      "epoch": 5.738464227452808,
+      "grad_norm": 0.3177572190761566,
+      "learning_rate": 0.0005314520547945206,
+      "loss": 3.5449,
+      "step": 19700
+    },
+    {
+      "epoch": 5.753029596830576,
+      "grad_norm": 0.32383835315704346,
+      "learning_rate": 0.0005312771786651121,
+      "loss": 3.5343,
+      "step": 19750
+    },
+    {
+      "epoch": 5.7675949662083426,
+      "grad_norm": 0.3220798671245575,
+      "learning_rate": 0.0005311023025357038,
+      "loss": 3.5266,
+      "step": 19800
+    },
+    {
+      "epoch": 5.78216033558611,
+      "grad_norm": 0.3043564558029175,
+      "learning_rate": 0.0005309274264062955,
+      "loss": 3.5257,
+      "step": 19850
+    },
+    {
+      "epoch": 5.796725704963878,
+      "grad_norm": 0.3111497461795807,
+      "learning_rate": 0.0005307525502768872,
+      "loss": 3.531,
+      "step": 19900
+    },
+    {
+      "epoch": 5.8112910743416455,
+      "grad_norm": 0.31145745515823364,
+      "learning_rate": 0.0005305776741474788,
+      "loss": 3.5113,
+      "step": 19950
+    },
+    {
+      "epoch": 5.825856443719413,
+      "grad_norm": 0.3015078008174896,
+      "learning_rate": 0.0005304027980180705,
+      "loss": 3.5334,
+      "step": 20000
+    },
+    {
+      "epoch": 5.825856443719413,
+      "eval_accuracy": 0.3619600799712967,
+      "eval_loss": 3.6139492988586426,
+      "eval_runtime": 181.0023,
+      "eval_samples_per_second": 91.938,
+      "eval_steps_per_second": 5.751,
+      "step": 20000
+    },
+    {
+      "epoch": 5.84042181309718,
+      "grad_norm": 0.3019517958164215,
+      "learning_rate": 0.0005302279218886622,
+      "loss": 3.5292,
+      "step": 20050
+    },
+    {
+      "epoch": 5.8549871824749475,
+      "grad_norm": 0.31948110461235046,
+      "learning_rate": 0.0005300530457592538,
+      "loss": 3.5435,
+      "step": 20100
+    },
+    {
+      "epoch": 5.869552551852715,
+      "grad_norm": 0.3611204922199249,
+      "learning_rate": 0.0005298781696298456,
+      "loss": 3.5351,
+      "step": 20150
+    },
+    {
+      "epoch": 5.884117921230482,
+      "grad_norm": 0.32438987493515015,
+      "learning_rate": 0.0005297032935004371,
+      "loss": 3.5399,
+      "step": 20200
+    },
+    {
+      "epoch": 5.89868329060825,
+      "grad_norm": 0.3309918940067291,
+      "learning_rate": 0.0005295284173710288,
+      "loss": 3.5355,
+      "step": 20250
+    },
+    {
+      "epoch": 5.913248659986017,
+      "grad_norm": 0.31693235039711,
+      "learning_rate": 0.0005293535412416205,
+      "loss": 3.525,
+      "step": 20300
+    },
+    {
+      "epoch": 5.927814029363785,
+      "grad_norm": 0.3211771249771118,
+      "learning_rate": 0.0005291786651122121,
+      "loss": 3.5434,
+      "step": 20350
+    },
+    {
+      "epoch": 5.9423793987415525,
+      "grad_norm": 0.3379644453525543,
+      "learning_rate": 0.0005290037889828038,
+      "loss": 3.548,
+      "step": 20400
+    },
+    {
+      "epoch": 5.956944768119319,
+      "grad_norm": 0.32558566331863403,
+      "learning_rate": 0.0005288289128533955,
+      "loss": 3.5169,
+      "step": 20450
+    },
+    {
+      "epoch": 5.971510137497087,
+      "grad_norm": 0.3327178359031677,
+      "learning_rate": 0.0005286540367239872,
+      "loss": 3.5232,
+      "step": 20500
+    },
+    {
+      "epoch": 5.986075506874855,
+      "grad_norm": 0.3202207684516907,
+      "learning_rate": 0.0005284791605945788,
+      "loss": 3.5359,
+      "step": 20550
+    },
+    {
+      "epoch": 6.0005826147751105,
+      "grad_norm": 0.32077357172966003,
+      "learning_rate": 0.0005283042844651704,
+      "loss": 3.5254,
+      "step": 20600
+    },
+    {
+      "epoch": 6.015147984152878,
+      "grad_norm": 0.3209148049354553,
+      "learning_rate": 0.0005281294083357621,
+      "loss": 3.4134,
+      "step": 20650
+    },
+    {
+      "epoch": 6.029713353530646,
+      "grad_norm": 0.31453171372413635,
+      "learning_rate": 0.0005279545322063538,
+      "loss": 3.4272,
+      "step": 20700
+    },
+    {
+      "epoch": 6.044278722908413,
+      "grad_norm": 0.32655394077301025,
+      "learning_rate": 0.0005277796560769455,
+      "loss": 3.4097,
+      "step": 20750
+    },
+    {
+      "epoch": 6.05884409228618,
+      "grad_norm": 0.3791479468345642,
+      "learning_rate": 0.0005276047799475371,
+      "loss": 3.4402,
+      "step": 20800
+    },
+    {
+      "epoch": 6.073409461663948,
+      "grad_norm": 0.31815582513809204,
+      "learning_rate": 0.0005274299038181288,
+      "loss": 3.4348,
+      "step": 20850
+    },
+    {
+      "epoch": 6.087974831041715,
+      "grad_norm": 0.3264419734477997,
+      "learning_rate": 0.0005272550276887205,
+      "loss": 3.4253,
+      "step": 20900
+    },
+    {
+      "epoch": 6.102540200419483,
+      "grad_norm": 0.31985941529273987,
+      "learning_rate": 0.0005270801515593121,
+      "loss": 3.4381,
+      "step": 20950
+    },
+    {
+      "epoch": 6.11710556979725,
+      "grad_norm": 0.33856844902038574,
+      "learning_rate": 0.0005269052754299037,
+      "loss": 3.4438,
+      "step": 21000
+    },
+    {
+      "epoch": 6.11710556979725,
+      "eval_accuracy": 0.3620460440585351,
+      "eval_loss": 3.6200692653656006,
+      "eval_runtime": 181.1616,
+      "eval_samples_per_second": 91.857,
+      "eval_steps_per_second": 5.746,
+      "step": 21000
+    },
+    {
+      "epoch": 6.1316709391750175,
+      "grad_norm": 0.315507709980011,
+      "learning_rate": 0.0005267303993004954,
+      "loss": 3.453,
+      "step": 21050
+    },
+    {
+      "epoch": 6.146236308552785,
+      "grad_norm": 0.3275434970855713,
+      "learning_rate": 0.000526555523171087,
+      "loss": 3.4556,
+      "step": 21100
+    },
+    {
+      "epoch": 6.160801677930552,
+      "grad_norm": 0.2962492108345032,
+      "learning_rate": 0.0005263806470416788,
+      "loss": 3.4371,
+      "step": 21150
+    },
+    {
+      "epoch": 6.1753670473083195,
+      "grad_norm": 0.3482397198677063,
+      "learning_rate": 0.0005262057709122704,
+      "loss": 3.4355,
+      "step": 21200
+    },
+    {
+      "epoch": 6.189932416686087,
+      "grad_norm": 0.3102310299873352,
+      "learning_rate": 0.0005260308947828621,
+      "loss": 3.4526,
+      "step": 21250
+    },
+    {
+      "epoch": 6.204497786063855,
+      "grad_norm": 0.314818412065506,
+      "learning_rate": 0.0005258560186534538,
+      "loss": 3.4675,
+      "step": 21300
+    },
+    {
+      "epoch": 6.219063155441622,
+      "grad_norm": 0.3258023262023926,
+      "learning_rate": 0.0005256811425240455,
+      "loss": 3.441,
+      "step": 21350
+    },
+    {
+      "epoch": 6.233628524819389,
+      "grad_norm": 0.3221692144870758,
+      "learning_rate": 0.0005255062663946371,
+      "loss": 3.4644,
+      "step": 21400
+    },
+    {
+      "epoch": 6.248193894197157,
+      "grad_norm": 0.32585564255714417,
+      "learning_rate": 0.0005253313902652287,
+      "loss": 3.4531,
+      "step": 21450
+    },
+    {
+      "epoch": 6.2627592635749245,
+      "grad_norm": 0.3138885498046875,
+      "learning_rate": 0.0005251565141358204,
+      "loss": 3.4766,
+      "step": 21500
+    },
+    {
+      "epoch": 6.277324632952691,
+      "grad_norm": 0.33104372024536133,
+      "learning_rate": 0.000524981638006412,
+      "loss": 3.4572,
+      "step": 21550
+    },
+    {
+      "epoch": 6.291890002330459,
+      "grad_norm": 0.3003644049167633,
+      "learning_rate": 0.0005248067618770038,
+      "loss": 3.4759,
+      "step": 21600
+    },
+    {
+      "epoch": 6.306455371708227,
+      "grad_norm": 0.30780765414237976,
+      "learning_rate": 0.0005246318857475954,
+      "loss": 3.4607,
+      "step": 21650
+    },
+    {
+      "epoch": 6.321020741085994,
+      "grad_norm": 0.3122555613517761,
+      "learning_rate": 0.0005244570096181871,
+      "loss": 3.463,
+      "step": 21700
+    },
+    {
+      "epoch": 6.335586110463761,
+      "grad_norm": 0.3137441575527191,
+      "learning_rate": 0.0005242821334887788,
+      "loss": 3.4694,
+      "step": 21750
+    },
+    {
+      "epoch": 6.350151479841529,
+      "grad_norm": 0.3140689432621002,
+      "learning_rate": 0.0005241072573593704,
+      "loss": 3.4803,
+      "step": 21800
+    },
+    {
+      "epoch": 6.364716849219296,
+      "grad_norm": 0.3386966288089752,
+      "learning_rate": 0.000523932381229962,
+      "loss": 3.4822,
+      "step": 21850
+    },
+    {
+      "epoch": 6.379282218597064,
+      "grad_norm": 0.3035133183002472,
+      "learning_rate": 0.0005237575051005537,
+      "loss": 3.4826,
+      "step": 21900
+    },
+    {
+      "epoch": 6.393847587974831,
+      "grad_norm": 0.3223668932914734,
+      "learning_rate": 0.0005235826289711454,
+      "loss": 3.4827,
+      "step": 21950
+    },
+    {
+      "epoch": 6.408412957352598,
+      "grad_norm": 0.3119979798793793,
+      "learning_rate": 0.000523407752841737,
+      "loss": 3.4704,
+      "step": 22000
+    },
+    {
+      "epoch": 6.408412957352598,
+      "eval_accuracy": 0.3628611153152371,
+      "eval_loss": 3.6088709831237793,
+      "eval_runtime": 181.1625,
+      "eval_samples_per_second": 91.857,
+      "eval_steps_per_second": 5.746,
+      "step": 22000
+    },
+    {
+      "epoch": 6.422978326730366,
+      "grad_norm": 0.3329072594642639,
+      "learning_rate": 0.0005232328767123287,
+      "loss": 3.4775,
+      "step": 22050
+    },
+    {
+      "epoch": 6.437543696108134,
+      "grad_norm": 0.309803307056427,
+      "learning_rate": 0.0005230580005829204,
+      "loss": 3.479,
+      "step": 22100
+    },
+    {
+      "epoch": 6.4521090654859,
+      "grad_norm": 0.3253338038921356,
+      "learning_rate": 0.0005228831244535121,
+      "loss": 3.481,
+      "step": 22150
+    },
+    {
+      "epoch": 6.466674434863668,
+      "grad_norm": 0.32025864720344543,
+      "learning_rate": 0.0005227082483241038,
+      "loss": 3.4885,
+      "step": 22200
+    },
+    {
+      "epoch": 6.481239804241436,
+      "grad_norm": 0.3189886510372162,
+      "learning_rate": 0.0005225333721946954,
+      "loss": 3.4745,
+      "step": 22250
+    },
+    {
+      "epoch": 6.495805173619203,
+      "grad_norm": 0.320529580116272,
+      "learning_rate": 0.000522358496065287,
+      "loss": 3.4672,
+      "step": 22300
+    },
+    {
+      "epoch": 6.51037054299697,
+      "grad_norm": 0.3055493235588074,
+      "learning_rate": 0.0005221836199358787,
+      "loss": 3.4893,
+      "step": 22350
+    },
+    {
+      "epoch": 6.524935912374738,
+      "grad_norm": 0.30660557746887207,
+      "learning_rate": 0.0005220087438064703,
+      "loss": 3.4739,
+      "step": 22400
+    },
+    {
+      "epoch": 6.539501281752505,
+      "grad_norm": 0.31690749526023865,
+      "learning_rate": 0.000521833867677062,
+      "loss": 3.4961,
+      "step": 22450
+    },
+    {
+      "epoch": 6.554066651130273,
+      "grad_norm": 0.3079032897949219,
+      "learning_rate": 0.0005216589915476537,
+      "loss": 3.484,
+      "step": 22500
+    },
+    {
+      "epoch": 6.56863202050804,
+      "grad_norm": 0.33683452010154724,
+      "learning_rate": 0.0005214841154182454,
+      "loss": 3.4659,
+      "step": 22550
+    },
+    {
+      "epoch": 6.583197389885807,
+      "grad_norm": 0.3125605881214142,
+      "learning_rate": 0.0005213092392888371,
+      "loss": 3.4816,
+      "step": 22600
+    },
+    {
+      "epoch": 6.597762759263575,
+      "grad_norm": 0.3357701897621155,
+      "learning_rate": 0.0005211343631594287,
+      "loss": 3.4893,
+      "step": 22650
+    },
+    {
+      "epoch": 6.612328128641343,
+      "grad_norm": 0.31705376505851746,
+      "learning_rate": 0.0005209594870300204,
+      "loss": 3.486,
+      "step": 22700
+    },
+    {
+      "epoch": 6.626893498019109,
+      "grad_norm": 0.31996995210647583,
+      "learning_rate": 0.000520784610900612,
+      "loss": 3.4754,
+      "step": 22750
+    },
+    {
+      "epoch": 6.641458867396877,
+      "grad_norm": 0.31893283128738403,
+      "learning_rate": 0.0005206097347712037,
+      "loss": 3.4731,
+      "step": 22800
+    },
+    {
+      "epoch": 6.656024236774645,
+      "grad_norm": 0.3142924904823303,
+      "learning_rate": 0.0005204348586417953,
+      "loss": 3.4902,
+      "step": 22850
+    },
+    {
+      "epoch": 6.670589606152412,
+      "grad_norm": 0.31704509258270264,
+      "learning_rate": 0.000520259982512387,
+      "loss": 3.4902,
+      "step": 22900
+    },
+    {
+      "epoch": 6.685154975530179,
+      "grad_norm": 0.31690555810928345,
+      "learning_rate": 0.0005200851063829787,
+      "loss": 3.4898,
+      "step": 22950
+    },
+    {
+      "epoch": 6.699720344907947,
+      "grad_norm": 0.31461670994758606,
+      "learning_rate": 0.0005199102302535703,
+      "loss": 3.4899,
+      "step": 23000
+    },
+    {
+      "epoch": 6.699720344907947,
+      "eval_accuracy": 0.36383706054094345,
+      "eval_loss": 3.600382089614868,
+      "eval_runtime": 181.0314,
+      "eval_samples_per_second": 91.923,
+      "eval_steps_per_second": 5.75,
+      "step": 23000
+    },
+    {
+      "epoch": 6.714285714285714,
+      "grad_norm": 0.29567137360572815,
+      "learning_rate": 0.0005197353541241621,
+      "loss": 3.48,
+      "step": 23050
+    },
+    {
+      "epoch": 6.728851083663482,
+      "grad_norm": 0.31721678376197815,
+      "learning_rate": 0.0005195604779947537,
+      "loss": 3.4899,
+      "step": 23100
+    },
+    {
+      "epoch": 6.743416453041249,
+      "grad_norm": 0.3300456702709198,
+      "learning_rate": 0.0005193856018653454,
+      "loss": 3.4828,
+      "step": 23150
+    },
+    {
+      "epoch": 6.7579818224190165,
+      "grad_norm": 0.32258230447769165,
+      "learning_rate": 0.000519210725735937,
+      "loss": 3.4954,
+      "step": 23200
+    },
+    {
+      "epoch": 6.772547191796784,
+      "grad_norm": 0.31654587388038635,
+      "learning_rate": 0.0005190358496065286,
+      "loss": 3.4821,
+      "step": 23250
+    },
+    {
+      "epoch": 6.787112561174552,
+      "grad_norm": 0.3079007863998413,
+      "learning_rate": 0.0005188609734771203,
+      "loss": 3.4949,
+      "step": 23300
+    },
+    {
+      "epoch": 6.8016779305523185,
+      "grad_norm": 0.32757923007011414,
+      "learning_rate": 0.000518686097347712,
+      "loss": 3.4913,
+      "step": 23350
+    },
+    {
+      "epoch": 6.816243299930086,
+      "grad_norm": 0.32713207602500916,
+      "learning_rate": 0.0005185112212183037,
+      "loss": 3.4813,
+      "step": 23400
+    },
+    {
+      "epoch": 6.830808669307854,
+      "grad_norm": 0.3143518567085266,
+      "learning_rate": 0.0005183363450888953,
+      "loss": 3.4858,
+      "step": 23450
+    },
+    {
+      "epoch": 6.845374038685621,
+      "grad_norm": 0.318615198135376,
+      "learning_rate": 0.000518161468959487,
+      "loss": 3.485,
+      "step": 23500
+    },
+    {
+      "epoch": 6.859939408063388,
+      "grad_norm": 0.31687304377555847,
+      "learning_rate": 0.0005179865928300787,
+      "loss": 3.4766,
+      "step": 23550
+    },
+    {
+      "epoch": 6.874504777441156,
+      "grad_norm": 0.304725706577301,
+      "learning_rate": 0.0005178117167006703,
+      "loss": 3.4977,
+      "step": 23600
+    },
+    {
+      "epoch": 6.8890701468189235,
+      "grad_norm": 0.3037196099758148,
+      "learning_rate": 0.000517636840571262,
+      "loss": 3.4889,
+      "step": 23650
+    },
+    {
+      "epoch": 6.903635516196691,
+      "grad_norm": 0.3093416094779968,
+      "learning_rate": 0.0005174619644418536,
+      "loss": 3.4898,
+      "step": 23700
+    },
+    {
+      "epoch": 6.918200885574458,
+      "grad_norm": 0.30834659934043884,
+      "learning_rate": 0.0005172870883124453,
+      "loss": 3.4879,
+      "step": 23750
+    },
+    {
+      "epoch": 6.9327662549522255,
+      "grad_norm": 0.3172142803668976,
+      "learning_rate": 0.000517112212183037,
+      "loss": 3.485,
+      "step": 23800
+    },
+    {
+      "epoch": 6.947331624329993,
+      "grad_norm": 0.309283047914505,
+      "learning_rate": 0.0005169373360536286,
+      "loss": 3.4967,
+      "step": 23850
+    },
+    {
+      "epoch": 6.961896993707761,
+      "grad_norm": 0.3254207372665405,
+      "learning_rate": 0.0005167624599242203,
+      "loss": 3.485,
+      "step": 23900
+    },
+    {
+      "epoch": 6.976462363085528,
+      "grad_norm": 0.336944580078125,
+      "learning_rate": 0.000516587583794812,
+      "loss": 3.4879,
+      "step": 23950
+    },
+    {
+      "epoch": 6.991027732463295,
+      "grad_norm": 0.3161655068397522,
+      "learning_rate": 0.0005164127076654037,
+      "loss": 3.4974,
+      "step": 24000
+    },
+    {
+      "epoch": 6.991027732463295,
+      "eval_accuracy": 0.36471328272153597,
+      "eval_loss": 3.5900657176971436,
+      "eval_runtime": 181.0515,
+      "eval_samples_per_second": 91.913,
+      "eval_steps_per_second": 5.75,
+      "step": 24000
+    },
+    {
+      "epoch": 7.005534840363552,
+      "grad_norm": 0.3199058175086975,
+      "learning_rate": 0.0005162378315359953,
+      "loss": 3.4538,
+      "step": 24050
+    },
+    {
+      "epoch": 7.020100209741319,
+      "grad_norm": 0.3226770758628845,
+      "learning_rate": 0.0005160629554065869,
+      "loss": 3.3792,
+      "step": 24100
+    },
+    {
+      "epoch": 7.034665579119086,
+      "grad_norm": 0.3223600387573242,
+      "learning_rate": 0.0005158880792771786,
+      "loss": 3.3864,
+      "step": 24150
+    },
+    {
+      "epoch": 7.049230948496854,
+      "grad_norm": 0.3161149024963379,
+      "learning_rate": 0.0005157132031477703,
+      "loss": 3.3869,
+      "step": 24200
+    },
+    {
+      "epoch": 7.063796317874622,
+      "grad_norm": 0.3132457137107849,
+      "learning_rate": 0.000515538327018362,
+      "loss": 3.4001,
+      "step": 24250
+    },
+    {
+      "epoch": 7.0783616872523885,
+      "grad_norm": 0.347877562046051,
+      "learning_rate": 0.0005153634508889536,
+      "loss": 3.4111,
+      "step": 24300
+    },
+    {
+      "epoch": 7.092927056630156,
+      "grad_norm": 0.30582693219184875,
+      "learning_rate": 0.0005151885747595453,
+      "loss": 3.407,
+      "step": 24350
+    },
+    {
+      "epoch": 7.107492426007924,
+      "grad_norm": 0.33908724784851074,
+      "learning_rate": 0.000515013698630137,
+      "loss": 3.397,
+      "step": 24400
+    },
+    {
+      "epoch": 7.122057795385691,
+      "grad_norm": 0.5248947143554688,
+      "learning_rate": 0.0005148388225007285,
+      "loss": 3.4088,
+      "step": 24450
+    },
+    {
+      "epoch": 7.136623164763458,
+      "grad_norm": 0.3276110291481018,
+      "learning_rate": 0.0005146639463713203,
+      "loss": 3.3978,
+      "step": 24500
+    },
+    {
+      "epoch": 7.151188534141226,
+      "grad_norm": 0.32857128977775574,
+      "learning_rate": 0.0005144890702419119,
+      "loss": 3.4162,
+      "step": 24550
+    },
+    {
+      "epoch": 7.165753903518993,
+      "grad_norm": 0.32675600051879883,
+      "learning_rate": 0.0005143141941125036,
+      "loss": 3.4251,
+      "step": 24600
+    },
+    {
+      "epoch": 7.180319272896761,
+      "grad_norm": 0.3161190152168274,
+      "learning_rate": 0.0005141393179830953,
+      "loss": 3.4046,
+      "step": 24650
+    },
+    {
+      "epoch": 7.194884642274528,
+      "grad_norm": 0.306657999753952,
+      "learning_rate": 0.0005139644418536869,
+      "loss": 3.4064,
+      "step": 24700
+    },
+    {
+      "epoch": 7.2094500116522955,
+      "grad_norm": 0.31195658445358276,
+      "learning_rate": 0.0005137895657242786,
+      "loss": 3.4211,
+      "step": 24750
+    },
+    {
+      "epoch": 7.224015381030063,
+      "grad_norm": 0.3211931586265564,
+      "learning_rate": 0.0005136146895948703,
+      "loss": 3.4182,
+      "step": 24800
+    },
+    {
+      "epoch": 7.238580750407831,
+      "grad_norm": 0.31733015179634094,
+      "learning_rate": 0.000513439813465462,
+      "loss": 3.4133,
+      "step": 24850
+    },
+    {
+      "epoch": 7.2531461197855975,
+      "grad_norm": 0.3402824401855469,
+      "learning_rate": 0.0005132649373360535,
+      "loss": 3.4146,
+      "step": 24900
+    },
+    {
+      "epoch": 7.267711489163365,
+      "grad_norm": 0.33483394980430603,
+      "learning_rate": 0.0005130900612066452,
+      "loss": 3.4263,
+      "step": 24950
+    },
+    {
+      "epoch": 7.282276858541133,
+      "grad_norm": 0.3118570148944855,
+      "learning_rate": 0.0005129151850772369,
+      "loss": 3.4278,
+      "step": 25000
+    },
+    {
+      "epoch": 7.282276858541133,
+      "eval_accuracy": 0.36476655458407903,
+      "eval_loss": 3.599691390991211,
+      "eval_runtime": 180.9668,
+      "eval_samples_per_second": 91.956,
+      "eval_steps_per_second": 5.752,
+      "step": 25000
+    },
+    {
+      "epoch": 7.2968422279189,
+      "grad_norm": 0.3210265636444092,
+      "learning_rate": 0.0005127403089478286,
+      "loss": 3.4248,
+      "step": 25050
+    },
+    {
+      "epoch": 7.311407597296667,
+      "grad_norm": 0.3390471041202545,
+      "learning_rate": 0.0005125654328184203,
+      "loss": 3.4295,
+      "step": 25100
+    },
+    {
+      "epoch": 7.325972966674435,
+      "grad_norm": 0.3226793706417084,
+      "learning_rate": 0.0005123905566890119,
+      "loss": 3.4171,
+      "step": 25150
+    },
+    {
+      "epoch": 7.3405383360522025,
+      "grad_norm": 0.3274274170398712,
+      "learning_rate": 0.0005122156805596036,
+      "loss": 3.4351,
+      "step": 25200
+    },
+    {
+      "epoch": 7.35510370542997,
+      "grad_norm": 0.3180485665798187,
+      "learning_rate": 0.0005120408044301953,
+      "loss": 3.4277,
+      "step": 25250
+    },
+    {
+      "epoch": 7.369669074807737,
+      "grad_norm": 0.32034561038017273,
+      "learning_rate": 0.0005118659283007868,
+      "loss": 3.4297,
+      "step": 25300
+    },
+    {
+      "epoch": 7.384234444185505,
+      "grad_norm": 0.3191639184951782,
+      "learning_rate": 0.0005116910521713785,
+      "loss": 3.4395,
+      "step": 25350
+    },
+    {
+      "epoch": 7.398799813563272,
+      "grad_norm": 0.3250245749950409,
+      "learning_rate": 0.0005115161760419702,
+      "loss": 3.4386,
+      "step": 25400
+    },
+    {
+      "epoch": 7.413365182941039,
+      "grad_norm": 0.31783708930015564,
+      "learning_rate": 0.0005113412999125619,
+      "loss": 3.4402,
+      "step": 25450
+    },
+    {
+      "epoch": 7.427930552318807,
+      "grad_norm": 0.3078354299068451,
+      "learning_rate": 0.0005111664237831536,
+      "loss": 3.4504,
+      "step": 25500
+    },
+    {
+      "epoch": 7.442495921696574,
+      "grad_norm": 0.3199695646762848,
+      "learning_rate": 0.0005109915476537452,
+      "loss": 3.4331,
+      "step": 25550
+    },
+    {
+      "epoch": 7.457061291074342,
+      "grad_norm": 0.33788904547691345,
+      "learning_rate": 0.0005108166715243369,
+      "loss": 3.4517,
+      "step": 25600
+    },
+    {
+      "epoch": 7.471626660452109,
+      "grad_norm": 0.3061235845088959,
+      "learning_rate": 0.0005106417953949286,
+      "loss": 3.4403,
+      "step": 25650
+    },
+    {
+      "epoch": 7.486192029829876,
+      "grad_norm": 0.30813688039779663,
+      "learning_rate": 0.0005104669192655203,
+      "loss": 3.4501,
+      "step": 25700
+    },
+    {
+      "epoch": 7.500757399207644,
+      "grad_norm": 0.32963570952415466,
+      "learning_rate": 0.0005102920431361118,
+      "loss": 3.4319,
+      "step": 25750
+    },
+    {
+      "epoch": 7.515322768585412,
+      "grad_norm": 0.3357667624950409,
+      "learning_rate": 0.0005101171670067035,
+      "loss": 3.4368,
+      "step": 25800
+    },
+    {
+      "epoch": 7.529888137963178,
+      "grad_norm": 0.3124500513076782,
+      "learning_rate": 0.0005099422908772952,
+      "loss": 3.444,
+      "step": 25850
+    },
+    {
+      "epoch": 7.544453507340946,
+      "grad_norm": 0.3235322833061218,
+      "learning_rate": 0.0005097674147478868,
+      "loss": 3.4464,
+      "step": 25900
+    },
+    {
+      "epoch": 7.559018876718714,
+      "grad_norm": 0.33699706196784973,
+      "learning_rate": 0.0005095925386184786,
+      "loss": 3.4557,
+      "step": 25950
+    },
+    {
+      "epoch": 7.573584246096481,
+      "grad_norm": 0.33560916781425476,
+      "learning_rate": 0.0005094176624890702,
+      "loss": 3.444,
+      "step": 26000
+    },
+    {
+      "epoch": 7.573584246096481,
+      "eval_accuracy": 0.36543262926276326,
+      "eval_loss": 3.588552713394165,
+      "eval_runtime": 181.0214,
+      "eval_samples_per_second": 91.928,
+      "eval_steps_per_second": 5.751,
+      "step": 26000
+    },
+    {
+      "epoch": 7.588149615474248,
+      "grad_norm": 0.3573473393917084,
+      "learning_rate": 0.0005092427863596619,
+      "loss": 3.4578,
+      "step": 26050
+    },
+    {
+      "epoch": 7.602714984852016,
+      "grad_norm": 0.30736207962036133,
+      "learning_rate": 0.0005090679102302536,
+      "loss": 3.4607,
+      "step": 26100
+    },
+    {
+      "epoch": 7.617280354229783,
+      "grad_norm": 0.3256712257862091,
+      "learning_rate": 0.0005088930341008451,
+      "loss": 3.4547,
+      "step": 26150
+    },
+    {
+      "epoch": 7.631845723607551,
+      "grad_norm": 0.31736353039741516,
+      "learning_rate": 0.0005087181579714368,
+      "loss": 3.446,
+      "step": 26200
+    },
+    {
+      "epoch": 7.646411092985318,
+      "grad_norm": 0.3086666464805603,
+      "learning_rate": 0.0005085432818420285,
+      "loss": 3.45,
+      "step": 26250
+    },
+    {
+      "epoch": 7.660976462363085,
+      "grad_norm": 0.3179473876953125,
+      "learning_rate": 0.0005083684057126202,
+      "loss": 3.4488,
+      "step": 26300
+    },
+    {
+      "epoch": 7.675541831740853,
+      "grad_norm": 0.32129475474357605,
+      "learning_rate": 0.0005081935295832118,
+      "loss": 3.4415,
+      "step": 26350
+    },
+    {
+      "epoch": 7.690107201118621,
+      "grad_norm": 0.31210270524024963,
+      "learning_rate": 0.0005080186534538035,
+      "loss": 3.4555,
+      "step": 26400
+    },
+    {
+      "epoch": 7.704672570496387,
+      "grad_norm": 0.32844117283821106,
+      "learning_rate": 0.0005078437773243952,
+      "loss": 3.4485,
+      "step": 26450
+    },
+    {
+      "epoch": 7.719237939874155,
+      "grad_norm": 0.32630887627601624,
+      "learning_rate": 0.0005076689011949869,
+      "loss": 3.444,
+      "step": 26500
+    },
+    {
+      "epoch": 7.733803309251923,
+      "grad_norm": 0.3128376603126526,
+      "learning_rate": 0.0005074940250655786,
+      "loss": 3.4498,
+      "step": 26550
+    },
+    {
+      "epoch": 7.74836867862969,
+      "grad_norm": 0.3320539891719818,
+      "learning_rate": 0.0005073191489361701,
+      "loss": 3.459,
+      "step": 26600
+    },
+    {
+      "epoch": 7.762934048007457,
+      "grad_norm": 0.3323521018028259,
+      "learning_rate": 0.0005071442728067618,
+      "loss": 3.4609,
+      "step": 26650
+    },
+    {
+      "epoch": 7.777499417385225,
+      "grad_norm": 0.3059885501861572,
+      "learning_rate": 0.0005069693966773535,
+      "loss": 3.4545,
+      "step": 26700
+    },
+    {
+      "epoch": 7.792064786762992,
+      "grad_norm": 0.31743645668029785,
+      "learning_rate": 0.0005067945205479451,
+      "loss": 3.4472,
+      "step": 26750
+    },
+    {
+      "epoch": 7.80663015614076,
+      "grad_norm": 0.3409293591976166,
+      "learning_rate": 0.0005066196444185368,
+      "loss": 3.4691,
+      "step": 26800
+    },
+    {
+      "epoch": 7.821195525518527,
+      "grad_norm": 0.3024209141731262,
+      "learning_rate": 0.0005064447682891285,
+      "loss": 3.4668,
+      "step": 26850
+    },
+    {
+      "epoch": 7.8357608948962945,
+      "grad_norm": 0.3276432752609253,
+      "learning_rate": 0.0005062698921597202,
+      "loss": 3.4733,
+      "step": 26900
+    },
+    {
+      "epoch": 7.850326264274062,
+      "grad_norm": 0.32540908455848694,
+      "learning_rate": 0.0005060950160303119,
+      "loss": 3.4538,
+      "step": 26950
+    },
+    {
+      "epoch": 7.86489163365183,
+      "grad_norm": 0.32031726837158203,
+      "learning_rate": 0.0005059201399009035,
+      "loss": 3.4694,
+      "step": 27000
+    },
+    {
+      "epoch": 7.86489163365183,
+      "eval_accuracy": 0.3658355197728572,
+      "eval_loss": 3.581731081008911,
+      "eval_runtime": 180.9719,
+      "eval_samples_per_second": 91.954,
+      "eval_steps_per_second": 5.752,
+      "step": 27000
+    },
+    {
+      "epoch": 7.8794570030295965,
+      "grad_norm": 0.32334402203559875,
+      "learning_rate": 0.0005057452637714951,
+      "loss": 3.4503,
+      "step": 27050
+    },
+    {
+      "epoch": 7.894022372407364,
+      "grad_norm": 0.3104668855667114,
+      "learning_rate": 0.0005055703876420868,
+      "loss": 3.4598,
+      "step": 27100
+    },
+    {
+      "epoch": 7.908587741785132,
+      "grad_norm": 0.32057490944862366,
+      "learning_rate": 0.0005053955115126785,
+      "loss": 3.4528,
+      "step": 27150
+    },
+    {
+      "epoch": 7.923153111162899,
+      "grad_norm": 0.32466772198677063,
+      "learning_rate": 0.0005052206353832701,
+      "loss": 3.4712,
+      "step": 27200
+    },
+    {
+      "epoch": 7.937718480540666,
+      "grad_norm": 0.31811457872390747,
+      "learning_rate": 0.0005050457592538618,
+      "loss": 3.4534,
+      "step": 27250
+    },
+    {
+      "epoch": 7.952283849918434,
+      "grad_norm": 0.322372168302536,
+      "learning_rate": 0.0005048708831244535,
+      "loss": 3.4765,
+      "step": 27300
+    },
+    {
+      "epoch": 7.9668492192962015,
+      "grad_norm": 0.31595420837402344,
+      "learning_rate": 0.0005046960069950451,
+      "loss": 3.4604,
+      "step": 27350
+    },
+    {
+      "epoch": 7.981414588673969,
+      "grad_norm": 0.3345142602920532,
+      "learning_rate": 0.0005045211308656369,
+      "loss": 3.4715,
+      "step": 27400
+    },
+    {
+      "epoch": 7.995979958051736,
+      "grad_norm": 0.31398117542266846,
+      "learning_rate": 0.0005043462547362284,
+      "loss": 3.4658,
+      "step": 27450
+    },
+    {
+      "epoch": 8.010487065951992,
+      "grad_norm": 0.3555661737918854,
+      "learning_rate": 0.0005041713786068201,
+      "loss": 3.3733,
+      "step": 27500
+    },
+    {
+      "epoch": 8.02505243532976,
+      "grad_norm": 0.33493924140930176,
+      "learning_rate": 0.0005039965024774118,
+      "loss": 3.3483,
+      "step": 27550
+    },
+    {
+      "epoch": 8.039617804707527,
+      "grad_norm": 0.30841153860092163,
+      "learning_rate": 0.0005038216263480034,
+      "loss": 3.3532,
+      "step": 27600
+    },
+    {
+      "epoch": 8.054183174085296,
+      "grad_norm": 0.32531553506851196,
+      "learning_rate": 0.0005036467502185951,
+      "loss": 3.3533,
+      "step": 27650
+    },
+    {
+      "epoch": 8.068748543463062,
+      "grad_norm": 0.30850234627723694,
+      "learning_rate": 0.0005034718740891868,
+      "loss": 3.3675,
+      "step": 27700
+    },
+    {
+      "epoch": 8.08331391284083,
+      "grad_norm": 0.3094310164451599,
+      "learning_rate": 0.0005032969979597785,
+      "loss": 3.3712,
+      "step": 27750
+    },
+    {
+      "epoch": 8.097879282218598,
+      "grad_norm": 0.3374312222003937,
+      "learning_rate": 0.0005031221218303701,
+      "loss": 3.3693,
+      "step": 27800
+    },
+    {
+      "epoch": 8.112444651596364,
+      "grad_norm": 0.36242911219596863,
+      "learning_rate": 0.0005029472457009618,
+      "loss": 3.3688,
+      "step": 27850
+    },
+    {
+      "epoch": 8.127010020974131,
+      "grad_norm": 0.33885157108306885,
+      "learning_rate": 0.0005027723695715534,
+      "loss": 3.3775,
+      "step": 27900
+    },
+    {
+      "epoch": 8.1415753903519,
+      "grad_norm": 0.3156500458717346,
+      "learning_rate": 0.0005025974934421451,
+      "loss": 3.3714,
+      "step": 27950
+    },
+    {
+      "epoch": 8.156140759729666,
+      "grad_norm": 0.31954655051231384,
+      "learning_rate": 0.0005024226173127368,
+      "loss": 3.3704,
+      "step": 28000
+    },
+    {
+      "epoch": 8.156140759729666,
+      "eval_accuracy": 0.36593935874554057,
+      "eval_loss": 3.5886237621307373,
+      "eval_runtime": 181.022,
+      "eval_samples_per_second": 91.928,
+      "eval_steps_per_second": 5.751,
+      "step": 28000
+    },
+    {
+      "epoch": 8.170706129107435,
+      "grad_norm": 0.32791200280189514,
+      "learning_rate": 0.0005022477411833284,
+      "loss": 3.3801,
+      "step": 28050
+    },
+    {
+      "epoch": 8.185271498485202,
+      "grad_norm": 0.35599347949028015,
+      "learning_rate": 0.0005020728650539201,
+      "loss": 3.389,
+      "step": 28100
+    },
+    {
+      "epoch": 8.199836867862969,
+      "grad_norm": 0.3596663475036621,
+      "learning_rate": 0.0005018979889245118,
+      "loss": 3.3856,
+      "step": 28150
+    },
+    {
+      "epoch": 8.214402237240737,
+      "grad_norm": 0.3202419579029083,
+      "learning_rate": 0.0005017231127951034,
+      "loss": 3.3962,
+      "step": 28200
+    },
+    {
+      "epoch": 8.228967606618504,
+      "grad_norm": 0.3104107975959778,
+      "learning_rate": 0.0005015482366656951,
+      "loss": 3.4003,
+      "step": 28250
+    },
+    {
+      "epoch": 8.24353297599627,
+      "grad_norm": 0.32614579796791077,
+      "learning_rate": 0.0005013733605362868,
+      "loss": 3.4007,
+      "step": 28300
+    },
+    {
+      "epoch": 8.258098345374039,
+      "grad_norm": 0.34570226073265076,
+      "learning_rate": 0.0005011984844068784,
+      "loss": 3.4056,
+      "step": 28350
+    },
+    {
+      "epoch": 8.272663714751806,
+      "grad_norm": 0.3463624119758606,
+      "learning_rate": 0.0005010236082774701,
+      "loss": 3.3834,
+      "step": 28400
+    },
+    {
+      "epoch": 8.287229084129574,
+      "grad_norm": 0.32712987065315247,
+      "learning_rate": 0.0005008487321480617,
+      "loss": 3.3932,
+      "step": 28450
+    },
+    {
+      "epoch": 8.301794453507341,
+      "grad_norm": 0.32646724581718445,
+      "learning_rate": 0.0005006738560186534,
+      "loss": 3.3973,
+      "step": 28500
+    },
+    {
+      "epoch": 8.316359822885108,
+      "grad_norm": 0.35286417603492737,
+      "learning_rate": 0.0005004989798892451,
+      "loss": 3.3867,
+      "step": 28550
+    },
+    {
+      "epoch": 8.330925192262876,
+      "grad_norm": 0.3331160545349121,
+      "learning_rate": 0.0005003241037598368,
+      "loss": 3.3967,
+      "step": 28600
+    },
+    {
+      "epoch": 8.345490561640643,
+      "grad_norm": 0.32697588205337524,
+      "learning_rate": 0.0005001492276304284,
+      "loss": 3.399,
+      "step": 28650
+    },
+    {
+      "epoch": 8.36005593101841,
+      "grad_norm": 0.31661832332611084,
+      "learning_rate": 0.0004999743515010201,
+      "loss": 3.3939,
+      "step": 28700
+    },
+    {
+      "epoch": 8.374621300396178,
+      "grad_norm": 0.33573973178863525,
+      "learning_rate": 0.0004997994753716117,
+      "loss": 3.415,
+      "step": 28750
+    },
+    {
+      "epoch": 8.389186669773945,
+      "grad_norm": 0.3202578127384186,
+      "learning_rate": 0.0004996245992422033,
+      "loss": 3.4068,
+      "step": 28800
+    },
+    {
+      "epoch": 8.403752039151712,
+      "grad_norm": 0.3525102436542511,
+      "learning_rate": 0.0004994497231127951,
+      "loss": 3.4111,
+      "step": 28850
+    },
+    {
+      "epoch": 8.41831740852948,
+      "grad_norm": 0.3182973265647888,
+      "learning_rate": 0.0004992748469833867,
+      "loss": 3.3995,
+      "step": 28900
+    },
+    {
+      "epoch": 8.432882777907247,
+      "grad_norm": 0.32942458987236023,
+      "learning_rate": 0.0004990999708539784,
+      "loss": 3.4205,
+      "step": 28950
+    },
+    {
+      "epoch": 8.447448147285016,
+      "grad_norm": 0.3333584666252136,
+      "learning_rate": 0.0004989250947245701,
+      "loss": 3.4165,
+      "step": 29000
+    },
+    {
+      "epoch": 8.447448147285016,
+      "eval_accuracy": 0.36646419830962385,
+      "eval_loss": 3.583397626876831,
+      "eval_runtime": 181.0212,
+      "eval_samples_per_second": 91.928,
+      "eval_steps_per_second": 5.751,
+      "step": 29000
+    },
+    {
+      "epoch": 8.462013516662783,
+      "grad_norm": 0.3410201072692871,
+      "learning_rate": 0.0004987502185951617,
+      "loss": 3.4048,
+      "step": 29050
+    },
+    {
+      "epoch": 8.47657888604055,
+      "grad_norm": 0.3245595395565033,
+      "learning_rate": 0.0004985753424657534,
+      "loss": 3.4134,
+      "step": 29100
+    },
+    {
+      "epoch": 8.491144255418318,
+      "grad_norm": 0.329243004322052,
+      "learning_rate": 0.000498400466336345,
+      "loss": 3.4072,
+      "step": 29150
+    },
+    {
+      "epoch": 8.505709624796085,
+      "grad_norm": 0.3384111225605011,
+      "learning_rate": 0.0004982255902069367,
+      "loss": 3.4097,
+      "step": 29200
+    },
+    {
+      "epoch": 8.520274994173853,
+      "grad_norm": 0.33028554916381836,
+      "learning_rate": 0.0004980507140775283,
+      "loss": 3.4127,
+      "step": 29250
+    },
+    {
+      "epoch": 8.53484036355162,
+      "grad_norm": 0.33368030190467834,
+      "learning_rate": 0.0004978758379481201,
+      "loss": 3.4092,
+      "step": 29300
+    },
+    {
+      "epoch": 8.549405732929387,
+      "grad_norm": 0.3354744613170624,
+      "learning_rate": 0.0004977009618187117,
+      "loss": 3.4099,
+      "step": 29350
+    },
+    {
+      "epoch": 8.563971102307155,
+      "grad_norm": 0.333848237991333,
+      "learning_rate": 0.0004975260856893034,
+      "loss": 3.44,
+      "step": 29400
+    },
+    {
+      "epoch": 8.578536471684922,
+      "grad_norm": 0.3305192291736603,
+      "learning_rate": 0.0004973512095598951,
+      "loss": 3.4258,
+      "step": 29450
+    },
+    {
+      "epoch": 8.593101841062689,
+      "grad_norm": 0.30741816759109497,
+      "learning_rate": 0.0004971763334304867,
+      "loss": 3.4281,
+      "step": 29500
+    },
+    {
+      "epoch": 8.607667210440457,
+      "grad_norm": 0.3138411343097687,
+      "learning_rate": 0.0004970014573010784,
+      "loss": 3.4276,
+      "step": 29550
+    },
+    {
+      "epoch": 8.622232579818224,
+      "grad_norm": 0.3112436532974243,
+      "learning_rate": 0.00049682658117167,
+      "loss": 3.4166,
+      "step": 29600
+    },
+    {
+      "epoch": 8.63679794919599,
+      "grad_norm": 0.3153873085975647,
+      "learning_rate": 0.0004966517050422616,
+      "loss": 3.4367,
+      "step": 29650
+    },
+    {
+      "epoch": 8.65136331857376,
+      "grad_norm": 0.3235262334346771,
+      "learning_rate": 0.0004964768289128533,
+      "loss": 3.419,
+      "step": 29700
+    },
+    {
+      "epoch": 8.665928687951526,
+      "grad_norm": 0.3494694232940674,
+      "learning_rate": 0.000496301952783445,
+      "loss": 3.4267,
+      "step": 29750
+    },
+    {
+      "epoch": 8.680494057329295,
+      "grad_norm": 0.3496847450733185,
+      "learning_rate": 0.0004961270766540367,
+      "loss": 3.4166,
+      "step": 29800
+    },
+    {
+      "epoch": 8.695059426707061,
+      "grad_norm": 0.32888874411582947,
+      "learning_rate": 0.0004959522005246284,
+      "loss": 3.4266,
+      "step": 29850
+    },
+    {
+      "epoch": 8.709624796084828,
+      "grad_norm": 0.32648685574531555,
+      "learning_rate": 0.00049577732439522,
+      "loss": 3.4292,
+      "step": 29900
+    },
+    {
+      "epoch": 8.724190165462597,
+      "grad_norm": 0.3119540214538574,
+      "learning_rate": 0.0004956024482658117,
+      "loss": 3.4145,
+      "step": 29950
+    },
+    {
+      "epoch": 8.738755534840363,
+      "grad_norm": 0.3203805983066559,
+      "learning_rate": 0.0004954275721364034,
+      "loss": 3.4266,
+      "step": 30000
+    },
+    {
+      "epoch": 8.738755534840363,
+      "eval_accuracy": 0.3671463839047946,
+      "eval_loss": 3.5733180046081543,
+      "eval_runtime": 187.0901,
+      "eval_samples_per_second": 88.946,
+      "eval_steps_per_second": 5.564,
+      "step": 30000
+    },
+    {
+      "epoch": 8.753320904218132,
+      "grad_norm": 0.30910488963127136,
+      "learning_rate": 0.000495252696006995,
+      "loss": 3.4274,
+      "step": 30050
+    },
+    {
+      "epoch": 8.767886273595899,
+      "grad_norm": 0.33857056498527527,
+      "learning_rate": 0.0004950778198775866,
+      "loss": 3.4322,
+      "step": 30100
+    },
+    {
+      "epoch": 8.782451642973665,
+      "grad_norm": 0.34136781096458435,
+      "learning_rate": 0.0004949029437481783,
+      "loss": 3.4361,
+      "step": 30150
+    },
+    {
+      "epoch": 8.797017012351434,
+      "grad_norm": 0.3273514211177826,
+      "learning_rate": 0.00049472806761877,
+      "loss": 3.4242,
+      "step": 30200
+    },
+    {
+      "epoch": 8.8115823817292,
+      "grad_norm": 0.31845390796661377,
+      "learning_rate": 0.0004945531914893616,
+      "loss": 3.4355,
+      "step": 30250
+    },
+    {
+      "epoch": 8.826147751106967,
+      "grad_norm": 0.3300946354866028,
+      "learning_rate": 0.0004943783153599534,
+      "loss": 3.4208,
+      "step": 30300
+    },
+    {
+      "epoch": 8.840713120484736,
+      "grad_norm": 0.30375778675079346,
+      "learning_rate": 0.000494203439230545,
+      "loss": 3.432,
+      "step": 30350
+    },
+    {
+      "epoch": 8.855278489862503,
+      "grad_norm": 0.3202904760837555,
+      "learning_rate": 0.0004940285631011367,
+      "loss": 3.4407,
+      "step": 30400
+    },
+    {
+      "epoch": 8.86984385924027,
+      "grad_norm": 0.35564732551574707,
+      "learning_rate": 0.0004938536869717284,
+      "loss": 3.4362,
+      "step": 30450
+    },
+    {
+      "epoch": 8.884409228618038,
+      "grad_norm": 0.34171804785728455,
+      "learning_rate": 0.0004936788108423199,
+      "loss": 3.4227,
+      "step": 30500
+    },
+    {
+      "epoch": 8.898974597995805,
+      "grad_norm": 0.32788291573524475,
+      "learning_rate": 0.0004935039347129116,
+      "loss": 3.4332,
+      "step": 30550
+    },
+    {
+      "epoch": 8.913539967373573,
+      "grad_norm": 0.33452388644218445,
+      "learning_rate": 0.0004933290585835033,
+      "loss": 3.4312,
+      "step": 30600
+    },
+    {
+      "epoch": 8.92810533675134,
+      "grad_norm": 0.30564260482788086,
+      "learning_rate": 0.000493154182454095,
+      "loss": 3.4444,
+      "step": 30650
+    },
+    {
+      "epoch": 8.942670706129107,
+      "grad_norm": 0.31938478350639343,
+      "learning_rate": 0.0004929793063246866,
+      "loss": 3.4333,
+      "step": 30700
+    },
+    {
+      "epoch": 8.957236075506875,
+      "grad_norm": 0.3310040533542633,
+      "learning_rate": 0.0004928044301952783,
+      "loss": 3.4342,
+      "step": 30750
+    },
+    {
+      "epoch": 8.971801444884642,
+      "grad_norm": 0.32065731287002563,
+      "learning_rate": 0.00049262955406587,
+      "loss": 3.4474,
+      "step": 30800
+    },
+    {
+      "epoch": 8.986366814262409,
+      "grad_norm": 0.3287248909473419,
+      "learning_rate": 0.0004924546779364617,
+      "loss": 3.441,
+      "step": 30850
+    },
+    {
+      "epoch": 9.000873922162667,
+      "grad_norm": 0.3414923846721649,
+      "learning_rate": 0.0004922798018070533,
+      "loss": 3.4264,
+      "step": 30900
+    },
+    {
+      "epoch": 9.015439291540433,
+      "grad_norm": 0.331150621175766,
+      "learning_rate": 0.0004921049256776449,
+      "loss": 3.3187,
+      "step": 30950
+    },
+    {
+      "epoch": 9.0300046609182,
+      "grad_norm": 0.30203065276145935,
+      "learning_rate": 0.0004919300495482366,
+      "loss": 3.3195,
+      "step": 31000
+    },
+    {
+      "epoch": 9.0300046609182,
+      "eval_accuracy": 0.36716261241921166,
+      "eval_loss": 3.5813663005828857,
+      "eval_runtime": 186.8171,
+      "eval_samples_per_second": 89.076,
+      "eval_steps_per_second": 5.572,
+      "step": 31000
+    },
+    {
+      "epoch": 9.044570030295969,
+      "grad_norm": 0.3459174931049347,
+      "learning_rate": 0.0004917551734188283,
+      "loss": 3.3252,
+      "step": 31050
+    },
+    {
+      "epoch": 9.059135399673735,
+      "grad_norm": 0.31476983428001404,
+      "learning_rate": 0.0004915802972894199,
+      "loss": 3.3272,
+      "step": 31100
+    },
+    {
+      "epoch": 9.073700769051504,
+      "grad_norm": 0.3300425410270691,
+      "learning_rate": 0.0004914054211600116,
+      "loss": 3.3371,
+      "step": 31150
+    },
+    {
+      "epoch": 9.08826613842927,
+      "grad_norm": 0.3382238745689392,
+      "learning_rate": 0.0004912305450306033,
+      "loss": 3.3348,
+      "step": 31200
+    },
+    {
+      "epoch": 9.102831507807037,
+      "grad_norm": 0.3423096239566803,
+      "learning_rate": 0.000491055668901195,
+      "loss": 3.3325,
+      "step": 31250
+    },
+    {
+      "epoch": 9.117396877184806,
+      "grad_norm": 0.3285750448703766,
+      "learning_rate": 0.0004908807927717865,
+      "loss": 3.3515,
+      "step": 31300
+    },
+    {
+      "epoch": 9.131962246562573,
+      "grad_norm": 0.3302641212940216,
+      "learning_rate": 0.0004907059166423783,
+      "loss": 3.351,
+      "step": 31350
+    },
+    {
+      "epoch": 9.14652761594034,
+      "grad_norm": 0.3364979922771454,
+      "learning_rate": 0.0004905310405129699,
+      "loss": 3.3537,
+      "step": 31400
+    },
+    {
+      "epoch": 9.161092985318108,
+      "grad_norm": 0.3409692943096161,
+      "learning_rate": 0.0004903561643835616,
+      "loss": 3.3599,
+      "step": 31450
+    },
+    {
+      "epoch": 9.175658354695875,
+      "grad_norm": 0.3414802849292755,
+      "learning_rate": 0.0004901812882541533,
+      "loss": 3.3434,
+      "step": 31500
+    },
+    {
+      "epoch": 9.190223724073643,
+      "grad_norm": 0.36004215478897095,
+      "learning_rate": 0.0004900064121247449,
+      "loss": 3.3608,
+      "step": 31550
+    },
+    {
+      "epoch": 9.20478909345141,
+      "grad_norm": 0.35667118430137634,
+      "learning_rate": 0.0004898315359953366,
+      "loss": 3.3651,
+      "step": 31600
+    },
+    {
+      "epoch": 9.219354462829177,
+      "grad_norm": 0.3091524839401245,
+      "learning_rate": 0.0004896566598659283,
+      "loss": 3.3685,
+      "step": 31650
+    },
+    {
+      "epoch": 9.233919832206945,
+      "grad_norm": 0.35599398612976074,
+      "learning_rate": 0.0004894817837365199,
+      "loss": 3.38,
+      "step": 31700
+    },
+    {
+      "epoch": 9.248485201584712,
+      "grad_norm": 0.33103418350219727,
+      "learning_rate": 0.0004893069076071115,
+      "loss": 3.3702,
+      "step": 31750
+    },
+    {
+      "epoch": 9.263050570962479,
+      "grad_norm": 0.3387484550476074,
+      "learning_rate": 0.0004891320314777032,
+      "loss": 3.3711,
+      "step": 31800
+    },
+    {
+      "epoch": 9.277615940340247,
+      "grad_norm": 0.33295512199401855,
+      "learning_rate": 0.0004889571553482949,
+      "loss": 3.382,
+      "step": 31850
+    },
+    {
+      "epoch": 9.292181309718014,
+      "grad_norm": 0.34174227714538574,
+      "learning_rate": 0.0004887822792188866,
+      "loss": 3.3672,
+      "step": 31900
+    },
+    {
+      "epoch": 9.306746679095783,
+      "grad_norm": 0.3584904074668884,
+      "learning_rate": 0.0004886074030894782,
+      "loss": 3.3721,
+      "step": 31950
+    },
+    {
+      "epoch": 9.32131204847355,
+      "grad_norm": 0.3444700837135315,
+      "learning_rate": 0.0004884325269600699,
+      "loss": 3.3913,
+      "step": 32000
+    },
+    {
+      "epoch": 9.32131204847355,
+      "eval_accuracy": 0.36725492679470023,
+      "eval_loss": 3.576958179473877,
+      "eval_runtime": 186.7182,
+      "eval_samples_per_second": 89.124,
+      "eval_steps_per_second": 5.575,
+      "step": 32000
+    },
+    {
+      "epoch": 9.335877417851316,
+      "grad_norm": 0.340363085269928,
+      "learning_rate": 0.0004882576508306615,
+      "loss": 3.3624,
+      "step": 32050
+    },
+    {
+      "epoch": 9.350442787229085,
+      "grad_norm": 0.3756904602050781,
+      "learning_rate": 0.00048808277470125327,
+      "loss": 3.3805,
+      "step": 32100
+    },
+    {
+      "epoch": 9.365008156606851,
+      "grad_norm": 0.32874080538749695,
+      "learning_rate": 0.0004879078985718449,
+      "loss": 3.373,
+      "step": 32150
+    },
+    {
+      "epoch": 9.379573525984618,
+      "grad_norm": 0.32386621832847595,
+      "learning_rate": 0.0004877330224424366,
+      "loss": 3.3735,
+      "step": 32200
+    },
+    {
+      "epoch": 9.394138895362387,
+      "grad_norm": 0.33558595180511475,
+      "learning_rate": 0.00048755814631302823,
+      "loss": 3.3827,
+      "step": 32250
+    },
+    {
+      "epoch": 9.408704264740154,
+      "grad_norm": 0.33117812871932983,
+      "learning_rate": 0.00048738327018361987,
+      "loss": 3.3877,
+      "step": 32300
+    },
+    {
+      "epoch": 9.423269634117922,
+      "grad_norm": 0.3605954647064209,
+      "learning_rate": 0.00048720839405421156,
+      "loss": 3.3848,
+      "step": 32350
+    },
+    {
+      "epoch": 9.437835003495689,
+      "grad_norm": 0.31175658106803894,
+      "learning_rate": 0.0004870335179248032,
+      "loss": 3.3758,
+      "step": 32400
+    },
+    {
+      "epoch": 9.452400372873456,
+      "grad_norm": 0.3200511634349823,
+      "learning_rate": 0.0004868586417953949,
+      "loss": 3.3908,
+      "step": 32450
+    },
+    {
+      "epoch": 9.466965742251224,
+      "grad_norm": 0.335527241230011,
+      "learning_rate": 0.0004866837656659865,
+      "loss": 3.392,
+      "step": 32500
+    },
+    {
+      "epoch": 9.48153111162899,
+      "grad_norm": 0.3885400593280792,
+      "learning_rate": 0.00048650888953657816,
+      "loss": 3.3818,
+      "step": 32550
+    },
+    {
+      "epoch": 9.496096481006758,
+      "grad_norm": 0.33358171582221985,
+      "learning_rate": 0.0004863340134071699,
+      "loss": 3.3745,
+      "step": 32600
+    },
+    {
+      "epoch": 9.510661850384526,
+      "grad_norm": 0.3522608280181885,
+      "learning_rate": 0.00048615913727776154,
+      "loss": 3.3791,
+      "step": 32650
+    },
+    {
+      "epoch": 9.525227219762293,
+      "grad_norm": 0.35827428102493286,
+      "learning_rate": 0.00048598426114835323,
+      "loss": 3.3967,
+      "step": 32700
+    },
+    {
+      "epoch": 9.53979258914006,
+      "grad_norm": 0.30217501521110535,
+      "learning_rate": 0.00048580938501894486,
+      "loss": 3.393,
+      "step": 32750
+    },
+    {
+      "epoch": 9.554357958517828,
+      "grad_norm": 0.3405115306377411,
+      "learning_rate": 0.00048563450888953655,
+      "loss": 3.3937,
+      "step": 32800
+    },
+    {
+      "epoch": 9.568923327895595,
+      "grad_norm": 0.3223709464073181,
+      "learning_rate": 0.0004854596327601282,
+      "loss": 3.3931,
+      "step": 32850
+    },
+    {
+      "epoch": 9.583488697273363,
+      "grad_norm": 0.3424103856086731,
+      "learning_rate": 0.0004852847566307198,
+      "loss": 3.4077,
+      "step": 32900
+    },
+    {
+      "epoch": 9.59805406665113,
+      "grad_norm": 0.34686312079429626,
+      "learning_rate": 0.0004851098805013115,
+      "loss": 3.3998,
+      "step": 32950
+    },
+    {
+      "epoch": 9.612619436028897,
+      "grad_norm": 0.32871878147125244,
+      "learning_rate": 0.00048493500437190315,
+      "loss": 3.4035,
+      "step": 33000
+    },
+    {
+      "epoch": 9.612619436028897,
+      "eval_accuracy": 0.3678857220942169,
+      "eval_loss": 3.5705971717834473,
+      "eval_runtime": 186.8723,
+      "eval_samples_per_second": 89.05,
+      "eval_steps_per_second": 5.571,
+      "step": 33000
+    },
+    {
+      "epoch": 9.627184805406666,
+      "grad_norm": 0.32823511958122253,
+      "learning_rate": 0.0004847601282424949,
+      "loss": 3.3981,
+      "step": 33050
+    },
+    {
+      "epoch": 9.641750174784432,
+      "grad_norm": 0.35547196865081787,
+      "learning_rate": 0.00048458525211308653,
+      "loss": 3.3911,
+      "step": 33100
+    },
+    {
+      "epoch": 9.6563155441622,
+      "grad_norm": 0.34656205773353577,
+      "learning_rate": 0.00048441037598367817,
+      "loss": 3.3973,
+      "step": 33150
+    },
+    {
+      "epoch": 9.670880913539968,
+      "grad_norm": 0.34557628631591797,
+      "learning_rate": 0.00048423549985426986,
+      "loss": 3.3916,
+      "step": 33200
+    },
+    {
+      "epoch": 9.685446282917734,
+      "grad_norm": 0.34200358390808105,
+      "learning_rate": 0.0004840606237248615,
+      "loss": 3.4006,
+      "step": 33250
+    },
+    {
+      "epoch": 9.700011652295503,
+      "grad_norm": 0.34963956475257874,
+      "learning_rate": 0.0004838857475954532,
+      "loss": 3.4008,
+      "step": 33300
+    },
+    {
+      "epoch": 9.71457702167327,
+      "grad_norm": 0.3219369053840637,
+      "learning_rate": 0.0004837108714660448,
+      "loss": 3.4052,
+      "step": 33350
+    },
+    {
+      "epoch": 9.729142391051036,
+      "grad_norm": 0.34709078073501587,
+      "learning_rate": 0.0004835359953366365,
+      "loss": 3.4039,
+      "step": 33400
+    },
+    {
+      "epoch": 9.743707760428805,
+      "grad_norm": 0.34171485900878906,
+      "learning_rate": 0.00048336111920722815,
+      "loss": 3.4107,
+      "step": 33450
+    },
+    {
+      "epoch": 9.758273129806572,
+      "grad_norm": 0.345464825630188,
+      "learning_rate": 0.0004831862430778198,
+      "loss": 3.3978,
+      "step": 33500
+    },
+    {
+      "epoch": 9.772838499184338,
+      "grad_norm": 0.31442996859550476,
+      "learning_rate": 0.00048301136694841153,
+      "loss": 3.4097,
+      "step": 33550
+    },
+    {
+      "epoch": 9.787403868562107,
+      "grad_norm": 0.35765334963798523,
+      "learning_rate": 0.00048283649081900317,
+      "loss": 3.4129,
+      "step": 33600
+    },
+    {
+      "epoch": 9.801969237939874,
+      "grad_norm": 0.34802165627479553,
+      "learning_rate": 0.00048266161468959486,
+      "loss": 3.4094,
+      "step": 33650
+    },
+    {
+      "epoch": 9.816534607317642,
+      "grad_norm": 0.321868360042572,
+      "learning_rate": 0.0004824867385601865,
+      "loss": 3.4063,
+      "step": 33700
+    },
+    {
+      "epoch": 9.831099976695409,
+      "grad_norm": 0.3318175673484802,
+      "learning_rate": 0.00048231186243077813,
+      "loss": 3.4035,
+      "step": 33750
+    },
+    {
+      "epoch": 9.845665346073176,
+      "grad_norm": 0.31054285168647766,
+      "learning_rate": 0.0004821369863013698,
+      "loss": 3.4011,
+      "step": 33800
+    },
+    {
+      "epoch": 9.860230715450944,
+      "grad_norm": 0.33249276876449585,
+      "learning_rate": 0.00048196211017196146,
+      "loss": 3.4145,
+      "step": 33850
+    },
+    {
+      "epoch": 9.874796084828711,
+      "grad_norm": 0.3437790870666504,
+      "learning_rate": 0.00048178723404255315,
+      "loss": 3.405,
+      "step": 33900
+    },
+    {
+      "epoch": 9.88936145420648,
+      "grad_norm": 0.3448680639266968,
+      "learning_rate": 0.0004816123579131448,
+      "loss": 3.4269,
+      "step": 33950
+    },
+    {
+      "epoch": 9.903926823584246,
+      "grad_norm": 0.3296782970428467,
+      "learning_rate": 0.0004814374817837364,
+      "loss": 3.4133,
+      "step": 34000
+    },
+    {
+      "epoch": 9.903926823584246,
+      "eval_accuracy": 0.36833894451858995,
+      "eval_loss": 3.563152551651001,
+      "eval_runtime": 186.8296,
+      "eval_samples_per_second": 89.07,
+      "eval_steps_per_second": 5.572,
+      "step": 34000
+    },
+    {
+      "epoch": 9.918492192962013,
+      "grad_norm": 0.3219934105873108,
+      "learning_rate": 0.00048126260565432816,
+      "loss": 3.4071,
+      "step": 34050
+    },
+    {
+      "epoch": 9.933057562339782,
+      "grad_norm": 0.34816163778305054,
+      "learning_rate": 0.0004810877295249198,
+      "loss": 3.39,
+      "step": 34100
+    },
+    {
+      "epoch": 9.947622931717548,
+      "grad_norm": 0.32388779520988464,
+      "learning_rate": 0.0004809128533955115,
+      "loss": 3.4089,
+      "step": 34150
+    },
+    {
+      "epoch": 9.962188301095315,
+      "grad_norm": 0.32343435287475586,
+      "learning_rate": 0.0004807379772661031,
+      "loss": 3.4061,
+      "step": 34200
+    },
+    {
+      "epoch": 9.976753670473084,
+      "grad_norm": 0.35213232040405273,
+      "learning_rate": 0.0004805631011366948,
+      "loss": 3.414,
+      "step": 34250
+    },
+    {
+      "epoch": 9.99131903985085,
+      "grad_norm": 0.3568871021270752,
+      "learning_rate": 0.00048038822500728645,
+      "loss": 3.4131,
+      "step": 34300
+    },
+    {
+      "epoch": 10.005826147751106,
+      "grad_norm": 0.3455922305583954,
+      "learning_rate": 0.0004802133488778781,
+      "loss": 3.3616,
+      "step": 34350
+    },
+    {
+      "epoch": 10.020391517128875,
+      "grad_norm": 0.3425789475440979,
+      "learning_rate": 0.0004800384727484698,
+      "loss": 3.3011,
+      "step": 34400
+    },
+    {
+      "epoch": 10.034956886506642,
+      "grad_norm": 0.32573023438453674,
+      "learning_rate": 0.0004798635966190614,
+      "loss": 3.2948,
+      "step": 34450
+    },
+    {
+      "epoch": 10.049522255884408,
+      "grad_norm": 0.358612984418869,
+      "learning_rate": 0.00047968872048965316,
+      "loss": 3.2988,
+      "step": 34500
+    },
+    {
+      "epoch": 10.064087625262177,
+      "grad_norm": 0.35645079612731934,
+      "learning_rate": 0.0004795138443602448,
+      "loss": 3.3224,
+      "step": 34550
+    },
+    {
+      "epoch": 10.078652994639944,
+      "grad_norm": 0.31410548090934753,
+      "learning_rate": 0.00047933896823083643,
+      "loss": 3.3127,
+      "step": 34600
+    },
+    {
+      "epoch": 10.093218364017712,
+      "grad_norm": 0.3214830756187439,
+      "learning_rate": 0.0004791640921014281,
+      "loss": 3.3221,
+      "step": 34650
+    },
+    {
+      "epoch": 10.107783733395479,
+      "grad_norm": 0.33103054761886597,
+      "learning_rate": 0.00047898921597201976,
+      "loss": 3.3177,
+      "step": 34700
+    },
+    {
+      "epoch": 10.122349102773246,
+      "grad_norm": 0.3237544000148773,
+      "learning_rate": 0.00047881433984261145,
+      "loss": 3.3272,
+      "step": 34750
+    },
+    {
+      "epoch": 10.136914472151014,
+      "grad_norm": 0.3658389747142792,
+      "learning_rate": 0.0004786394637132031,
+      "loss": 3.3313,
+      "step": 34800
+    },
+    {
+      "epoch": 10.151479841528781,
+      "grad_norm": 0.3435143530368805,
+      "learning_rate": 0.0004784645875837948,
+      "loss": 3.3187,
+      "step": 34850
+    },
+    {
+      "epoch": 10.166045210906548,
+      "grad_norm": 0.3427942097187042,
+      "learning_rate": 0.0004782897114543864,
+      "loss": 3.3263,
+      "step": 34900
+    },
+    {
+      "epoch": 10.180610580284316,
+      "grad_norm": 0.33091989159584045,
+      "learning_rate": 0.00047811483532497805,
+      "loss": 3.326,
+      "step": 34950
+    },
+    {
+      "epoch": 10.195175949662083,
+      "grad_norm": 0.3355826437473297,
+      "learning_rate": 0.0004779399591955698,
+      "loss": 3.3355,
+      "step": 35000
+    },
+    {
+      "epoch": 10.195175949662083,
+      "eval_accuracy": 0.36803942258945704,
+      "eval_loss": 3.576420307159424,
+      "eval_runtime": 186.8678,
+      "eval_samples_per_second": 89.052,
+      "eval_steps_per_second": 5.571,
+      "step": 35000
+    },
+    {
+      "epoch": 10.209741319039852,
+      "grad_norm": 0.3341238796710968,
+      "learning_rate": 0.00047776508306616143,
+      "loss": 3.3377,
+      "step": 35050
+    },
+    {
+      "epoch": 10.224306688417618,
+      "grad_norm": 0.3380385637283325,
+      "learning_rate": 0.0004775902069367531,
+      "loss": 3.3465,
+      "step": 35100
+    },
+    {
+      "epoch": 10.238872057795385,
+      "grad_norm": 0.3392700254917145,
+      "learning_rate": 0.00047741533080734476,
+      "loss": 3.3357,
+      "step": 35150
+    },
+    {
+      "epoch": 10.253437427173154,
+      "grad_norm": 0.3265944719314575,
+      "learning_rate": 0.0004772404546779364,
+      "loss": 3.3374,
+      "step": 35200
+    },
+    {
+      "epoch": 10.26800279655092,
+      "grad_norm": 0.3445483148097992,
+      "learning_rate": 0.0004770655785485281,
+      "loss": 3.3394,
+      "step": 35250
+    },
+    {
+      "epoch": 10.282568165928687,
+      "grad_norm": 0.34125420451164246,
+      "learning_rate": 0.0004768907024191197,
+      "loss": 3.3448,
+      "step": 35300
+    },
+    {
+      "epoch": 10.297133535306456,
+      "grad_norm": 0.32139939069747925,
+      "learning_rate": 0.0004767158262897114,
+      "loss": 3.3384,
+      "step": 35350
+    },
+    {
+      "epoch": 10.311698904684222,
+      "grad_norm": 0.3343341052532196,
+      "learning_rate": 0.00047654095016030305,
+      "loss": 3.35,
+      "step": 35400
+    },
+    {
+      "epoch": 10.326264274061991,
+      "grad_norm": 0.3511643409729004,
+      "learning_rate": 0.0004763660740308948,
+      "loss": 3.3704,
+      "step": 35450
+    },
+    {
+      "epoch": 10.340829643439758,
+      "grad_norm": 0.3447152376174927,
+      "learning_rate": 0.0004761911979014864,
+      "loss": 3.3533,
+      "step": 35500
+    },
+    {
+      "epoch": 10.355395012817525,
+      "grad_norm": 0.31971412897109985,
+      "learning_rate": 0.00047601632177207806,
+      "loss": 3.3617,
+      "step": 35550
+    },
+    {
+      "epoch": 10.369960382195293,
+      "grad_norm": 0.33184388279914856,
+      "learning_rate": 0.00047584144564266975,
+      "loss": 3.3606,
+      "step": 35600
+    },
+    {
+      "epoch": 10.38452575157306,
+      "grad_norm": 0.32978543639183044,
+      "learning_rate": 0.0004756665695132614,
+      "loss": 3.351,
+      "step": 35650
+    },
+    {
+      "epoch": 10.399091120950827,
+      "grad_norm": 0.3334619104862213,
+      "learning_rate": 0.0004754916933838531,
+      "loss": 3.3725,
+      "step": 35700
+    },
+    {
+      "epoch": 10.413656490328595,
+      "grad_norm": 0.33430349826812744,
+      "learning_rate": 0.0004753168172544447,
+      "loss": 3.3543,
+      "step": 35750
+    },
+    {
+      "epoch": 10.428221859706362,
+      "grad_norm": 0.34066829085350037,
+      "learning_rate": 0.00047514194112503635,
+      "loss": 3.363,
+      "step": 35800
+    },
+    {
+      "epoch": 10.44278722908413,
+      "grad_norm": 0.3592451214790344,
+      "learning_rate": 0.00047496706499562804,
+      "loss": 3.3558,
+      "step": 35850
+    },
+    {
+      "epoch": 10.457352598461897,
+      "grad_norm": 0.3279033601284027,
+      "learning_rate": 0.0004747921888662197,
+      "loss": 3.3752,
+      "step": 35900
+    },
+    {
+      "epoch": 10.471917967839664,
+      "grad_norm": 0.3477780520915985,
+      "learning_rate": 0.0004746173127368114,
+      "loss": 3.3651,
+      "step": 35950
+    },
+    {
+      "epoch": 10.486483337217432,
+      "grad_norm": 0.33582985401153564,
+      "learning_rate": 0.00047444243660740306,
+      "loss": 3.3782,
+      "step": 36000
+    },
+    {
+      "epoch": 10.486483337217432,
+      "eval_accuracy": 0.3685552071128873,
+      "eval_loss": 3.5681331157684326,
+      "eval_runtime": 186.617,
+      "eval_samples_per_second": 89.172,
+      "eval_steps_per_second": 5.578,
+      "step": 36000
+    },
+    {
+      "epoch": 10.5010487065952,
+      "grad_norm": 0.3560413122177124,
+      "learning_rate": 0.0004742675604779947,
+      "loss": 3.3591,
+      "step": 36050
+    },
+    {
+      "epoch": 10.515614075972966,
+      "grad_norm": 0.3362480700016022,
+      "learning_rate": 0.0004740926843485864,
+      "loss": 3.3706,
+      "step": 36100
+    },
+    {
+      "epoch": 10.530179445350734,
+      "grad_norm": 0.3344756066799164,
+      "learning_rate": 0.000473917808219178,
+      "loss": 3.3684,
+      "step": 36150
+    },
+    {
+      "epoch": 10.544744814728501,
+      "grad_norm": 0.3948496878147125,
+      "learning_rate": 0.0004737429320897697,
+      "loss": 3.3719,
+      "step": 36200
+    },
+    {
+      "epoch": 10.55931018410627,
+      "grad_norm": 0.3138646185398102,
+      "learning_rate": 0.00047356805596036135,
+      "loss": 3.37,
+      "step": 36250
+    },
+    {
+      "epoch": 10.573875553484037,
+      "grad_norm": 0.36229807138442993,
+      "learning_rate": 0.00047339317983095304,
+      "loss": 3.3733,
+      "step": 36300
+    },
+    {
+      "epoch": 10.588440922861803,
+      "grad_norm": 0.3171110451221466,
+      "learning_rate": 0.0004732183037015447,
+      "loss": 3.3716,
+      "step": 36350
+    },
+    {
+      "epoch": 10.603006292239572,
+      "grad_norm": 0.34809014201164246,
+      "learning_rate": 0.0004730434275721363,
+      "loss": 3.374,
+      "step": 36400
+    },
+    {
+      "epoch": 10.617571661617339,
+      "grad_norm": 0.35348230600357056,
+      "learning_rate": 0.00047286855144272806,
+      "loss": 3.3829,
+      "step": 36450
+    },
+    {
+      "epoch": 10.632137030995105,
+      "grad_norm": 0.3688938319683075,
+      "learning_rate": 0.0004726936753133197,
+      "loss": 3.3774,
+      "step": 36500
+    },
+    {
+      "epoch": 10.646702400372874,
+      "grad_norm": 0.3461929261684418,
+      "learning_rate": 0.0004725187991839114,
+      "loss": 3.3732,
+      "step": 36550
+    },
+    {
+      "epoch": 10.66126776975064,
+      "grad_norm": 0.3369111120700836,
+      "learning_rate": 0.000472343923054503,
+      "loss": 3.3673,
+      "step": 36600
+    },
+    {
+      "epoch": 10.675833139128407,
+      "grad_norm": 0.3469894826412201,
+      "learning_rate": 0.00047216904692509465,
+      "loss": 3.3842,
+      "step": 36650
+    },
+    {
+      "epoch": 10.690398508506176,
+      "grad_norm": 0.31455203890800476,
+      "learning_rate": 0.00047199417079568634,
+      "loss": 3.3849,
+      "step": 36700
+    },
+    {
+      "epoch": 10.704963877883943,
+      "grad_norm": 0.35142651200294495,
+      "learning_rate": 0.000471819294666278,
+      "loss": 3.3756,
+      "step": 36750
+    },
+    {
+      "epoch": 10.719529247261711,
+      "grad_norm": 0.3406135141849518,
+      "learning_rate": 0.00047164441853686967,
+      "loss": 3.3783,
+      "step": 36800
+    },
+    {
+      "epoch": 10.734094616639478,
+      "grad_norm": 0.33253002166748047,
+      "learning_rate": 0.0004714695424074613,
+      "loss": 3.3786,
+      "step": 36850
+    },
+    {
+      "epoch": 10.748659986017245,
+      "grad_norm": 0.33225905895233154,
+      "learning_rate": 0.00047129466627805305,
+      "loss": 3.3757,
+      "step": 36900
+    },
+    {
+      "epoch": 10.763225355395013,
+      "grad_norm": 0.33042529225349426,
+      "learning_rate": 0.0004711197901486447,
+      "loss": 3.375,
+      "step": 36950
+    },
+    {
+      "epoch": 10.77779072477278,
+      "grad_norm": 0.34693804383277893,
+      "learning_rate": 0.0004709449140192363,
+      "loss": 3.3796,
+      "step": 37000
+    },
+    {
+      "epoch": 10.77779072477278,
+      "eval_accuracy": 0.36912449869472175,
+      "eval_loss": 3.560312271118164,
+      "eval_runtime": 187.0056,
+      "eval_samples_per_second": 88.987,
+      "eval_steps_per_second": 5.567,
+      "step": 37000
+    },
+    {
+      "epoch": 10.792356094150549,
+      "grad_norm": 0.3276842534542084,
+      "learning_rate": 0.000470770037889828,
+      "loss": 3.3861,
+      "step": 37050
+    },
+    {
+      "epoch": 10.806921463528315,
+      "grad_norm": 0.3207048773765564,
+      "learning_rate": 0.00047059516176041965,
+      "loss": 3.3866,
+      "step": 37100
+    },
+    {
+      "epoch": 10.821486832906082,
+      "grad_norm": 0.3365728557109833,
+      "learning_rate": 0.00047042028563101134,
+      "loss": 3.3855,
+      "step": 37150
+    },
+    {
+      "epoch": 10.83605220228385,
+      "grad_norm": 0.3523632884025574,
+      "learning_rate": 0.000470245409501603,
+      "loss": 3.3849,
+      "step": 37200
+    },
+    {
+      "epoch": 10.850617571661617,
+      "grad_norm": 0.3342358469963074,
+      "learning_rate": 0.0004700705333721946,
+      "loss": 3.3786,
+      "step": 37250
+    },
+    {
+      "epoch": 10.865182941039384,
+      "grad_norm": 0.34308284521102905,
+      "learning_rate": 0.0004698956572427863,
+      "loss": 3.3877,
+      "step": 37300
+    },
+    {
+      "epoch": 10.879748310417153,
+      "grad_norm": 0.3864139914512634,
+      "learning_rate": 0.00046972078111337794,
+      "loss": 3.387,
+      "step": 37350
+    },
+    {
+      "epoch": 10.89431367979492,
+      "grad_norm": 0.340122789144516,
+      "learning_rate": 0.0004695459049839697,
+      "loss": 3.3867,
+      "step": 37400
+    },
+    {
+      "epoch": 10.908879049172686,
+      "grad_norm": 0.3356240391731262,
+      "learning_rate": 0.0004693710288545613,
+      "loss": 3.392,
+      "step": 37450
+    },
+    {
+      "epoch": 10.923444418550455,
+      "grad_norm": 0.3422701358795166,
+      "learning_rate": 0.000469196152725153,
+      "loss": 3.3981,
+      "step": 37500
+    },
+    {
+      "epoch": 10.938009787928221,
+      "grad_norm": 0.3323993682861328,
+      "learning_rate": 0.00046902127659574465,
+      "loss": 3.3845,
+      "step": 37550
+    },
+    {
+      "epoch": 10.95257515730599,
+      "grad_norm": 0.32822585105895996,
+      "learning_rate": 0.0004688464004663363,
+      "loss": 3.4054,
+      "step": 37600
+    },
+    {
+      "epoch": 10.967140526683757,
+      "grad_norm": 0.3223362863063812,
+      "learning_rate": 0.000468671524336928,
+      "loss": 3.3982,
+      "step": 37650
+    },
+    {
+      "epoch": 10.981705896061523,
+      "grad_norm": 0.347657710313797,
+      "learning_rate": 0.0004684966482075196,
+      "loss": 3.3923,
+      "step": 37700
+    },
+    {
+      "epoch": 10.996271265439292,
+      "grad_norm": 0.37350377440452576,
+      "learning_rate": 0.0004683217720781113,
+      "loss": 3.3975,
+      "step": 37750
+    },
+    {
+      "epoch": 11.010778373339548,
+      "grad_norm": 0.3476003110408783,
+      "learning_rate": 0.00046814689594870294,
+      "loss": 3.3105,
+      "step": 37800
+    },
+    {
+      "epoch": 11.025343742717315,
+      "grad_norm": 0.3537498712539673,
+      "learning_rate": 0.0004679720198192946,
+      "loss": 3.2683,
+      "step": 37850
+    },
+    {
+      "epoch": 11.039909112095083,
+      "grad_norm": 0.35762515664100647,
+      "learning_rate": 0.0004677971436898863,
+      "loss": 3.2801,
+      "step": 37900
+    },
+    {
+      "epoch": 11.05447448147285,
+      "grad_norm": 0.33562251925468445,
+      "learning_rate": 0.00046762226756047795,
+      "loss": 3.2888,
+      "step": 37950
+    },
+    {
+      "epoch": 11.069039850850617,
+      "grad_norm": 0.3734108805656433,
+      "learning_rate": 0.00046744739143106964,
+      "loss": 3.2878,
+      "step": 38000
+    },
+    {
+      "epoch": 11.069039850850617,
+      "eval_accuracy": 0.3688991810597714,
+      "eval_loss": 3.5698390007019043,
+      "eval_runtime": 187.2877,
+      "eval_samples_per_second": 88.853,
+      "eval_steps_per_second": 5.558,
+      "step": 38000
+    },
+    {
+      "epoch": 11.083605220228385,
+      "grad_norm": 0.3412722945213318,
+      "learning_rate": 0.0004672725153016613,
+      "loss": 3.2812,
+      "step": 38050
+    },
+    {
+      "epoch": 11.098170589606152,
+      "grad_norm": 0.332423597574234,
+      "learning_rate": 0.00046709763917225297,
+      "loss": 3.3108,
+      "step": 38100
+    },
+    {
+      "epoch": 11.11273595898392,
+      "grad_norm": 0.34238865971565247,
+      "learning_rate": 0.0004669227630428446,
+      "loss": 3.2887,
+      "step": 38150
+    },
+    {
+      "epoch": 11.127301328361687,
+      "grad_norm": 0.34268054366111755,
+      "learning_rate": 0.00046674788691343624,
+      "loss": 3.2998,
+      "step": 38200
+    },
+    {
+      "epoch": 11.141866697739454,
+      "grad_norm": 0.3180009722709656,
+      "learning_rate": 0.00046657301078402793,
+      "loss": 3.3092,
+      "step": 38250
+    },
+    {
+      "epoch": 11.156432067117223,
+      "grad_norm": 0.3360149562358856,
+      "learning_rate": 0.00046639813465461957,
+      "loss": 3.3097,
+      "step": 38300
+    },
+    {
+      "epoch": 11.17099743649499,
+      "grad_norm": 0.319097101688385,
+      "learning_rate": 0.0004662232585252113,
+      "loss": 3.3147,
+      "step": 38350
+    },
+    {
+      "epoch": 11.185562805872756,
+      "grad_norm": 0.3515799343585968,
+      "learning_rate": 0.00046604838239580295,
+      "loss": 3.3159,
+      "step": 38400
+    },
+    {
+      "epoch": 11.200128175250525,
+      "grad_norm": 0.3431311845779419,
+      "learning_rate": 0.0004658735062663946,
+      "loss": 3.3138,
+      "step": 38450
+    },
+    {
+      "epoch": 11.214693544628291,
+      "grad_norm": 0.3624405860900879,
+      "learning_rate": 0.0004656986301369863,
+      "loss": 3.3163,
+      "step": 38500
+    },
+    {
+      "epoch": 11.22925891400606,
+      "grad_norm": 0.34174567461013794,
+      "learning_rate": 0.0004655237540075779,
+      "loss": 3.3089,
+      "step": 38550
+    },
+    {
+      "epoch": 11.243824283383827,
+      "grad_norm": 0.35303977131843567,
+      "learning_rate": 0.0004653488778781696,
+      "loss": 3.3176,
+      "step": 38600
+    },
+    {
+      "epoch": 11.258389652761593,
+      "grad_norm": 0.3335185647010803,
+      "learning_rate": 0.00046517400174876124,
+      "loss": 3.3263,
+      "step": 38650
+    },
+    {
+      "epoch": 11.272955022139362,
+      "grad_norm": 0.3662583529949188,
+      "learning_rate": 0.0004649991256193529,
+      "loss": 3.3295,
+      "step": 38700
+    },
+    {
+      "epoch": 11.287520391517129,
+      "grad_norm": 0.361371785402298,
+      "learning_rate": 0.00046482424948994457,
+      "loss": 3.3321,
+      "step": 38750
+    },
+    {
+      "epoch": 11.302085760894895,
+      "grad_norm": 0.32191139459609985,
+      "learning_rate": 0.0004646493733605362,
+      "loss": 3.3366,
+      "step": 38800
+    },
+    {
+      "epoch": 11.316651130272664,
+      "grad_norm": 0.3527620732784271,
+      "learning_rate": 0.00046447449723112795,
+      "loss": 3.3228,
+      "step": 38850
+    },
+    {
+      "epoch": 11.33121649965043,
+      "grad_norm": 0.3786145746707916,
+      "learning_rate": 0.0004642996211017196,
+      "loss": 3.3236,
+      "step": 38900
+    },
+    {
+      "epoch": 11.3457818690282,
+      "grad_norm": 0.3184213638305664,
+      "learning_rate": 0.0004641247449723113,
+      "loss": 3.336,
+      "step": 38950
+    },
+    {
+      "epoch": 11.360347238405966,
+      "grad_norm": 0.38115012645721436,
+      "learning_rate": 0.0004639498688429029,
+      "loss": 3.3437,
+      "step": 39000
+    },
+    {
+      "epoch": 11.360347238405966,
+      "eval_accuracy": 0.36905405753431714,
+      "eval_loss": 3.567598342895508,
+      "eval_runtime": 186.745,
+      "eval_samples_per_second": 89.111,
+      "eval_steps_per_second": 5.574,
+      "step": 39000
+    },
+    {
+      "epoch": 11.374912607783733,
+      "grad_norm": 0.33753758668899536,
+      "learning_rate": 0.00046377499271349455,
+      "loss": 3.3427,
+      "step": 39050
+    },
+    {
+      "epoch": 11.389477977161501,
+      "grad_norm": 0.37911510467529297,
+      "learning_rate": 0.00046360011658408624,
+      "loss": 3.339,
+      "step": 39100
+    },
+    {
+      "epoch": 11.404043346539268,
+      "grad_norm": 0.34134548902511597,
+      "learning_rate": 0.00046342524045467787,
+      "loss": 3.3449,
+      "step": 39150
+    },
+    {
+      "epoch": 11.418608715917035,
+      "grad_norm": 0.3794786036014557,
+      "learning_rate": 0.00046325036432526956,
+      "loss": 3.3471,
+      "step": 39200
+    },
+    {
+      "epoch": 11.433174085294803,
+      "grad_norm": 0.36745432019233704,
+      "learning_rate": 0.0004630754881958612,
+      "loss": 3.3449,
+      "step": 39250
+    },
+    {
+      "epoch": 11.44773945467257,
+      "grad_norm": 0.34845367074012756,
+      "learning_rate": 0.00046290061206645284,
+      "loss": 3.3361,
+      "step": 39300
+    },
+    {
+      "epoch": 11.462304824050339,
+      "grad_norm": 0.33139854669570923,
+      "learning_rate": 0.0004627257359370446,
+      "loss": 3.3314,
+      "step": 39350
+    },
+    {
+      "epoch": 11.476870193428105,
+      "grad_norm": 0.342301607131958,
+      "learning_rate": 0.0004625508598076362,
+      "loss": 3.3476,
+      "step": 39400
+    },
+    {
+      "epoch": 11.491435562805872,
+      "grad_norm": 0.36012282967567444,
+      "learning_rate": 0.0004623759836782279,
+      "loss": 3.3472,
+      "step": 39450
+    },
+    {
+      "epoch": 11.50600093218364,
+      "grad_norm": 0.3556145131587982,
+      "learning_rate": 0.00046220110754881954,
+      "loss": 3.3567,
+      "step": 39500
+    },
+    {
+      "epoch": 11.520566301561407,
+      "grad_norm": 0.3342931270599365,
+      "learning_rate": 0.00046202623141941123,
+      "loss": 3.3395,
+      "step": 39550
+    },
+    {
+      "epoch": 11.535131670939174,
+      "grad_norm": 0.32155027985572815,
+      "learning_rate": 0.00046185135529000287,
+      "loss": 3.3591,
+      "step": 39600
+    },
+    {
+      "epoch": 11.549697040316943,
+      "grad_norm": 0.31837499141693115,
+      "learning_rate": 0.0004616764791605945,
+      "loss": 3.3535,
+      "step": 39650
+    },
+    {
+      "epoch": 11.56426240969471,
+      "grad_norm": 0.34017813205718994,
+      "learning_rate": 0.0004615016030311862,
+      "loss": 3.3599,
+      "step": 39700
+    },
+    {
+      "epoch": 11.578827779072478,
+      "grad_norm": 0.3498923182487488,
+      "learning_rate": 0.00046132672690177783,
+      "loss": 3.356,
+      "step": 39750
+    },
+    {
+      "epoch": 11.593393148450245,
+      "grad_norm": 0.33563312888145447,
+      "learning_rate": 0.0004611518507723696,
+      "loss": 3.3638,
+      "step": 39800
+    },
+    {
+      "epoch": 11.607958517828012,
+      "grad_norm": 0.3288760781288147,
+      "learning_rate": 0.0004609769746429612,
+      "loss": 3.3532,
+      "step": 39850
+    },
+    {
+      "epoch": 11.62252388720578,
+      "grad_norm": 0.3533109426498413,
+      "learning_rate": 0.00046080209851355285,
+      "loss": 3.3562,
+      "step": 39900
+    },
+    {
+      "epoch": 11.637089256583547,
+      "grad_norm": 0.33913081884384155,
+      "learning_rate": 0.00046062722238414454,
+      "loss": 3.3551,
+      "step": 39950
+    },
+    {
+      "epoch": 11.651654625961314,
+      "grad_norm": 0.33748742938041687,
+      "learning_rate": 0.0004604523462547362,
+      "loss": 3.3602,
+      "step": 40000
+    },
+    {
+      "epoch": 11.651654625961314,
+      "eval_accuracy": 0.3697631730555858,
+      "eval_loss": 3.556750535964966,
+      "eval_runtime": 187.7282,
+      "eval_samples_per_second": 88.644,
+      "eval_steps_per_second": 5.545,
+      "step": 40000
+    },
+    {
+      "epoch": 11.666219995339082,
+      "grad_norm": 0.3496758043766022,
+      "learning_rate": 0.00046027747012532787,
+      "loss": 3.3549,
+      "step": 40050
+    },
+    {
+      "epoch": 11.680785364716849,
+      "grad_norm": 0.3555567264556885,
+      "learning_rate": 0.0004601025939959195,
+      "loss": 3.3545,
+      "step": 40100
+    },
+    {
+      "epoch": 11.695350734094617,
+      "grad_norm": 0.33175376057624817,
+      "learning_rate": 0.0004599277178665112,
+      "loss": 3.3642,
+      "step": 40150
+    },
+    {
+      "epoch": 11.709916103472384,
+      "grad_norm": 0.33108267188072205,
+      "learning_rate": 0.00045975284173710283,
+      "loss": 3.3554,
+      "step": 40200
+    },
+    {
+      "epoch": 11.724481472850151,
+      "grad_norm": 0.3640613555908203,
+      "learning_rate": 0.00045957796560769446,
+      "loss": 3.3579,
+      "step": 40250
+    },
+    {
+      "epoch": 11.73904684222792,
+      "grad_norm": 0.32153892517089844,
+      "learning_rate": 0.0004594030894782862,
+      "loss": 3.3662,
+      "step": 40300
+    },
+    {
+      "epoch": 11.753612211605686,
+      "grad_norm": 0.34505143761634827,
+      "learning_rate": 0.00045922821334887785,
+      "loss": 3.3772,
+      "step": 40350
+    },
+    {
+      "epoch": 11.768177580983453,
+      "grad_norm": 0.341985285282135,
+      "learning_rate": 0.00045905333721946954,
+      "loss": 3.3526,
+      "step": 40400
+    },
+    {
+      "epoch": 11.782742950361222,
+      "grad_norm": 0.3537483513355255,
+      "learning_rate": 0.00045887846109006117,
+      "loss": 3.3576,
+      "step": 40450
+    },
+    {
+      "epoch": 11.797308319738988,
+      "grad_norm": 0.3518037497997284,
+      "learning_rate": 0.0004587035849606528,
+      "loss": 3.3552,
+      "step": 40500
+    },
+    {
+      "epoch": 11.811873689116755,
+      "grad_norm": 0.35557788610458374,
+      "learning_rate": 0.0004585287088312445,
+      "loss": 3.3785,
+      "step": 40550
+    },
+    {
+      "epoch": 11.826439058494524,
+      "grad_norm": 0.36040130257606506,
+      "learning_rate": 0.00045835383270183613,
+      "loss": 3.3717,
+      "step": 40600
+    },
+    {
+      "epoch": 11.84100442787229,
+      "grad_norm": 0.32966625690460205,
+      "learning_rate": 0.0004581789565724278,
+      "loss": 3.369,
+      "step": 40650
+    },
+    {
+      "epoch": 11.855569797250059,
+      "grad_norm": 0.33325114846229553,
+      "learning_rate": 0.00045800408044301946,
+      "loss": 3.3558,
+      "step": 40700
+    },
+    {
+      "epoch": 11.870135166627826,
+      "grad_norm": 0.34718069434165955,
+      "learning_rate": 0.0004578292043136111,
+      "loss": 3.3729,
+      "step": 40750
+    },
+    {
+      "epoch": 11.884700536005592,
+      "grad_norm": 0.3371071517467499,
+      "learning_rate": 0.00045765432818420284,
+      "loss": 3.3678,
+      "step": 40800
+    },
+    {
+      "epoch": 11.899265905383361,
+      "grad_norm": 0.3415924310684204,
+      "learning_rate": 0.0004574794520547945,
+      "loss": 3.3638,
+      "step": 40850
+    },
+    {
+      "epoch": 11.913831274761128,
+      "grad_norm": 0.3594110310077667,
+      "learning_rate": 0.00045730457592538617,
+      "loss": 3.3659,
+      "step": 40900
+    },
+    {
+      "epoch": 11.928396644138896,
+      "grad_norm": 0.3098519444465637,
+      "learning_rate": 0.0004571296997959778,
+      "loss": 3.3683,
+      "step": 40950
+    },
+    {
+      "epoch": 11.942962013516663,
+      "grad_norm": 0.32152944803237915,
+      "learning_rate": 0.0004569548236665695,
+      "loss": 3.3774,
+      "step": 41000
+    },
+    {
+      "epoch": 11.942962013516663,
+      "eval_accuracy": 0.3702212169951118,
+      "eval_loss": 3.5490400791168213,
+      "eval_runtime": 187.3283,
+      "eval_samples_per_second": 88.833,
+      "eval_steps_per_second": 5.557,
+      "step": 41000
+    },
+    {
+      "epoch": 11.95752738289443,
+      "grad_norm": 0.3719773292541504,
+      "learning_rate": 0.00045677994753716113,
+      "loss": 3.3698,
+      "step": 41050
+    },
+    {
+      "epoch": 11.972092752272198,
+      "grad_norm": 0.329070121049881,
+      "learning_rate": 0.00045660507140775277,
+      "loss": 3.3743,
+      "step": 41100
+    },
+    {
+      "epoch": 11.986658121649965,
+      "grad_norm": 0.3313333988189697,
+      "learning_rate": 0.00045643019527834446,
+      "loss": 3.3775,
+      "step": 41150
+    },
+    {
+      "epoch": 12.001165229550221,
+      "grad_norm": 0.33316850662231445,
+      "learning_rate": 0.0004562553191489361,
+      "loss": 3.3676,
+      "step": 41200
+    },
+    {
+      "epoch": 12.01573059892799,
+      "grad_norm": 0.33400437235832214,
+      "learning_rate": 0.00045608044301952784,
+      "loss": 3.2501,
+      "step": 41250
+    },
+    {
+      "epoch": 12.030295968305756,
+      "grad_norm": 0.3558497130870819,
+      "learning_rate": 0.0004559055668901195,
+      "loss": 3.2538,
+      "step": 41300
+    },
+    {
+      "epoch": 12.044861337683523,
+      "grad_norm": 0.3651934862136841,
+      "learning_rate": 0.0004557306907607111,
+      "loss": 3.266,
+      "step": 41350
+    },
+    {
+      "epoch": 12.059426707061291,
+      "grad_norm": 0.39346328377723694,
+      "learning_rate": 0.0004555558146313028,
+      "loss": 3.2725,
+      "step": 41400
+    },
+    {
+      "epoch": 12.073992076439058,
+      "grad_norm": 0.38937070965766907,
+      "learning_rate": 0.00045538093850189444,
+      "loss": 3.2637,
+      "step": 41450
+    },
+    {
+      "epoch": 12.088557445816827,
+      "grad_norm": 0.39181190729141235,
+      "learning_rate": 0.00045520606237248613,
+      "loss": 3.2686,
+      "step": 41500
+    },
+    {
+      "epoch": 12.103122815194594,
+      "grad_norm": 0.3455702066421509,
+      "learning_rate": 0.00045503118624307776,
+      "loss": 3.2885,
+      "step": 41550
+    },
+    {
+      "epoch": 12.11768818457236,
+      "grad_norm": 0.3350291848182678,
+      "learning_rate": 0.00045485631011366945,
+      "loss": 3.2849,
+      "step": 41600
+    },
+    {
+      "epoch": 12.132253553950129,
+      "grad_norm": 0.36060675978660583,
+      "learning_rate": 0.0004546814339842611,
+      "loss": 3.2866,
+      "step": 41650
+    },
+    {
+      "epoch": 12.146818923327896,
+      "grad_norm": 0.33322539925575256,
+      "learning_rate": 0.0004545065578548527,
+      "loss": 3.2868,
+      "step": 41700
+    },
+    {
+      "epoch": 12.161384292705662,
+      "grad_norm": 0.3414677679538727,
+      "learning_rate": 0.00045433168172544447,
+      "loss": 3.2949,
+      "step": 41750
+    },
+    {
+      "epoch": 12.17594966208343,
+      "grad_norm": 0.36268654465675354,
+      "learning_rate": 0.0004541568055960361,
+      "loss": 3.2938,
+      "step": 41800
+    },
+    {
+      "epoch": 12.190515031461198,
+      "grad_norm": 0.3349114656448364,
+      "learning_rate": 0.0004539819294666278,
+      "loss": 3.2975,
+      "step": 41850
+    },
+    {
+      "epoch": 12.205080400838966,
+      "grad_norm": 0.39525845646858215,
+      "learning_rate": 0.00045380705333721943,
+      "loss": 3.3016,
+      "step": 41900
+    },
+    {
+      "epoch": 12.219645770216733,
+      "grad_norm": 0.3538644313812256,
+      "learning_rate": 0.00045363217720781107,
+      "loss": 3.2817,
+      "step": 41950
+    },
+    {
+      "epoch": 12.2342111395945,
+      "grad_norm": 0.33876872062683105,
+      "learning_rate": 0.00045345730107840276,
+      "loss": 3.3016,
+      "step": 42000
+    },
+    {
+      "epoch": 12.2342111395945,
+      "eval_accuracy": 0.3695183341641627,
+      "eval_loss": 3.563988447189331,
+      "eval_runtime": 187.028,
+      "eval_samples_per_second": 88.976,
+      "eval_steps_per_second": 5.566,
+      "step": 42000
+    },
+    {
+      "epoch": 12.248776508972268,
+      "grad_norm": 0.3659180998802185,
+      "learning_rate": 0.0004532824249489944,
+      "loss": 3.2961,
+      "step": 42050
+    },
+    {
+      "epoch": 12.263341878350035,
+      "grad_norm": 0.3452215790748596,
+      "learning_rate": 0.0004531075488195861,
+      "loss": 3.3187,
+      "step": 42100
+    },
+    {
+      "epoch": 12.277907247727802,
+      "grad_norm": 0.34895065426826477,
+      "learning_rate": 0.0004529326726901777,
+      "loss": 3.3038,
+      "step": 42150
+    },
+    {
+      "epoch": 12.29247261710557,
+      "grad_norm": 0.3614833652973175,
+      "learning_rate": 0.00045275779656076947,
+      "loss": 3.3102,
+      "step": 42200
+    },
+    {
+      "epoch": 12.307037986483337,
+      "grad_norm": 0.34445762634277344,
+      "learning_rate": 0.0004525829204313611,
+      "loss": 3.2949,
+      "step": 42250
+    },
+    {
+      "epoch": 12.321603355861104,
+      "grad_norm": 0.36873099207878113,
+      "learning_rate": 0.00045240804430195274,
+      "loss": 3.3138,
+      "step": 42300
+    },
+    {
+      "epoch": 12.336168725238872,
+      "grad_norm": 0.3713667094707489,
+      "learning_rate": 0.00045223316817254443,
+      "loss": 3.3318,
+      "step": 42350
+    },
+    {
+      "epoch": 12.350734094616639,
+      "grad_norm": 0.35013601183891296,
+      "learning_rate": 0.00045205829204313607,
+      "loss": 3.3158,
+      "step": 42400
+    },
+    {
+      "epoch": 12.365299463994408,
+      "grad_norm": 0.35453933477401733,
+      "learning_rate": 0.00045188341591372776,
+      "loss": 3.314,
+      "step": 42450
+    },
+    {
+      "epoch": 12.379864833372174,
+      "grad_norm": 0.3458472490310669,
+      "learning_rate": 0.0004517085397843194,
+      "loss": 3.3239,
+      "step": 42500
+    },
+    {
+      "epoch": 12.394430202749941,
+      "grad_norm": 0.34224337339401245,
+      "learning_rate": 0.00045153366365491103,
+      "loss": 3.3146,
+      "step": 42550
+    },
+    {
+      "epoch": 12.40899557212771,
+      "grad_norm": 0.35080021619796753,
+      "learning_rate": 0.0004513587875255027,
+      "loss": 3.3188,
+      "step": 42600
+    },
+    {
+      "epoch": 12.423560941505476,
+      "grad_norm": 0.35965898633003235,
+      "learning_rate": 0.00045118391139609436,
+      "loss": 3.3247,
+      "step": 42650
+    },
+    {
+      "epoch": 12.438126310883243,
+      "grad_norm": 0.36585733294487,
+      "learning_rate": 0.0004510090352666861,
+      "loss": 3.3236,
+      "step": 42700
+    },
+    {
+      "epoch": 12.452691680261012,
+      "grad_norm": 0.38818955421447754,
+      "learning_rate": 0.00045083415913727774,
+      "loss": 3.3228,
+      "step": 42750
+    },
+    {
+      "epoch": 12.467257049638778,
+      "grad_norm": 0.3432595431804657,
+      "learning_rate": 0.0004506592830078694,
+      "loss": 3.3295,
+      "step": 42800
+    },
+    {
+      "epoch": 12.481822419016547,
+      "grad_norm": 0.3374660611152649,
+      "learning_rate": 0.00045048440687846106,
+      "loss": 3.3338,
+      "step": 42850
+    },
+    {
+      "epoch": 12.496387788394314,
+      "grad_norm": 0.3394157886505127,
+      "learning_rate": 0.0004503095307490527,
+      "loss": 3.337,
+      "step": 42900
+    },
+    {
+      "epoch": 12.51095315777208,
+      "grad_norm": 0.3551163375377655,
+      "learning_rate": 0.0004501346546196444,
+      "loss": 3.3385,
+      "step": 42950
+    },
+    {
+      "epoch": 12.525518527149849,
+      "grad_norm": 0.33821481466293335,
+      "learning_rate": 0.000449959778490236,
+      "loss": 3.3469,
+      "step": 43000
+    },
+    {
+      "epoch": 12.525518527149849,
+      "eval_accuracy": 0.370278722383155,
+      "eval_loss": 3.5560638904571533,
+      "eval_runtime": 186.841,
+      "eval_samples_per_second": 89.065,
+      "eval_steps_per_second": 5.572,
+      "step": 43000
+    },
+    {
+      "epoch": 12.540083896527616,
+      "grad_norm": 0.33071276545524597,
+      "learning_rate": 0.0004497849023608277,
+      "loss": 3.3416,
+      "step": 43050
+    },
+    {
+      "epoch": 12.554649265905383,
+      "grad_norm": 0.3239411413669586,
+      "learning_rate": 0.00044961002623141935,
+      "loss": 3.3453,
+      "step": 43100
+    },
+    {
+      "epoch": 12.569214635283151,
+      "grad_norm": 0.34707823395729065,
+      "learning_rate": 0.000449435150102011,
+      "loss": 3.3233,
+      "step": 43150
+    },
+    {
+      "epoch": 12.583780004660918,
+      "grad_norm": 0.3594145178794861,
+      "learning_rate": 0.00044926027397260273,
+      "loss": 3.3387,
+      "step": 43200
+    },
+    {
+      "epoch": 12.598345374038686,
+      "grad_norm": 0.3301078975200653,
+      "learning_rate": 0.00044908539784319437,
+      "loss": 3.3321,
+      "step": 43250
+    },
+    {
+      "epoch": 12.612910743416453,
+      "grad_norm": 0.3485218286514282,
+      "learning_rate": 0.00044891052171378606,
+      "loss": 3.3371,
+      "step": 43300
+    },
+    {
+      "epoch": 12.62747611279422,
+      "grad_norm": 0.36146172881126404,
+      "learning_rate": 0.0004487356455843777,
+      "loss": 3.3446,
+      "step": 43350
+    },
+    {
+      "epoch": 12.642041482171988,
+      "grad_norm": 0.38193467259407043,
+      "learning_rate": 0.00044856076945496933,
+      "loss": 3.3344,
+      "step": 43400
+    },
+    {
+      "epoch": 12.656606851549755,
+      "grad_norm": 0.38987061381340027,
+      "learning_rate": 0.000448385893325561,
+      "loss": 3.3418,
+      "step": 43450
+    },
+    {
+      "epoch": 12.671172220927522,
+      "grad_norm": 0.3399363160133362,
+      "learning_rate": 0.00044821101719615266,
+      "loss": 3.3405,
+      "step": 43500
+    },
+    {
+      "epoch": 12.68573759030529,
+      "grad_norm": 0.3345790505409241,
+      "learning_rate": 0.00044803614106674435,
+      "loss": 3.3451,
+      "step": 43550
+    },
+    {
+      "epoch": 12.700302959683057,
+      "grad_norm": 0.3493204712867737,
+      "learning_rate": 0.000447861264937336,
+      "loss": 3.3492,
+      "step": 43600
+    },
+    {
+      "epoch": 12.714868329060826,
+      "grad_norm": 0.3261966407299042,
+      "learning_rate": 0.00044768638880792773,
+      "loss": 3.3379,
+      "step": 43650
+    },
+    {
+      "epoch": 12.729433698438593,
+      "grad_norm": 0.3602188229560852,
+      "learning_rate": 0.00044751151267851937,
+      "loss": 3.3499,
+      "step": 43700
+    },
+    {
+      "epoch": 12.74399906781636,
+      "grad_norm": 0.3429162800312042,
+      "learning_rate": 0.000447336636549111,
+      "loss": 3.3506,
+      "step": 43750
+    },
+    {
+      "epoch": 12.758564437194128,
+      "grad_norm": 0.3623040020465851,
+      "learning_rate": 0.0004471617604197027,
+      "loss": 3.3488,
+      "step": 43800
+    },
+    {
+      "epoch": 12.773129806571895,
+      "grad_norm": 0.3544643521308899,
+      "learning_rate": 0.00044698688429029433,
+      "loss": 3.3387,
+      "step": 43850
+    },
+    {
+      "epoch": 12.787695175949661,
+      "grad_norm": 0.358529657125473,
+      "learning_rate": 0.000446812008160886,
+      "loss": 3.3481,
+      "step": 43900
+    },
+    {
+      "epoch": 12.80226054532743,
+      "grad_norm": 0.3519335091114044,
+      "learning_rate": 0.00044663713203147766,
+      "loss": 3.3496,
+      "step": 43950
+    },
+    {
+      "epoch": 12.816825914705197,
+      "grad_norm": 0.3490459620952606,
+      "learning_rate": 0.0004464622559020693,
+      "loss": 3.3608,
+      "step": 44000
+    },
+    {
+      "epoch": 12.816825914705197,
+      "eval_accuracy": 0.37078063035077935,
+      "eval_loss": 3.5487096309661865,
+      "eval_runtime": 185.64,
+      "eval_samples_per_second": 89.641,
+      "eval_steps_per_second": 5.608,
+      "step": 44000
+    },
+    {
+      "epoch": 12.831391284082965,
+      "grad_norm": 0.36724475026130676,
+      "learning_rate": 0.000446287379772661,
+      "loss": 3.3458,
+      "step": 44050
+    },
+    {
+      "epoch": 12.845956653460732,
+      "grad_norm": 0.3556743562221527,
+      "learning_rate": 0.0004461125036432526,
+      "loss": 3.3565,
+      "step": 44100
+    },
+    {
+      "epoch": 12.860522022838499,
+      "grad_norm": 0.4028126001358032,
+      "learning_rate": 0.00044593762751384436,
+      "loss": 3.3458,
+      "step": 44150
+    },
+    {
+      "epoch": 12.875087392216267,
+      "grad_norm": 0.34439170360565186,
+      "learning_rate": 0.000445762751384436,
+      "loss": 3.3495,
+      "step": 44200
+    },
+    {
+      "epoch": 12.889652761594034,
+      "grad_norm": 0.3576850891113281,
+      "learning_rate": 0.0004455878752550277,
+      "loss": 3.3563,
+      "step": 44250
+    },
+    {
+      "epoch": 12.9042181309718,
+      "grad_norm": 0.346293568611145,
+      "learning_rate": 0.0004454129991256193,
+      "loss": 3.343,
+      "step": 44300
+    },
+    {
+      "epoch": 12.91878350034957,
+      "grad_norm": 0.3490327000617981,
+      "learning_rate": 0.00044523812299621096,
+      "loss": 3.3499,
+      "step": 44350
+    },
+    {
+      "epoch": 12.933348869727336,
+      "grad_norm": 0.34292006492614746,
+      "learning_rate": 0.00044506324686680265,
+      "loss": 3.3389,
+      "step": 44400
+    },
+    {
+      "epoch": 12.947914239105105,
+      "grad_norm": 0.35687047243118286,
+      "learning_rate": 0.0004448883707373943,
+      "loss": 3.3452,
+      "step": 44450
+    },
+    {
+      "epoch": 12.962479608482871,
+      "grad_norm": 0.3330571949481964,
+      "learning_rate": 0.000444713494607986,
+      "loss": 3.3588,
+      "step": 44500
+    },
+    {
+      "epoch": 12.977044977860638,
+      "grad_norm": 0.33951443433761597,
+      "learning_rate": 0.0004445386184785776,
+      "loss": 3.3641,
+      "step": 44550
+    },
+    {
+      "epoch": 12.991610347238407,
+      "grad_norm": 0.34642261266708374,
+      "learning_rate": 0.00044436374234916925,
+      "loss": 3.3546,
+      "step": 44600
+    },
+    {
+      "epoch": 13.006117455138662,
+      "grad_norm": 0.36387667059898376,
+      "learning_rate": 0.000444188866219761,
+      "loss": 3.3089,
+      "step": 44650
+    },
+    {
+      "epoch": 13.02068282451643,
+      "grad_norm": 0.3527930676937103,
+      "learning_rate": 0.00044401399009035263,
+      "loss": 3.2279,
+      "step": 44700
+    },
+    {
+      "epoch": 13.035248193894198,
+      "grad_norm": 0.3520139455795288,
+      "learning_rate": 0.0004438391139609443,
+      "loss": 3.2397,
+      "step": 44750
+    },
+    {
+      "epoch": 13.049813563271965,
+      "grad_norm": 0.3651536703109741,
+      "learning_rate": 0.00044366423783153596,
+      "loss": 3.2539,
+      "step": 44800
+    },
+    {
+      "epoch": 13.064378932649731,
+      "grad_norm": 0.3563643991947174,
+      "learning_rate": 0.0004434893617021276,
+      "loss": 3.2584,
+      "step": 44850
+    },
+    {
+      "epoch": 13.0789443020275,
+      "grad_norm": 0.3816784918308258,
+      "learning_rate": 0.0004433144855727193,
+      "loss": 3.2598,
+      "step": 44900
+    },
+    {
+      "epoch": 13.093509671405267,
+      "grad_norm": 0.3494201898574829,
+      "learning_rate": 0.0004431396094433109,
+      "loss": 3.2528,
+      "step": 44950
+    },
+    {
+      "epoch": 13.108075040783035,
+      "grad_norm": 0.36257123947143555,
+      "learning_rate": 0.0004429647333139026,
+      "loss": 3.2594,
+      "step": 45000
+    },
+    {
+      "epoch": 13.108075040783035,
+      "eval_accuracy": 0.37042713097152,
+      "eval_loss": 3.5596282482147217,
+      "eval_runtime": 181.0392,
+      "eval_samples_per_second": 91.919,
+      "eval_steps_per_second": 5.75,
+      "step": 45000
+    },
+    {
+      "epoch": 13.122640410160802,
+      "grad_norm": 0.36224088072776794,
+      "learning_rate": 0.00044278985718449425,
+      "loss": 3.2588,
+      "step": 45050
+    },
+    {
+      "epoch": 13.137205779538569,
+      "grad_norm": 0.3430086374282837,
+      "learning_rate": 0.000442614981055086,
+      "loss": 3.2742,
+      "step": 45100
+    },
+    {
+      "epoch": 13.151771148916337,
+      "grad_norm": 0.3522573411464691,
+      "learning_rate": 0.00044244010492567763,
+      "loss": 3.2793,
+      "step": 45150
+    },
+    {
+      "epoch": 13.166336518294104,
+      "grad_norm": 0.36241790652275085,
+      "learning_rate": 0.00044226522879626927,
+      "loss": 3.2691,
+      "step": 45200
+    },
+    {
+      "epoch": 13.18090188767187,
+      "grad_norm": 0.3618699312210083,
+      "learning_rate": 0.00044209035266686096,
+      "loss": 3.279,
+      "step": 45250
+    },
+    {
+      "epoch": 13.19546725704964,
+      "grad_norm": 0.3411371409893036,
+      "learning_rate": 0.0004419154765374526,
+      "loss": 3.2829,
+      "step": 45300
+    },
+    {
+      "epoch": 13.210032626427406,
+      "grad_norm": 0.3545594811439514,
+      "learning_rate": 0.0004417406004080443,
+      "loss": 3.2742,
+      "step": 45350
+    },
+    {
+      "epoch": 13.224597995805174,
+      "grad_norm": 0.3439912796020508,
+      "learning_rate": 0.0004415657242786359,
+      "loss": 3.295,
+      "step": 45400
+    },
+    {
+      "epoch": 13.239163365182941,
+      "grad_norm": 0.3650055527687073,
+      "learning_rate": 0.00044139084814922755,
+      "loss": 3.2871,
+      "step": 45450
+    },
+    {
+      "epoch": 13.253728734560708,
+      "grad_norm": 0.3532308340072632,
+      "learning_rate": 0.00044121597201981924,
+      "loss": 3.2812,
+      "step": 45500
+    },
+    {
+      "epoch": 13.268294103938477,
+      "grad_norm": 0.35890576243400574,
+      "learning_rate": 0.0004410410958904109,
+      "loss": 3.2823,
+      "step": 45550
+    },
+    {
+      "epoch": 13.282859473316243,
+      "grad_norm": 0.34750354290008545,
+      "learning_rate": 0.0004408662197610026,
+      "loss": 3.2712,
+      "step": 45600
+    },
+    {
+      "epoch": 13.29742484269401,
+      "grad_norm": 0.3511049151420593,
+      "learning_rate": 0.00044069134363159426,
+      "loss": 3.2994,
+      "step": 45650
+    },
+    {
+      "epoch": 13.311990212071779,
+      "grad_norm": 0.3906053602695465,
+      "learning_rate": 0.00044051646750218595,
+      "loss": 3.2986,
+      "step": 45700
+    },
+    {
+      "epoch": 13.326555581449545,
+      "grad_norm": 0.3588992655277252,
+      "learning_rate": 0.0004403415913727776,
+      "loss": 3.2932,
+      "step": 45750
+    },
+    {
+      "epoch": 13.341120950827314,
+      "grad_norm": 0.343471884727478,
+      "learning_rate": 0.0004401667152433692,
+      "loss": 3.3092,
+      "step": 45800
+    },
+    {
+      "epoch": 13.35568632020508,
+      "grad_norm": 0.3413912057876587,
+      "learning_rate": 0.0004399918391139609,
+      "loss": 3.298,
+      "step": 45850
+    },
+    {
+      "epoch": 13.370251689582847,
+      "grad_norm": 0.3421027660369873,
+      "learning_rate": 0.00043981696298455255,
+      "loss": 3.3044,
+      "step": 45900
+    },
+    {
+      "epoch": 13.384817058960616,
+      "grad_norm": 0.3687370717525482,
+      "learning_rate": 0.00043964208685514424,
+      "loss": 3.3,
+      "step": 45950
+    },
+    {
+      "epoch": 13.399382428338383,
+      "grad_norm": 0.3572240173816681,
+      "learning_rate": 0.0004394672107257359,
+      "loss": 3.2975,
+      "step": 46000
+    },
+    {
+      "epoch": 13.399382428338383,
+      "eval_accuracy": 0.37047711009200746,
+      "eval_loss": 3.555755615234375,
+      "eval_runtime": 180.926,
+      "eval_samples_per_second": 91.977,
+      "eval_steps_per_second": 5.754,
+      "step": 46000
+    },
+    {
+      "epoch": 13.41394779771615,
+      "grad_norm": 0.3360699713230133,
+      "learning_rate": 0.0004392923345963275,
+      "loss": 3.3111,
+      "step": 46050
+    },
+    {
+      "epoch": 13.428513167093918,
+      "grad_norm": 0.38085097074508667,
+      "learning_rate": 0.00043911745846691926,
+      "loss": 3.3174,
+      "step": 46100
+    },
+    {
+      "epoch": 13.443078536471685,
+      "grad_norm": 0.36990615725517273,
+      "learning_rate": 0.0004389425823375109,
+      "loss": 3.3099,
+      "step": 46150
+    },
+    {
+      "epoch": 13.457643905849451,
+      "grad_norm": 0.36363065242767334,
+      "learning_rate": 0.0004387677062081026,
+      "loss": 3.3142,
+      "step": 46200
+    },
+    {
+      "epoch": 13.47220927522722,
+      "grad_norm": 0.34396278858184814,
+      "learning_rate": 0.0004385928300786942,
+      "loss": 3.319,
+      "step": 46250
+    },
+    {
+      "epoch": 13.486774644604987,
+      "grad_norm": 0.36657220125198364,
+      "learning_rate": 0.0004384179539492859,
+      "loss": 3.3065,
+      "step": 46300
+    },
+    {
+      "epoch": 13.501340013982755,
+      "grad_norm": 0.36169782280921936,
+      "learning_rate": 0.00043824307781987755,
+      "loss": 3.3197,
+      "step": 46350
+    },
+    {
+      "epoch": 13.515905383360522,
+      "grad_norm": 0.36206182837486267,
+      "learning_rate": 0.0004380682016904692,
+      "loss": 3.3157,
+      "step": 46400
+    },
+    {
+      "epoch": 13.530470752738289,
+      "grad_norm": 0.35609057545661926,
+      "learning_rate": 0.0004378933255610609,
+      "loss": 3.3094,
+      "step": 46450
+    },
+    {
+      "epoch": 13.545036122116057,
+      "grad_norm": 0.3487824499607086,
+      "learning_rate": 0.0004377184494316525,
+      "loss": 3.3041,
+      "step": 46500
+    },
+    {
+      "epoch": 13.559601491493824,
+      "grad_norm": 0.3546448051929474,
+      "learning_rate": 0.00043754357330224426,
+      "loss": 3.3157,
+      "step": 46550
+    },
+    {
+      "epoch": 13.574166860871593,
+      "grad_norm": 0.3269023001194,
+      "learning_rate": 0.0004373686971728359,
+      "loss": 3.3176,
+      "step": 46600
+    },
+    {
+      "epoch": 13.58873223024936,
+      "grad_norm": 0.3617735505104065,
+      "learning_rate": 0.00043719382104342753,
+      "loss": 3.3243,
+      "step": 46650
+    },
+    {
+      "epoch": 13.603297599627126,
+      "grad_norm": 0.3514387011528015,
+      "learning_rate": 0.0004370189449140192,
+      "loss": 3.3227,
+      "step": 46700
+    },
+    {
+      "epoch": 13.617862969004895,
+      "grad_norm": 0.3604964315891266,
+      "learning_rate": 0.00043684406878461085,
+      "loss": 3.3264,
+      "step": 46750
+    },
+    {
+      "epoch": 13.632428338382661,
+      "grad_norm": 0.34199249744415283,
+      "learning_rate": 0.00043666919265520254,
+      "loss": 3.3331,
+      "step": 46800
+    },
+    {
+      "epoch": 13.646993707760428,
+      "grad_norm": 0.3650147616863251,
+      "learning_rate": 0.0004364943165257942,
+      "loss": 3.3322,
+      "step": 46850
+    },
+    {
+      "epoch": 13.661559077138197,
+      "grad_norm": 0.350155770778656,
+      "learning_rate": 0.0004363194403963858,
+      "loss": 3.3266,
+      "step": 46900
+    },
+    {
+      "epoch": 13.676124446515963,
+      "grad_norm": 0.36043399572372437,
+      "learning_rate": 0.0004361445642669775,
+      "loss": 3.3327,
+      "step": 46950
+    },
+    {
+      "epoch": 13.69068981589373,
+      "grad_norm": 0.35996055603027344,
+      "learning_rate": 0.00043596968813756914,
+      "loss": 3.3248,
+      "step": 47000
+    },
+    {
+      "epoch": 13.69068981589373,
+      "eval_accuracy": 0.3708637720876843,
+      "eval_loss": 3.549927234649658,
+      "eval_runtime": 181.1356,
+      "eval_samples_per_second": 91.87,
+      "eval_steps_per_second": 5.747,
+      "step": 47000
+    },
+    {
+      "epoch": 13.705255185271499,
+      "grad_norm": 0.35664069652557373,
+      "learning_rate": 0.0004357948120081609,
+      "loss": 3.3268,
+      "step": 47050
+    },
+    {
+      "epoch": 13.719820554649266,
+      "grad_norm": 0.3575288653373718,
+      "learning_rate": 0.0004356199358787525,
+      "loss": 3.3279,
+      "step": 47100
+    },
+    {
+      "epoch": 13.734385924027034,
+      "grad_norm": 0.3597092926502228,
+      "learning_rate": 0.0004354450597493442,
+      "loss": 3.3257,
+      "step": 47150
+    },
+    {
+      "epoch": 13.7489512934048,
+      "grad_norm": 0.3677772581577301,
+      "learning_rate": 0.00043527018361993585,
+      "loss": 3.3305,
+      "step": 47200
+    },
+    {
+      "epoch": 13.763516662782568,
+      "grad_norm": 0.34374454617500305,
+      "learning_rate": 0.0004350953074905275,
+      "loss": 3.347,
+      "step": 47250
+    },
+    {
+      "epoch": 13.778082032160336,
+      "grad_norm": 0.3679696023464203,
+      "learning_rate": 0.0004349204313611192,
+      "loss": 3.3403,
+      "step": 47300
+    },
+    {
+      "epoch": 13.792647401538103,
+      "grad_norm": 0.3548065721988678,
+      "learning_rate": 0.0004347455552317108,
+      "loss": 3.3284,
+      "step": 47350
+    },
+    {
+      "epoch": 13.80721277091587,
+      "grad_norm": 0.3433007001876831,
+      "learning_rate": 0.0004345706791023025,
+      "loss": 3.3307,
+      "step": 47400
+    },
+    {
+      "epoch": 13.821778140293638,
+      "grad_norm": 0.3408163785934448,
+      "learning_rate": 0.00043439580297289414,
+      "loss": 3.3236,
+      "step": 47450
+    },
+    {
+      "epoch": 13.836343509671405,
+      "grad_norm": 0.37055566906929016,
+      "learning_rate": 0.0004342209268434858,
+      "loss": 3.3317,
+      "step": 47500
+    },
+    {
+      "epoch": 13.850908879049173,
+      "grad_norm": 0.3566333055496216,
+      "learning_rate": 0.0004340460507140775,
+      "loss": 3.332,
+      "step": 47550
+    },
+    {
+      "epoch": 13.86547424842694,
+      "grad_norm": 0.360100656747818,
+      "learning_rate": 0.00043387117458466916,
+      "loss": 3.3341,
+      "step": 47600
+    },
+    {
+      "epoch": 13.880039617804707,
+      "grad_norm": 0.352517694234848,
+      "learning_rate": 0.00043369629845526085,
+      "loss": 3.3212,
+      "step": 47650
+    },
+    {
+      "epoch": 13.894604987182475,
+      "grad_norm": 0.35851529240608215,
+      "learning_rate": 0.0004335214223258525,
+      "loss": 3.3385,
+      "step": 47700
+    },
+    {
+      "epoch": 13.909170356560242,
+      "grad_norm": 0.3513929545879364,
+      "learning_rate": 0.0004333465461964442,
+      "loss": 3.3328,
+      "step": 47750
+    },
+    {
+      "epoch": 13.923735725938009,
+      "grad_norm": 0.342102974653244,
+      "learning_rate": 0.0004331716700670358,
+      "loss": 3.3326,
+      "step": 47800
+    },
+    {
+      "epoch": 13.938301095315778,
+      "grad_norm": 0.3388274013996124,
+      "learning_rate": 0.00043299679393762745,
+      "loss": 3.3507,
+      "step": 47850
+    },
+    {
+      "epoch": 13.952866464693544,
+      "grad_norm": 0.34295016527175903,
+      "learning_rate": 0.00043282191780821914,
+      "loss": 3.3478,
+      "step": 47900
+    },
+    {
+      "epoch": 13.967431834071313,
+      "grad_norm": 0.38172635436058044,
+      "learning_rate": 0.00043264704167881077,
+      "loss": 3.3322,
+      "step": 47950
+    },
+    {
+      "epoch": 13.98199720344908,
+      "grad_norm": 0.34662753343582153,
+      "learning_rate": 0.0004324721655494025,
+      "loss": 3.3497,
+      "step": 48000
+    },
+    {
+      "epoch": 13.98199720344908,
+      "eval_accuracy": 0.37183277903548767,
+      "eval_loss": 3.5397329330444336,
+      "eval_runtime": 181.0372,
+      "eval_samples_per_second": 91.92,
+      "eval_steps_per_second": 5.75,
+      "step": 48000
+    },
+    {
+      "epoch": 13.996562572826846,
+      "grad_norm": 0.33236080408096313,
+      "learning_rate": 0.00043229728941999415,
+      "loss": 3.3395,
+      "step": 48050
+    },
+    {
+      "epoch": 14.011069680727104,
+      "grad_norm": 0.37694936990737915,
+      "learning_rate": 0.0004321224132905858,
+      "loss": 3.2425,
+      "step": 48100
+    },
+    {
+      "epoch": 14.02563505010487,
+      "grad_norm": 0.38224613666534424,
+      "learning_rate": 0.0004319475371611775,
+      "loss": 3.2292,
+      "step": 48150
+    },
+    {
+      "epoch": 14.040200419482638,
+      "grad_norm": 0.3885684907436371,
+      "learning_rate": 0.0004317726610317691,
+      "loss": 3.2317,
+      "step": 48200
+    },
+    {
+      "epoch": 14.054765788860406,
+      "grad_norm": 0.364306777715683,
+      "learning_rate": 0.0004315977849023608,
+      "loss": 3.2383,
+      "step": 48250
+    },
+    {
+      "epoch": 14.069331158238173,
+      "grad_norm": 0.3601495921611786,
+      "learning_rate": 0.00043142290877295244,
+      "loss": 3.2439,
+      "step": 48300
+    },
+    {
+      "epoch": 14.08389652761594,
+      "grad_norm": 0.38885897397994995,
+      "learning_rate": 0.00043124803264354413,
+      "loss": 3.2479,
+      "step": 48350
+    },
+    {
+      "epoch": 14.098461896993708,
+      "grad_norm": 0.3615412414073944,
+      "learning_rate": 0.00043107315651413577,
+      "loss": 3.2454,
+      "step": 48400
+    },
+    {
+      "epoch": 14.113027266371475,
+      "grad_norm": 0.3586154580116272,
+      "learning_rate": 0.0004308982803847274,
+      "loss": 3.254,
+      "step": 48450
+    },
+    {
+      "epoch": 14.127592635749243,
+      "grad_norm": 0.3595428466796875,
+      "learning_rate": 0.00043072340425531915,
+      "loss": 3.2512,
+      "step": 48500
+    },
+    {
+      "epoch": 14.14215800512701,
+      "grad_norm": 0.3684776723384857,
+      "learning_rate": 0.0004305485281259108,
+      "loss": 3.2605,
+      "step": 48550
+    },
+    {
+      "epoch": 14.156723374504777,
+      "grad_norm": 0.3991050720214844,
+      "learning_rate": 0.0004303736519965025,
+      "loss": 3.2668,
+      "step": 48600
+    },
+    {
+      "epoch": 14.171288743882545,
+      "grad_norm": 0.3561571538448334,
+      "learning_rate": 0.0004301987758670941,
+      "loss": 3.2567,
+      "step": 48650
+    },
+    {
+      "epoch": 14.185854113260312,
+      "grad_norm": 0.3748098611831665,
+      "learning_rate": 0.00043002389973768575,
+      "loss": 3.2589,
+      "step": 48700
+    },
+    {
+      "epoch": 14.200419482638079,
+      "grad_norm": 0.3451537787914276,
+      "learning_rate": 0.00042984902360827744,
+      "loss": 3.2672,
+      "step": 48750
+    },
+    {
+      "epoch": 14.214984852015847,
+      "grad_norm": 0.36239251494407654,
+      "learning_rate": 0.0004296741474788691,
+      "loss": 3.2614,
+      "step": 48800
+    },
+    {
+      "epoch": 14.229550221393614,
+      "grad_norm": 0.3522060513496399,
+      "learning_rate": 0.00042949927134946077,
+      "loss": 3.2558,
+      "step": 48850
+    },
+    {
+      "epoch": 14.244115590771383,
+      "grad_norm": 0.36641281843185425,
+      "learning_rate": 0.0004293243952200524,
+      "loss": 3.2774,
+      "step": 48900
+    },
+    {
+      "epoch": 14.25868096014915,
+      "grad_norm": 0.37674859166145325,
+      "learning_rate": 0.00042914951909064415,
+      "loss": 3.2762,
+      "step": 48950
+    },
+    {
+      "epoch": 14.273246329526916,
+      "grad_norm": 0.3353112041950226,
+      "learning_rate": 0.0004289746429612358,
+      "loss": 3.2825,
+      "step": 49000
+    },
+    {
+      "epoch": 14.273246329526916,
+      "eval_accuracy": 0.3710279387987442,
+      "eval_loss": 3.5576698780059814,
+      "eval_runtime": 181.0153,
+      "eval_samples_per_second": 91.931,
+      "eval_steps_per_second": 5.751,
+      "step": 49000
+    },
+    {
+      "epoch": 14.287811698904685,
+      "grad_norm": 0.3597235381603241,
+      "learning_rate": 0.0004287997668318274,
+      "loss": 3.2874,
+      "step": 49050
+    },
+    {
+      "epoch": 14.302377068282452,
+      "grad_norm": 0.3678181767463684,
+      "learning_rate": 0.0004286248907024191,
+      "loss": 3.2931,
+      "step": 49100
+    },
+    {
+      "epoch": 14.316942437660218,
+      "grad_norm": 0.3513096570968628,
+      "learning_rate": 0.00042845001457301075,
+      "loss": 3.2802,
+      "step": 49150
+    },
+    {
+      "epoch": 14.331507807037987,
+      "grad_norm": 0.3584321439266205,
+      "learning_rate": 0.00042827513844360244,
+      "loss": 3.2823,
+      "step": 49200
+    },
+    {
+      "epoch": 14.346073176415754,
+      "grad_norm": 0.3550703227519989,
+      "learning_rate": 0.00042810026231419407,
+      "loss": 3.2715,
+      "step": 49250
+    },
+    {
+      "epoch": 14.360638545793522,
+      "grad_norm": 0.37994474172592163,
+      "learning_rate": 0.0004279253861847857,
+      "loss": 3.2978,
+      "step": 49300
+    },
+    {
+      "epoch": 14.375203915171289,
+      "grad_norm": 0.38211414217948914,
+      "learning_rate": 0.0004277505100553774,
+      "loss": 3.2701,
+      "step": 49350
+    },
+    {
+      "epoch": 14.389769284549056,
+      "grad_norm": 0.3565288782119751,
+      "learning_rate": 0.00042757563392596904,
+      "loss": 3.299,
+      "step": 49400
+    },
+    {
+      "epoch": 14.404334653926824,
+      "grad_norm": 0.36878564953804016,
+      "learning_rate": 0.0004274007577965608,
+      "loss": 3.2761,
+      "step": 49450
+    },
+    {
+      "epoch": 14.418900023304591,
+      "grad_norm": 0.38277795910835266,
+      "learning_rate": 0.0004272258816671524,
+      "loss": 3.2846,
+      "step": 49500
+    },
+    {
+      "epoch": 14.433465392682358,
+      "grad_norm": 0.3722183108329773,
+      "learning_rate": 0.00042705100553774405,
+      "loss": 3.2987,
+      "step": 49550
+    },
+    {
+      "epoch": 14.448030762060126,
+      "grad_norm": 0.3671691417694092,
+      "learning_rate": 0.00042687612940833574,
+      "loss": 3.2881,
+      "step": 49600
+    },
+    {
+      "epoch": 14.462596131437893,
+      "grad_norm": 0.3849431276321411,
+      "learning_rate": 0.0004267012532789274,
+      "loss": 3.287,
+      "step": 49650
+    },
+    {
+      "epoch": 14.477161500815662,
+      "grad_norm": 0.34819379448890686,
+      "learning_rate": 0.00042652637714951907,
+      "loss": 3.287,
+      "step": 49700
+    },
+    {
+      "epoch": 14.491726870193428,
+      "grad_norm": 0.35875314474105835,
+      "learning_rate": 0.0004263515010201107,
+      "loss": 3.2997,
+      "step": 49750
+    },
+    {
+      "epoch": 14.506292239571195,
+      "grad_norm": 0.3784511685371399,
+      "learning_rate": 0.0004261766248907024,
+      "loss": 3.3047,
+      "step": 49800
+    },
+    {
+      "epoch": 14.520857608948964,
+      "grad_norm": 0.3357415199279785,
+      "learning_rate": 0.00042600174876129403,
+      "loss": 3.3013,
+      "step": 49850
+    },
+    {
+      "epoch": 14.53542297832673,
+      "grad_norm": 0.38833755254745483,
+      "learning_rate": 0.00042582687263188567,
+      "loss": 3.2928,
+      "step": 49900
+    },
+    {
+      "epoch": 14.549988347704497,
+      "grad_norm": 0.3690185248851776,
+      "learning_rate": 0.0004256519965024774,
+      "loss": 3.2939,
+      "step": 49950
+    },
+    {
+      "epoch": 14.564553717082266,
+      "grad_norm": 0.3692043125629425,
+      "learning_rate": 0.00042547712037306905,
+      "loss": 3.2991,
+      "step": 50000
+    },
+    {
+      "epoch": 14.564553717082266,
+      "eval_accuracy": 0.3709371531963529,
+      "eval_loss": 3.5529329776763916,
+      "eval_runtime": 181.1317,
+      "eval_samples_per_second": 91.872,
+      "eval_steps_per_second": 5.747,
+      "step": 50000
+    },
+    {
+      "epoch": 14.579119086460032,
+      "grad_norm": 0.3446950316429138,
+      "learning_rate": 0.00042530224424366074,
+      "loss": 3.3086,
+      "step": 50050
+    },
+    {
+      "epoch": 14.5936844558378,
+      "grad_norm": 0.38580191135406494,
+      "learning_rate": 0.0004251273681142524,
+      "loss": 3.3067,
+      "step": 50100
+    },
+    {
+      "epoch": 14.608249825215568,
+      "grad_norm": 0.35636115074157715,
+      "learning_rate": 0.000424952491984844,
+      "loss": 3.3112,
+      "step": 50150
+    },
+    {
+      "epoch": 14.622815194593334,
+      "grad_norm": 0.3655899465084076,
+      "learning_rate": 0.0004247776158554357,
+      "loss": 3.3111,
+      "step": 50200
+    },
+    {
+      "epoch": 14.637380563971103,
+      "grad_norm": 0.35176143050193787,
+      "learning_rate": 0.00042460273972602734,
+      "loss": 3.3112,
+      "step": 50250
+    },
+    {
+      "epoch": 14.65194593334887,
+      "grad_norm": 0.3397698402404785,
+      "learning_rate": 0.00042442786359661903,
+      "loss": 3.3112,
+      "step": 50300
+    },
+    {
+      "epoch": 14.666511302726637,
+      "grad_norm": 0.37180206179618835,
+      "learning_rate": 0.00042425298746721066,
+      "loss": 3.3068,
+      "step": 50350
+    },
+    {
+      "epoch": 14.681076672104405,
+      "grad_norm": 0.40762200951576233,
+      "learning_rate": 0.0004240781113378024,
+      "loss": 3.3146,
+      "step": 50400
+    },
+    {
+      "epoch": 14.695642041482172,
+      "grad_norm": 0.3677314221858978,
+      "learning_rate": 0.00042390323520839405,
+      "loss": 3.315,
+      "step": 50450
+    },
+    {
+      "epoch": 14.71020741085994,
+      "grad_norm": 0.3551027178764343,
+      "learning_rate": 0.0004237283590789857,
+      "loss": 3.3202,
+      "step": 50500
+    },
+    {
+      "epoch": 14.724772780237707,
+      "grad_norm": 0.3665424585342407,
+      "learning_rate": 0.00042355348294957737,
+      "loss": 3.3175,
+      "step": 50550
+    },
+    {
+      "epoch": 14.739338149615474,
+      "grad_norm": 0.37414008378982544,
+      "learning_rate": 0.000423378606820169,
+      "loss": 3.3133,
+      "step": 50600
+    },
+    {
+      "epoch": 14.753903518993242,
+      "grad_norm": 0.33545243740081787,
+      "learning_rate": 0.0004232037306907607,
+      "loss": 3.321,
+      "step": 50650
+    },
+    {
+      "epoch": 14.76846888837101,
+      "grad_norm": 0.35436394810676575,
+      "learning_rate": 0.00042302885456135233,
+      "loss": 3.3096,
+      "step": 50700
+    },
+    {
+      "epoch": 14.783034257748776,
+      "grad_norm": 0.3778373897075653,
+      "learning_rate": 0.00042285397843194397,
+      "loss": 3.324,
+      "step": 50750
+    },
+    {
+      "epoch": 14.797599627126544,
+      "grad_norm": 0.3442220091819763,
+      "learning_rate": 0.00042267910230253566,
+      "loss": 3.3186,
+      "step": 50800
+    },
+    {
+      "epoch": 14.812164996504311,
+      "grad_norm": 0.36831867694854736,
+      "learning_rate": 0.0004225042261731273,
+      "loss": 3.3144,
+      "step": 50850
+    },
+    {
+      "epoch": 14.826730365882078,
+      "grad_norm": 0.3595118522644043,
+      "learning_rate": 0.00042232935004371904,
+      "loss": 3.3203,
+      "step": 50900
+    },
+    {
+      "epoch": 14.841295735259846,
+      "grad_norm": 0.3834764063358307,
+      "learning_rate": 0.0004221544739143107,
+      "loss": 3.3204,
+      "step": 50950
+    },
+    {
+      "epoch": 14.855861104637613,
+      "grad_norm": 0.35744091868400574,
+      "learning_rate": 0.00042197959778490237,
+      "loss": 3.312,
+      "step": 51000
+    },
+    {
+      "epoch": 14.855861104637613,
+      "eval_accuracy": 0.3720758539579524,
+      "eval_loss": 3.5414035320281982,
+      "eval_runtime": 180.9982,
+      "eval_samples_per_second": 91.94,
+      "eval_steps_per_second": 5.751,
+      "step": 51000
+    },
+    {
+      "epoch": 14.870426474015382,
+      "grad_norm": 0.3626762926578522,
+      "learning_rate": 0.000421804721655494,
+      "loss": 3.3163,
+      "step": 51050
+    },
+    {
+      "epoch": 14.884991843393149,
+      "grad_norm": 0.3768766224384308,
+      "learning_rate": 0.00042162984552608564,
+      "loss": 3.3191,
+      "step": 51100
+    },
+    {
+      "epoch": 14.899557212770915,
+      "grad_norm": 0.3746778666973114,
+      "learning_rate": 0.00042145496939667733,
+      "loss": 3.3165,
+      "step": 51150
+    },
+    {
+      "epoch": 14.914122582148684,
+      "grad_norm": 0.38692834973335266,
+      "learning_rate": 0.00042128009326726897,
+      "loss": 3.3282,
+      "step": 51200
+    },
+    {
+      "epoch": 14.92868795152645,
+      "grad_norm": 0.3417520523071289,
+      "learning_rate": 0.00042110521713786066,
+      "loss": 3.3214,
+      "step": 51250
+    },
+    {
+      "epoch": 14.943253320904217,
+      "grad_norm": 0.3445229232311249,
+      "learning_rate": 0.0004209303410084523,
+      "loss": 3.3292,
+      "step": 51300
+    },
+    {
+      "epoch": 14.957818690281986,
+      "grad_norm": 0.3492046892642975,
+      "learning_rate": 0.00042075546487904393,
+      "loss": 3.3202,
+      "step": 51350
+    },
+    {
+      "epoch": 14.972384059659753,
+      "grad_norm": 0.350506454706192,
+      "learning_rate": 0.0004205805887496357,
+      "loss": 3.3179,
+      "step": 51400
+    },
+    {
+      "epoch": 14.986949429037521,
+      "grad_norm": 0.36864006519317627,
+      "learning_rate": 0.0004204057126202273,
+      "loss": 3.3209,
+      "step": 51450
+    },
+    {
+      "epoch": 15.001456536937777,
+      "grad_norm": 0.38432347774505615,
+      "learning_rate": 0.000420230836490819,
+      "loss": 3.3159,
+      "step": 51500
+    },
+    {
+      "epoch": 15.016021906315544,
+      "grad_norm": 0.3642912805080414,
+      "learning_rate": 0.00042005596036141064,
+      "loss": 3.2077,
+      "step": 51550
+    },
+    {
+      "epoch": 15.030587275693312,
+      "grad_norm": 0.38347211480140686,
+      "learning_rate": 0.0004198810842320023,
+      "loss": 3.2167,
+      "step": 51600
+    },
+    {
+      "epoch": 15.045152645071079,
+      "grad_norm": 0.36690157651901245,
+      "learning_rate": 0.00041970620810259396,
+      "loss": 3.2239,
+      "step": 51650
+    },
+    {
+      "epoch": 15.059718014448846,
+      "grad_norm": 0.35979774594306946,
+      "learning_rate": 0.0004195313319731856,
+      "loss": 3.2159,
+      "step": 51700
+    },
+    {
+      "epoch": 15.074283383826614,
+      "grad_norm": 0.38011622428894043,
+      "learning_rate": 0.0004193564558437773,
+      "loss": 3.212,
+      "step": 51750
+    },
+    {
+      "epoch": 15.088848753204381,
+      "grad_norm": 0.35243844985961914,
+      "learning_rate": 0.0004191815797143689,
+      "loss": 3.2494,
+      "step": 51800
+    },
+    {
+      "epoch": 15.103414122582148,
+      "grad_norm": 0.36541128158569336,
+      "learning_rate": 0.00041900670358496067,
+      "loss": 3.2261,
+      "step": 51850
+    },
+    {
+      "epoch": 15.117979491959916,
+      "grad_norm": 0.3625212609767914,
+      "learning_rate": 0.0004188318274555523,
+      "loss": 3.2432,
+      "step": 51900
+    },
+    {
+      "epoch": 15.132544861337683,
+      "grad_norm": 0.35310661792755127,
+      "learning_rate": 0.00041865695132614394,
+      "loss": 3.2397,
+      "step": 51950
+    },
+    {
+      "epoch": 15.147110230715452,
+      "grad_norm": 0.35972100496292114,
+      "learning_rate": 0.00041848207519673563,
+      "loss": 3.2651,
+      "step": 52000
+    },
+    {
+      "epoch": 15.147110230715452,
+      "eval_accuracy": 0.3710567502917311,
+      "eval_loss": 3.557687282562256,
+      "eval_runtime": 180.9823,
+      "eval_samples_per_second": 91.948,
+      "eval_steps_per_second": 5.752,
+      "step": 52000
+    },
+    {
+      "epoch": 15.161675600093218,
+      "grad_norm": 0.365764856338501,
+      "learning_rate": 0.00041830719906732727,
+      "loss": 3.2409,
+      "step": 52050
+    },
+    {
+      "epoch": 15.176240969470985,
+      "grad_norm": 0.36164504289627075,
+      "learning_rate": 0.00041813232293791896,
+      "loss": 3.2456,
+      "step": 52100
+    },
+    {
+      "epoch": 15.190806338848754,
+      "grad_norm": 0.3860973119735718,
+      "learning_rate": 0.0004179574468085106,
+      "loss": 3.2516,
+      "step": 52150
+    },
+    {
+      "epoch": 15.20537170822652,
+      "grad_norm": 0.39706432819366455,
+      "learning_rate": 0.00041778257067910223,
+      "loss": 3.2573,
+      "step": 52200
+    },
+    {
+      "epoch": 15.219937077604287,
+      "grad_norm": 0.3509387671947479,
+      "learning_rate": 0.0004176076945496939,
+      "loss": 3.2564,
+      "step": 52250
+    },
+    {
+      "epoch": 15.234502446982056,
+      "grad_norm": 0.39134401082992554,
+      "learning_rate": 0.00041743281842028556,
+      "loss": 3.2633,
+      "step": 52300
+    },
+    {
+      "epoch": 15.249067816359823,
+      "grad_norm": 0.4164292812347412,
+      "learning_rate": 0.0004172579422908773,
+      "loss": 3.263,
+      "step": 52350
+    },
+    {
+      "epoch": 15.263633185737591,
+      "grad_norm": 0.3817541003227234,
+      "learning_rate": 0.00041708306616146894,
+      "loss": 3.2711,
+      "step": 52400
+    },
+    {
+      "epoch": 15.278198555115358,
+      "grad_norm": 0.3729110360145569,
+      "learning_rate": 0.00041690819003206063,
+      "loss": 3.2614,
+      "step": 52450
+    },
+    {
+      "epoch": 15.292763924493125,
+      "grad_norm": 0.38113781809806824,
+      "learning_rate": 0.00041673331390265227,
+      "loss": 3.2674,
+      "step": 52500
+    },
+    {
+      "epoch": 15.307329293870893,
+      "grad_norm": 0.3448164165019989,
+      "learning_rate": 0.0004165584377732439,
+      "loss": 3.2662,
+      "step": 52550
+    },
+    {
+      "epoch": 15.32189466324866,
+      "grad_norm": 0.34779250621795654,
+      "learning_rate": 0.0004163835616438356,
+      "loss": 3.2699,
+      "step": 52600
+    },
+    {
+      "epoch": 15.336460032626427,
+      "grad_norm": 0.3631135821342468,
+      "learning_rate": 0.00041620868551442723,
+      "loss": 3.2681,
+      "step": 52650
+    },
+    {
+      "epoch": 15.351025402004195,
+      "grad_norm": 0.3445514738559723,
+      "learning_rate": 0.0004160338093850189,
+      "loss": 3.2722,
+      "step": 52700
+    },
+    {
+      "epoch": 15.365590771381962,
+      "grad_norm": 0.3631516993045807,
+      "learning_rate": 0.00041585893325561056,
+      "loss": 3.2699,
+      "step": 52750
+    },
+    {
+      "epoch": 15.38015614075973,
+      "grad_norm": 0.38682258129119873,
+      "learning_rate": 0.0004156840571262022,
+      "loss": 3.2674,
+      "step": 52800
+    },
+    {
+      "epoch": 15.394721510137497,
+      "grad_norm": 0.36580127477645874,
+      "learning_rate": 0.00041550918099679394,
+      "loss": 3.2734,
+      "step": 52850
+    },
+    {
+      "epoch": 15.409286879515264,
+      "grad_norm": 0.3573247790336609,
+      "learning_rate": 0.0004153343048673856,
+      "loss": 3.2679,
+      "step": 52900
+    },
+    {
+      "epoch": 15.423852248893033,
+      "grad_norm": 0.3864547908306122,
+      "learning_rate": 0.00041515942873797726,
+      "loss": 3.2686,
+      "step": 52950
+    },
+    {
+      "epoch": 15.4384176182708,
+      "grad_norm": 0.3987586200237274,
+      "learning_rate": 0.0004149845526085689,
+      "loss": 3.2765,
+      "step": 53000
+    },
+    {
+      "epoch": 15.4384176182708,
+      "eval_accuracy": 0.3716245131004683,
+      "eval_loss": 3.551037311553955,
+      "eval_runtime": 181.1195,
+      "eval_samples_per_second": 91.879,
+      "eval_steps_per_second": 5.748,
+      "step": 53000
+    },
+    {
+      "epoch": 15.452982987648566,
+      "grad_norm": 0.3875608742237091,
+      "learning_rate": 0.0004148096764791606,
+      "loss": 3.2837,
+      "step": 53050
+    },
+    {
+      "epoch": 15.467548357026335,
+      "grad_norm": 0.3841041624546051,
+      "learning_rate": 0.0004146348003497522,
+      "loss": 3.2813,
+      "step": 53100
+    },
+    {
+      "epoch": 15.482113726404101,
+      "grad_norm": 0.40137773752212524,
+      "learning_rate": 0.00041445992422034386,
+      "loss": 3.2902,
+      "step": 53150
+    },
+    {
+      "epoch": 15.49667909578187,
+      "grad_norm": 0.3851076662540436,
+      "learning_rate": 0.00041428504809093555,
+      "loss": 3.2801,
+      "step": 53200
+    },
+    {
+      "epoch": 15.511244465159637,
+      "grad_norm": 0.3758155107498169,
+      "learning_rate": 0.0004141101719615272,
+      "loss": 3.3033,
+      "step": 53250
+    },
+    {
+      "epoch": 15.525809834537403,
+      "grad_norm": 0.3832157254219055,
+      "learning_rate": 0.00041393529583211893,
+      "loss": 3.2831,
+      "step": 53300
+    },
+    {
+      "epoch": 15.540375203915172,
+      "grad_norm": 0.41620537638664246,
+      "learning_rate": 0.00041376041970271057,
+      "loss": 3.2932,
+      "step": 53350
+    },
+    {
+      "epoch": 15.554940573292939,
+      "grad_norm": 0.39079970121383667,
+      "learning_rate": 0.0004135855435733022,
+      "loss": 3.2954,
+      "step": 53400
+    },
+    {
+      "epoch": 15.569505942670705,
+      "grad_norm": 0.3655340075492859,
+      "learning_rate": 0.0004134106674438939,
+      "loss": 3.2992,
+      "step": 53450
+    },
+    {
+      "epoch": 15.584071312048474,
+      "grad_norm": 0.38787537813186646,
+      "learning_rate": 0.00041323579131448553,
+      "loss": 3.2933,
+      "step": 53500
+    },
+    {
+      "epoch": 15.59863668142624,
+      "grad_norm": 0.36919912695884705,
+      "learning_rate": 0.0004130609151850772,
+      "loss": 3.2779,
+      "step": 53550
+    },
+    {
+      "epoch": 15.61320205080401,
+      "grad_norm": 0.3712100684642792,
+      "learning_rate": 0.00041288603905566886,
+      "loss": 3.296,
+      "step": 53600
+    },
+    {
+      "epoch": 15.627767420181776,
+      "grad_norm": 0.3865123689174652,
+      "learning_rate": 0.0004127111629262605,
+      "loss": 3.2939,
+      "step": 53650
+    },
+    {
+      "epoch": 15.642332789559543,
+      "grad_norm": 0.3723639249801636,
+      "learning_rate": 0.0004125362867968522,
+      "loss": 3.301,
+      "step": 53700
+    },
+    {
+      "epoch": 15.656898158937311,
+      "grad_norm": 0.3666393458843231,
+      "learning_rate": 0.0004123614106674438,
+      "loss": 3.2966,
+      "step": 53750
+    },
+    {
+      "epoch": 15.671463528315078,
+      "grad_norm": 0.38118353486061096,
+      "learning_rate": 0.00041218653453803557,
+      "loss": 3.2996,
+      "step": 53800
+    },
+    {
+      "epoch": 15.686028897692845,
+      "grad_norm": 0.3505551218986511,
+      "learning_rate": 0.0004120116584086272,
+      "loss": 3.302,
+      "step": 53850
+    },
+    {
+      "epoch": 15.700594267070613,
+      "grad_norm": 0.3584941625595093,
+      "learning_rate": 0.0004118367822792189,
+      "loss": 3.3016,
+      "step": 53900
+    },
+    {
+      "epoch": 15.71515963644838,
+      "grad_norm": 0.37399595975875854,
+      "learning_rate": 0.00041166190614981053,
+      "loss": 3.2957,
+      "step": 53950
+    },
+    {
+      "epoch": 15.729725005826147,
+      "grad_norm": 0.3469371497631073,
+      "learning_rate": 0.00041148703002040217,
+      "loss": 3.3027,
+      "step": 54000
+    },
+    {
+      "epoch": 15.729725005826147,
+      "eval_accuracy": 0.3722208522063312,
+      "eval_loss": 3.543367624282837,
+      "eval_runtime": 181.1215,
+      "eval_samples_per_second": 91.878,
+      "eval_steps_per_second": 5.748,
+      "step": 54000
+    },
+    {
+      "epoch": 15.744290375203915,
+      "grad_norm": 0.37410426139831543,
+      "learning_rate": 0.00041131215389099386,
+      "loss": 3.2895,
+      "step": 54050
+    },
+    {
+      "epoch": 15.758855744581682,
+      "grad_norm": 0.3597539961338043,
+      "learning_rate": 0.0004111372777615855,
+      "loss": 3.2943,
+      "step": 54100
+    },
+    {
+      "epoch": 15.77342111395945,
+      "grad_norm": 0.3461931645870209,
+      "learning_rate": 0.0004109624016321772,
+      "loss": 3.3047,
+      "step": 54150
+    },
+    {
+      "epoch": 15.787986483337217,
+      "grad_norm": 0.37114080786705017,
+      "learning_rate": 0.0004107875255027688,
+      "loss": 3.3032,
+      "step": 54200
+    },
+    {
+      "epoch": 15.802551852714984,
+      "grad_norm": 0.3615650236606598,
+      "learning_rate": 0.00041061264937336045,
+      "loss": 3.3004,
+      "step": 54250
+    },
+    {
+      "epoch": 15.817117222092753,
+      "grad_norm": 0.34998172521591187,
+      "learning_rate": 0.0004104377732439522,
+      "loss": 3.3015,
+      "step": 54300
+    },
+    {
+      "epoch": 15.83168259147052,
+      "grad_norm": 0.4051649868488312,
+      "learning_rate": 0.00041026289711454384,
+      "loss": 3.2955,
+      "step": 54350
+    },
+    {
+      "epoch": 15.846247960848288,
+      "grad_norm": 0.3691783845424652,
+      "learning_rate": 0.0004100880209851355,
+      "loss": 3.3104,
+      "step": 54400
+    },
+    {
+      "epoch": 15.860813330226055,
+      "grad_norm": 0.3434380292892456,
+      "learning_rate": 0.00040991314485572716,
+      "loss": 3.3057,
+      "step": 54450
+    },
+    {
+      "epoch": 15.875378699603822,
+      "grad_norm": 0.372763991355896,
+      "learning_rate": 0.00040973826872631885,
+      "loss": 3.313,
+      "step": 54500
+    },
+    {
+      "epoch": 15.88994406898159,
+      "grad_norm": 0.3684810698032379,
+      "learning_rate": 0.0004095633925969105,
+      "loss": 3.3076,
+      "step": 54550
+    },
+    {
+      "epoch": 15.904509438359357,
+      "grad_norm": 0.38107824325561523,
+      "learning_rate": 0.0004093885164675021,
+      "loss": 3.2928,
+      "step": 54600
+    },
+    {
+      "epoch": 15.919074807737124,
+      "grad_norm": 0.35651078820228577,
+      "learning_rate": 0.0004092136403380938,
+      "loss": 3.3029,
+      "step": 54650
+    },
+    {
+      "epoch": 15.933640177114892,
+      "grad_norm": 0.3670404553413391,
+      "learning_rate": 0.00040903876420868545,
+      "loss": 3.3014,
+      "step": 54700
+    },
+    {
+      "epoch": 15.948205546492659,
+      "grad_norm": 0.37488341331481934,
+      "learning_rate": 0.00040886388807927714,
+      "loss": 3.3087,
+      "step": 54750
+    },
+    {
+      "epoch": 15.962770915870426,
+      "grad_norm": 0.3732067048549652,
+      "learning_rate": 0.00040868901194986883,
+      "loss": 3.3063,
+      "step": 54800
+    },
+    {
+      "epoch": 15.977336285248194,
+      "grad_norm": 0.3323022723197937,
+      "learning_rate": 0.00040851413582046047,
+      "loss": 3.3041,
+      "step": 54850
+    },
+    {
+      "epoch": 15.991901654625961,
+      "grad_norm": 0.38767266273498535,
+      "learning_rate": 0.00040833925969105216,
+      "loss": 3.3147,
+      "step": 54900
+    },
+    {
+      "epoch": 16.006408762526217,
+      "grad_norm": 0.37091103196144104,
+      "learning_rate": 0.0004081643835616438,
+      "loss": 3.2616,
+      "step": 54950
+    },
+    {
+      "epoch": 16.020974131903984,
+      "grad_norm": 0.3541449010372162,
+      "learning_rate": 0.0004079895074322355,
+      "loss": 3.2035,
+      "step": 55000
+    },
+    {
+      "epoch": 16.020974131903984,
+      "eval_accuracy": 0.3714964489540899,
+      "eval_loss": 3.555342674255371,
+      "eval_runtime": 180.9467,
+      "eval_samples_per_second": 91.966,
+      "eval_steps_per_second": 5.753,
+      "step": 55000
+    },
+    {
+      "epoch": 16.035539501281754,
+      "grad_norm": 0.3793613910675049,
+      "learning_rate": 0.0004078146313028271,
+      "loss": 3.2082,
+      "step": 55050
+    },
+    {
+      "epoch": 16.05010487065952,
+      "grad_norm": 0.4013975262641907,
+      "learning_rate": 0.0004076397551734188,
+      "loss": 3.2126,
+      "step": 55100
+    },
+    {
+      "epoch": 16.064670240037287,
+      "grad_norm": 0.3848033547401428,
+      "learning_rate": 0.00040746487904401045,
+      "loss": 3.2174,
+      "step": 55150
+    },
+    {
+      "epoch": 16.079235609415054,
+      "grad_norm": 0.39623206853866577,
+      "learning_rate": 0.0004072900029146021,
+      "loss": 3.2172,
+      "step": 55200
+    },
+    {
+      "epoch": 16.09380097879282,
+      "grad_norm": 0.3713245689868927,
+      "learning_rate": 0.0004071151267851938,
+      "loss": 3.2191,
+      "step": 55250
+    },
+    {
+      "epoch": 16.10836634817059,
+      "grad_norm": 0.38348889350891113,
+      "learning_rate": 0.00040694025065578546,
+      "loss": 3.2284,
+      "step": 55300
+    },
+    {
+      "epoch": 16.122931717548358,
+      "grad_norm": 0.37850069999694824,
+      "learning_rate": 0.00040676537452637716,
+      "loss": 3.2369,
+      "step": 55350
+    },
+    {
+      "epoch": 16.137497086926125,
+      "grad_norm": 0.38939031958580017,
+      "learning_rate": 0.0004065904983969688,
+      "loss": 3.236,
+      "step": 55400
+    },
+    {
+      "epoch": 16.15206245630389,
+      "grad_norm": 0.39782270789146423,
+      "learning_rate": 0.00040641562226756043,
+      "loss": 3.2381,
+      "step": 55450
+    },
+    {
+      "epoch": 16.16662782568166,
+      "grad_norm": 0.39229854941368103,
+      "learning_rate": 0.0004062407461381521,
+      "loss": 3.2319,
+      "step": 55500
+    },
+    {
+      "epoch": 16.181193195059425,
+      "grad_norm": 0.379708856344223,
+      "learning_rate": 0.00040606587000874375,
+      "loss": 3.2324,
+      "step": 55550
+    },
+    {
+      "epoch": 16.195758564437195,
+      "grad_norm": 0.3870663344860077,
+      "learning_rate": 0.00040589099387933544,
+      "loss": 3.2388,
+      "step": 55600
+    },
+    {
+      "epoch": 16.210323933814962,
+      "grad_norm": 0.36951544880867004,
+      "learning_rate": 0.0004057161177499271,
+      "loss": 3.2338,
+      "step": 55650
+    },
+    {
+      "epoch": 16.22488930319273,
+      "grad_norm": 0.3752133548259735,
+      "learning_rate": 0.0004055412416205187,
+      "loss": 3.253,
+      "step": 55700
+    },
+    {
+      "epoch": 16.239454672570496,
+      "grad_norm": 0.39170658588409424,
+      "learning_rate": 0.0004053663654911104,
+      "loss": 3.2319,
+      "step": 55750
+    },
+    {
+      "epoch": 16.254020041948262,
+      "grad_norm": 0.3717292845249176,
+      "learning_rate": 0.0004051914893617021,
+      "loss": 3.2366,
+      "step": 55800
+    },
+    {
+      "epoch": 16.268585411326033,
+      "grad_norm": 0.3885873556137085,
+      "learning_rate": 0.0004050166132322938,
+      "loss": 3.2377,
+      "step": 55850
+    },
+    {
+      "epoch": 16.2831507807038,
+      "grad_norm": 0.3768942058086395,
+      "learning_rate": 0.0004048417371028854,
+      "loss": 3.2438,
+      "step": 55900
+    },
+    {
+      "epoch": 16.297716150081566,
+      "grad_norm": 0.3943030536174774,
+      "learning_rate": 0.0004046668609734771,
+      "loss": 3.25,
+      "step": 55950
+    },
+    {
+      "epoch": 16.312281519459333,
+      "grad_norm": 0.38262632489204407,
+      "learning_rate": 0.00040449198484406875,
+      "loss": 3.2571,
+      "step": 56000
+    },
+    {
+      "epoch": 16.312281519459333,
+      "eval_accuracy": 0.37204022178499313,
+      "eval_loss": 3.552259922027588,
+      "eval_runtime": 181.0729,
+      "eval_samples_per_second": 91.902,
+      "eval_steps_per_second": 5.749,
+      "step": 56000
+    },
+    {
+      "epoch": 16.3268468888371,
+      "grad_norm": 0.37317734956741333,
+      "learning_rate": 0.0004043171087146604,
+      "loss": 3.2472,
+      "step": 56050
+    },
+    {
+      "epoch": 16.34141225821487,
+      "grad_norm": 0.39976489543914795,
+      "learning_rate": 0.0004041422325852521,
+      "loss": 3.2512,
+      "step": 56100
+    },
+    {
+      "epoch": 16.355977627592637,
+      "grad_norm": 0.3861803412437439,
+      "learning_rate": 0.0004039673564558437,
+      "loss": 3.2619,
+      "step": 56150
+    },
+    {
+      "epoch": 16.370542996970403,
+      "grad_norm": 0.39547422528266907,
+      "learning_rate": 0.0004037924803264354,
+      "loss": 3.2572,
+      "step": 56200
+    },
+    {
+      "epoch": 16.38510836634817,
+      "grad_norm": 0.3525276184082031,
+      "learning_rate": 0.00040361760419702704,
+      "loss": 3.2686,
+      "step": 56250
+    },
+    {
+      "epoch": 16.399673735725937,
+      "grad_norm": 0.3637978434562683,
+      "learning_rate": 0.00040344272806761873,
+      "loss": 3.2522,
+      "step": 56300
+    },
+    {
+      "epoch": 16.414239105103704,
+      "grad_norm": 0.39342567324638367,
+      "learning_rate": 0.0004032678519382104,
+      "loss": 3.2612,
+      "step": 56350
+    },
+    {
+      "epoch": 16.428804474481474,
+      "grad_norm": 0.3710050582885742,
+      "learning_rate": 0.00040309297580880206,
+      "loss": 3.2516,
+      "step": 56400
+    },
+    {
+      "epoch": 16.44336984385924,
+      "grad_norm": 0.37674885988235474,
+      "learning_rate": 0.00040291809967939375,
+      "loss": 3.2625,
+      "step": 56450
+    },
+    {
+      "epoch": 16.457935213237008,
+      "grad_norm": 0.3522658944129944,
+      "learning_rate": 0.0004027432235499854,
+      "loss": 3.266,
+      "step": 56500
+    },
+    {
+      "epoch": 16.472500582614774,
+      "grad_norm": 0.3829957842826843,
+      "learning_rate": 0.0004025683474205771,
+      "loss": 3.2711,
+      "step": 56550
+    },
+    {
+      "epoch": 16.48706595199254,
+      "grad_norm": 0.3912467062473297,
+      "learning_rate": 0.0004023934712911687,
+      "loss": 3.2626,
+      "step": 56600
+    },
+    {
+      "epoch": 16.50163132137031,
+      "grad_norm": 0.36625564098358154,
+      "learning_rate": 0.00040221859516176035,
+      "loss": 3.2808,
+      "step": 56650
+    },
+    {
+      "epoch": 16.516196690748078,
+      "grad_norm": 0.345903605222702,
+      "learning_rate": 0.00040204371903235204,
+      "loss": 3.2792,
+      "step": 56700
+    },
+    {
+      "epoch": 16.530762060125845,
+      "grad_norm": 0.398399293422699,
+      "learning_rate": 0.0004018688429029437,
+      "loss": 3.27,
+      "step": 56750
+    },
+    {
+      "epoch": 16.54532742950361,
+      "grad_norm": 0.3665493130683899,
+      "learning_rate": 0.0004016939667735354,
+      "loss": 3.2772,
+      "step": 56800
+    },
+    {
+      "epoch": 16.55989279888138,
+      "grad_norm": 0.37150660157203674,
+      "learning_rate": 0.00040151909064412705,
+      "loss": 3.278,
+      "step": 56850
+    },
+    {
+      "epoch": 16.57445816825915,
+      "grad_norm": 0.37004363536834717,
+      "learning_rate": 0.0004013442145147187,
+      "loss": 3.2706,
+      "step": 56900
+    },
+    {
+      "epoch": 16.589023537636916,
+      "grad_norm": 0.373983234167099,
+      "learning_rate": 0.0004011693383853104,
+      "loss": 3.2701,
+      "step": 56950
+    },
+    {
+      "epoch": 16.603588907014682,
+      "grad_norm": 0.3955393135547638,
+      "learning_rate": 0.000400994462255902,
+      "loss": 3.2808,
+      "step": 57000
+    },
+    {
+      "epoch": 16.603588907014682,
+      "eval_accuracy": 0.3721092517702311,
+      "eval_loss": 3.5462496280670166,
+      "eval_runtime": 181.037,
+      "eval_samples_per_second": 91.92,
+      "eval_steps_per_second": 5.75,
+      "step": 57000
+    },
+    {
+      "epoch": 16.61815427639245,
+      "grad_norm": 0.3694080710411072,
+      "learning_rate": 0.0004008195861264937,
+      "loss": 3.2767,
+      "step": 57050
+    },
+    {
+      "epoch": 16.632719645770216,
+      "grad_norm": 0.37952783703804016,
+      "learning_rate": 0.00040064470999708534,
+      "loss": 3.2749,
+      "step": 57100
+    },
+    {
+      "epoch": 16.647285015147983,
+      "grad_norm": 0.37776124477386475,
+      "learning_rate": 0.00040046983386767703,
+      "loss": 3.2817,
+      "step": 57150
+    },
+    {
+      "epoch": 16.661850384525753,
+      "grad_norm": 0.36142295598983765,
+      "learning_rate": 0.00040029495773826867,
+      "loss": 3.283,
+      "step": 57200
+    },
+    {
+      "epoch": 16.67641575390352,
+      "grad_norm": 0.42377597093582153,
+      "learning_rate": 0.0004001200816088603,
+      "loss": 3.2864,
+      "step": 57250
+    },
+    {
+      "epoch": 16.690981123281286,
+      "grad_norm": 0.3531087636947632,
+      "learning_rate": 0.00039994520547945205,
+      "loss": 3.2862,
+      "step": 57300
+    },
+    {
+      "epoch": 16.705546492659053,
+      "grad_norm": 0.3619745075702667,
+      "learning_rate": 0.0003997703293500437,
+      "loss": 3.2888,
+      "step": 57350
+    },
+    {
+      "epoch": 16.72011186203682,
+      "grad_norm": 0.36244842410087585,
+      "learning_rate": 0.0003995954532206354,
+      "loss": 3.2967,
+      "step": 57400
+    },
+    {
+      "epoch": 16.73467723141459,
+      "grad_norm": 0.37257516384124756,
+      "learning_rate": 0.000399420577091227,
+      "loss": 3.2953,
+      "step": 57450
+    },
+    {
+      "epoch": 16.749242600792357,
+      "grad_norm": 0.3775019943714142,
+      "learning_rate": 0.00039924570096181865,
+      "loss": 3.2896,
+      "step": 57500
+    },
+    {
+      "epoch": 16.763807970170124,
+      "grad_norm": 0.3789142072200775,
+      "learning_rate": 0.00039907082483241034,
+      "loss": 3.2906,
+      "step": 57550
+    },
+    {
+      "epoch": 16.77837333954789,
+      "grad_norm": 0.36364027857780457,
+      "learning_rate": 0.000398895948703002,
+      "loss": 3.2832,
+      "step": 57600
+    },
+    {
+      "epoch": 16.792938708925657,
+      "grad_norm": 0.33775755763053894,
+      "learning_rate": 0.00039872107257359367,
+      "loss": 3.2906,
+      "step": 57650
+    },
+    {
+      "epoch": 16.807504078303424,
+      "grad_norm": 0.3706631362438202,
+      "learning_rate": 0.0003985461964441853,
+      "loss": 3.2908,
+      "step": 57700
+    },
+    {
+      "epoch": 16.822069447681194,
+      "grad_norm": 0.39492863416671753,
+      "learning_rate": 0.00039837132031477694,
+      "loss": 3.2892,
+      "step": 57750
+    },
+    {
+      "epoch": 16.83663481705896,
+      "grad_norm": 0.41402751207351685,
+      "learning_rate": 0.0003981964441853687,
+      "loss": 3.2873,
+      "step": 57800
+    },
+    {
+      "epoch": 16.851200186436728,
+      "grad_norm": 0.3844890892505646,
+      "learning_rate": 0.0003980215680559603,
+      "loss": 3.2812,
+      "step": 57850
+    },
+    {
+      "epoch": 16.865765555814495,
+      "grad_norm": 0.35271942615509033,
+      "learning_rate": 0.000397846691926552,
+      "loss": 3.302,
+      "step": 57900
+    },
+    {
+      "epoch": 16.88033092519226,
+      "grad_norm": 0.3879646062850952,
+      "learning_rate": 0.00039767181579714365,
+      "loss": 3.306,
+      "step": 57950
+    },
+    {
+      "epoch": 16.89489629457003,
+      "grad_norm": 0.3731992244720459,
+      "learning_rate": 0.00039749693966773534,
+      "loss": 3.2875,
+      "step": 58000
+    },
+    {
+      "epoch": 16.89489629457003,
+      "eval_accuracy": 0.3725817602552157,
+      "eval_loss": 3.5377631187438965,
+      "eval_runtime": 180.9279,
+      "eval_samples_per_second": 91.976,
+      "eval_steps_per_second": 5.754,
+      "step": 58000
+    },
+    {
+      "epoch": 16.9094616639478,
+      "grad_norm": 0.35738855600357056,
+      "learning_rate": 0.00039732206353832697,
+      "loss": 3.2992,
+      "step": 58050
+    },
+    {
+      "epoch": 16.924027033325565,
+      "grad_norm": 0.37029650807380676,
+      "learning_rate": 0.0003971471874089186,
+      "loss": 3.2952,
+      "step": 58100
+    },
+    {
+      "epoch": 16.938592402703332,
+      "grad_norm": 0.37805014848709106,
+      "learning_rate": 0.0003969723112795103,
+      "loss": 3.2882,
+      "step": 58150
+    },
+    {
+      "epoch": 16.9531577720811,
+      "grad_norm": 0.37777912616729736,
+      "learning_rate": 0.00039679743515010194,
+      "loss": 3.2945,
+      "step": 58200
+    },
+    {
+      "epoch": 16.96772314145887,
+      "grad_norm": 0.3731626570224762,
+      "learning_rate": 0.0003966225590206937,
+      "loss": 3.2981,
+      "step": 58250
+    },
+    {
+      "epoch": 16.982288510836636,
+      "grad_norm": 0.3958134055137634,
+      "learning_rate": 0.0003964476828912853,
+      "loss": 3.2938,
+      "step": 58300
+    },
+    {
+      "epoch": 16.996853880214402,
+      "grad_norm": 0.3792884051799774,
+      "learning_rate": 0.00039627280676187695,
+      "loss": 3.3124,
+      "step": 58350
+    },
+    {
+      "epoch": 17.01136098811466,
+      "grad_norm": 0.3800676167011261,
+      "learning_rate": 0.00039609793063246864,
+      "loss": 3.2102,
+      "step": 58400
+    },
+    {
+      "epoch": 17.025926357492427,
+      "grad_norm": 0.39205148816108704,
+      "learning_rate": 0.0003959230545030603,
+      "loss": 3.1951,
+      "step": 58450
+    },
+    {
+      "epoch": 17.040491726870194,
+      "grad_norm": 0.35533279180526733,
+      "learning_rate": 0.00039574817837365197,
+      "loss": 3.1817,
+      "step": 58500
+    },
+    {
+      "epoch": 17.05505709624796,
+      "grad_norm": 0.3837513327598572,
+      "learning_rate": 0.0003955733022442436,
+      "loss": 3.2115,
+      "step": 58550
+    },
+    {
+      "epoch": 17.069622465625727,
+      "grad_norm": 0.40589404106140137,
+      "learning_rate": 0.0003953984261148353,
+      "loss": 3.2004,
+      "step": 58600
+    },
+    {
+      "epoch": 17.084187835003497,
+      "grad_norm": 0.3653603494167328,
+      "learning_rate": 0.00039522354998542693,
+      "loss": 3.2003,
+      "step": 58650
+    },
+    {
+      "epoch": 17.098753204381264,
+      "grad_norm": 0.4216572940349579,
+      "learning_rate": 0.00039504867385601857,
+      "loss": 3.1984,
+      "step": 58700
+    },
+    {
+      "epoch": 17.11331857375903,
+      "grad_norm": 0.37719815969467163,
+      "learning_rate": 0.0003948737977266103,
+      "loss": 3.2251,
+      "step": 58750
+    },
+    {
+      "epoch": 17.127883943136798,
+      "grad_norm": 0.3737868666648865,
+      "learning_rate": 0.00039469892159720195,
+      "loss": 3.2223,
+      "step": 58800
+    },
+    {
+      "epoch": 17.142449312514564,
+      "grad_norm": 0.42152318358421326,
+      "learning_rate": 0.00039452404546779364,
+      "loss": 3.2086,
+      "step": 58850
+    },
+    {
+      "epoch": 17.15701468189233,
+      "grad_norm": 0.3755936324596405,
+      "learning_rate": 0.0003943491693383853,
+      "loss": 3.2352,
+      "step": 58900
+    },
+    {
+      "epoch": 17.1715800512701,
+      "grad_norm": 0.3840259313583374,
+      "learning_rate": 0.0003941742932089769,
+      "loss": 3.2167,
+      "step": 58950
+    },
+    {
+      "epoch": 17.18614542064787,
+      "grad_norm": 0.41081106662750244,
+      "learning_rate": 0.0003939994170795686,
+      "loss": 3.2199,
+      "step": 59000
+    },
+    {
+      "epoch": 17.18614542064787,
+      "eval_accuracy": 0.37207597155588296,
+      "eval_loss": 3.552147626876831,
+      "eval_runtime": 180.9159,
+      "eval_samples_per_second": 91.982,
+      "eval_steps_per_second": 5.754,
+      "step": 59000
+    },
+    {
+      "epoch": 17.200710790025635,
+      "grad_norm": 0.3771253228187561,
+      "learning_rate": 0.00039382454095016024,
+      "loss": 3.2284,
+      "step": 59050
+    },
+    {
+      "epoch": 17.215276159403402,
+      "grad_norm": 0.3621974587440491,
+      "learning_rate": 0.00039364966482075193,
+      "loss": 3.2189,
+      "step": 59100
+    },
+    {
+      "epoch": 17.22984152878117,
+      "grad_norm": 0.3764497637748718,
+      "learning_rate": 0.00039347478869134356,
+      "loss": 3.229,
+      "step": 59150
+    },
+    {
+      "epoch": 17.24440689815894,
+      "grad_norm": 0.3832677900791168,
+      "learning_rate": 0.0003932999125619353,
+      "loss": 3.2331,
+      "step": 59200
+    },
+    {
+      "epoch": 17.258972267536706,
+      "grad_norm": 0.36371684074401855,
+      "learning_rate": 0.00039312503643252695,
+      "loss": 3.2278,
+      "step": 59250
+    },
+    {
+      "epoch": 17.273537636914472,
+      "grad_norm": 0.36630338430404663,
+      "learning_rate": 0.0003929501603031186,
+      "loss": 3.239,
+      "step": 59300
+    },
+    {
+      "epoch": 17.28810300629224,
+      "grad_norm": 0.3638499975204468,
+      "learning_rate": 0.00039277528417371027,
+      "loss": 3.2384,
+      "step": 59350
+    },
+    {
+      "epoch": 17.302668375670006,
+      "grad_norm": 0.369156152009964,
+      "learning_rate": 0.0003926004080443019,
+      "loss": 3.2434,
+      "step": 59400
+    },
+    {
+      "epoch": 17.317233745047773,
+      "grad_norm": 0.3857540786266327,
+      "learning_rate": 0.0003924255319148936,
+      "loss": 3.2502,
+      "step": 59450
+    },
+    {
+      "epoch": 17.331799114425543,
+      "grad_norm": 0.3861338198184967,
+      "learning_rate": 0.00039225065578548523,
+      "loss": 3.2288,
+      "step": 59500
+    },
+    {
+      "epoch": 17.34636448380331,
+      "grad_norm": 0.3690037131309509,
+      "learning_rate": 0.00039207577965607687,
+      "loss": 3.2342,
+      "step": 59550
+    },
+    {
+      "epoch": 17.360929853181077,
+      "grad_norm": 0.38043132424354553,
+      "learning_rate": 0.00039190090352666856,
+      "loss": 3.2399,
+      "step": 59600
+    },
+    {
+      "epoch": 17.375495222558843,
+      "grad_norm": 0.38716545701026917,
+      "learning_rate": 0.0003917260273972602,
+      "loss": 3.251,
+      "step": 59650
+    },
+    {
+      "epoch": 17.39006059193661,
+      "grad_norm": 0.3937844932079315,
+      "learning_rate": 0.00039155115126785194,
+      "loss": 3.2406,
+      "step": 59700
+    },
+    {
+      "epoch": 17.40462596131438,
+      "grad_norm": 0.3667075037956238,
+      "learning_rate": 0.0003913762751384436,
+      "loss": 3.2496,
+      "step": 59750
+    },
+    {
+      "epoch": 17.419191330692147,
+      "grad_norm": 0.38368871808052063,
+      "learning_rate": 0.00039120139900903527,
+      "loss": 3.2542,
+      "step": 59800
+    },
+    {
+      "epoch": 17.433756700069914,
+      "grad_norm": 0.3750975728034973,
+      "learning_rate": 0.0003910265228796269,
+      "loss": 3.2565,
+      "step": 59850
+    },
+    {
+      "epoch": 17.44832206944768,
+      "grad_norm": 0.3607097268104553,
+      "learning_rate": 0.00039085164675021854,
+      "loss": 3.2575,
+      "step": 59900
+    },
+    {
+      "epoch": 17.462887438825447,
+      "grad_norm": 0.36779287457466125,
+      "learning_rate": 0.00039067677062081023,
+      "loss": 3.2499,
+      "step": 59950
+    },
+    {
+      "epoch": 17.477452808203218,
+      "grad_norm": 0.3928638696670532,
+      "learning_rate": 0.00039050189449140187,
+      "loss": 3.2427,
+      "step": 60000
+    },
+    {
+      "epoch": 17.477452808203218,
+      "eval_accuracy": 0.3728227184149304,
+      "eval_loss": 3.5461106300354004,
+      "eval_runtime": 180.899,
+      "eval_samples_per_second": 91.991,
+      "eval_steps_per_second": 5.755,
+      "step": 60000
+    },
+    {
+      "epoch": 17.492018177580984,
+      "grad_norm": 0.3889346718788147,
+      "learning_rate": 0.00039032701836199356,
+      "loss": 3.2534,
+      "step": 60050
+    },
+    {
+      "epoch": 17.50658354695875,
+      "grad_norm": 0.3726585805416107,
+      "learning_rate": 0.0003901521422325852,
+      "loss": 3.264,
+      "step": 60100
+    },
+    {
+      "epoch": 17.521148916336518,
+      "grad_norm": 0.38900190591812134,
+      "learning_rate": 0.00038997726610317683,
+      "loss": 3.2553,
+      "step": 60150
+    },
+    {
+      "epoch": 17.535714285714285,
+      "grad_norm": 0.38281652331352234,
+      "learning_rate": 0.0003898023899737686,
+      "loss": 3.2629,
+      "step": 60200
+    },
+    {
+      "epoch": 17.55027965509205,
+      "grad_norm": 0.40601515769958496,
+      "learning_rate": 0.0003896275138443602,
+      "loss": 3.2761,
+      "step": 60250
+    },
+    {
+      "epoch": 17.56484502446982,
+      "grad_norm": 0.3916570842266083,
+      "learning_rate": 0.0003894526377149519,
+      "loss": 3.2667,
+      "step": 60300
+    },
+    {
+      "epoch": 17.57941039384759,
+      "grad_norm": 0.3838474154472351,
+      "learning_rate": 0.00038927776158554354,
+      "loss": 3.2703,
+      "step": 60350
+    },
+    {
+      "epoch": 17.593975763225355,
+      "grad_norm": 0.36894601583480835,
+      "learning_rate": 0.0003891028854561352,
+      "loss": 3.2631,
+      "step": 60400
+    },
+    {
+      "epoch": 17.608541132603122,
+      "grad_norm": 0.37180987000465393,
+      "learning_rate": 0.00038892800932672686,
+      "loss": 3.2787,
+      "step": 60450
+    },
+    {
+      "epoch": 17.62310650198089,
+      "grad_norm": 0.3622785210609436,
+      "learning_rate": 0.0003887531331973185,
+      "loss": 3.2679,
+      "step": 60500
+    },
+    {
+      "epoch": 17.63767187135866,
+      "grad_norm": 0.4070373475551605,
+      "learning_rate": 0.0003885782570679102,
+      "loss": 3.2623,
+      "step": 60550
+    },
+    {
+      "epoch": 17.652237240736426,
+      "grad_norm": 0.40512874722480774,
+      "learning_rate": 0.0003884033809385018,
+      "loss": 3.2693,
+      "step": 60600
+    },
+    {
+      "epoch": 17.666802610114193,
+      "grad_norm": 0.3866254985332489,
+      "learning_rate": 0.00038822850480909357,
+      "loss": 3.27,
+      "step": 60650
+    },
+    {
+      "epoch": 17.68136797949196,
+      "grad_norm": 0.3540954887866974,
+      "learning_rate": 0.0003880536286796852,
+      "loss": 3.2668,
+      "step": 60700
+    },
+    {
+      "epoch": 17.695933348869726,
+      "grad_norm": 0.3930191695690155,
+      "learning_rate": 0.00038787875255027684,
+      "loss": 3.2668,
+      "step": 60750
+    },
+    {
+      "epoch": 17.710498718247496,
+      "grad_norm": 0.3699732720851898,
+      "learning_rate": 0.00038770387642086853,
+      "loss": 3.2756,
+      "step": 60800
+    },
+    {
+      "epoch": 17.725064087625263,
+      "grad_norm": 0.3773103952407837,
+      "learning_rate": 0.00038752900029146017,
+      "loss": 3.2764,
+      "step": 60850
+    },
+    {
+      "epoch": 17.73962945700303,
+      "grad_norm": 0.38592496514320374,
+      "learning_rate": 0.00038735412416205186,
+      "loss": 3.2725,
+      "step": 60900
+    },
+    {
+      "epoch": 17.754194826380797,
+      "grad_norm": 0.37976184487342834,
+      "learning_rate": 0.0003871792480326435,
+      "loss": 3.269,
+      "step": 60950
+    },
+    {
+      "epoch": 17.768760195758563,
+      "grad_norm": 0.41382166743278503,
+      "learning_rate": 0.00038700437190323513,
+      "loss": 3.2725,
+      "step": 61000
+    },
+    {
+      "epoch": 17.768760195758563,
+      "eval_accuracy": 0.3728652888657927,
+      "eval_loss": 3.539224863052368,
+      "eval_runtime": 180.9152,
+      "eval_samples_per_second": 91.982,
+      "eval_steps_per_second": 5.754,
+      "step": 61000
+    },
+    {
+      "epoch": 17.78332556513633,
+      "grad_norm": 0.37411078810691833,
+      "learning_rate": 0.0003868294957738268,
+      "loss": 3.28,
+      "step": 61050
+    },
+    {
+      "epoch": 17.7978909345141,
+      "grad_norm": 0.36364153027534485,
+      "learning_rate": 0.00038665461964441846,
+      "loss": 3.2845,
+      "step": 61100
+    },
+    {
+      "epoch": 17.812456303891867,
+      "grad_norm": 0.3570358157157898,
+      "learning_rate": 0.0003864797435150102,
+      "loss": 3.2845,
+      "step": 61150
+    },
+    {
+      "epoch": 17.827021673269634,
+      "grad_norm": 0.37199866771698,
+      "learning_rate": 0.00038630486738560184,
+      "loss": 3.2909,
+      "step": 61200
+    },
+    {
+      "epoch": 17.8415870426474,
+      "grad_norm": 0.39399829506874084,
+      "learning_rate": 0.00038612999125619353,
+      "loss": 3.271,
+      "step": 61250
+    },
+    {
+      "epoch": 17.856152412025168,
+      "grad_norm": 0.3831261396408081,
+      "learning_rate": 0.00038595511512678517,
+      "loss": 3.2751,
+      "step": 61300
+    },
+    {
+      "epoch": 17.870717781402938,
+      "grad_norm": 0.3945658504962921,
+      "learning_rate": 0.0003857802389973768,
+      "loss": 3.2822,
+      "step": 61350
+    },
+    {
+      "epoch": 17.885283150780705,
+      "grad_norm": 0.36668267846107483,
+      "learning_rate": 0.0003856053628679685,
+      "loss": 3.2872,
+      "step": 61400
+    },
+    {
+      "epoch": 17.89984852015847,
+      "grad_norm": 0.4171694219112396,
+      "learning_rate": 0.00038543048673856013,
+      "loss": 3.2733,
+      "step": 61450
+    },
+    {
+      "epoch": 17.914413889536238,
+      "grad_norm": 0.38371917605400085,
+      "learning_rate": 0.0003852556106091518,
+      "loss": 3.2823,
+      "step": 61500
+    },
+    {
+      "epoch": 17.928979258914005,
+      "grad_norm": 0.3641377389431,
+      "learning_rate": 0.00038508073447974346,
+      "loss": 3.2867,
+      "step": 61550
+    },
+    {
+      "epoch": 17.943544628291775,
+      "grad_norm": 0.36372044682502747,
+      "learning_rate": 0.0003849058583503351,
+      "loss": 3.2991,
+      "step": 61600
+    },
+    {
+      "epoch": 17.958109997669542,
+      "grad_norm": 0.3715762794017792,
+      "learning_rate": 0.00038473098222092684,
+      "loss": 3.2819,
+      "step": 61650
+    },
+    {
+      "epoch": 17.97267536704731,
+      "grad_norm": 0.3832816183567047,
+      "learning_rate": 0.0003845561060915185,
+      "loss": 3.2944,
+      "step": 61700
+    },
+    {
+      "epoch": 17.987240736425075,
+      "grad_norm": 0.389822393655777,
+      "learning_rate": 0.00038438122996211016,
+      "loss": 3.2822,
+      "step": 61750
+    },
+    {
+      "epoch": 18.001747844325333,
+      "grad_norm": 0.3840530812740326,
+      "learning_rate": 0.0003842063538327018,
+      "loss": 3.2846,
+      "step": 61800
+    },
+    {
+      "epoch": 18.0163132137031,
+      "grad_norm": 0.3474888205528259,
+      "learning_rate": 0.0003840314777032935,
+      "loss": 3.1638,
+      "step": 61850
+    },
+    {
+      "epoch": 18.030878583080867,
+      "grad_norm": 0.37700438499450684,
+      "learning_rate": 0.0003838566015738851,
+      "loss": 3.1834,
+      "step": 61900
+    },
+    {
+      "epoch": 18.045443952458633,
+      "grad_norm": 0.38227853178977966,
+      "learning_rate": 0.00038368172544447676,
+      "loss": 3.1793,
+      "step": 61950
+    },
+    {
+      "epoch": 18.0600093218364,
+      "grad_norm": 0.4074285328388214,
+      "learning_rate": 0.00038350684931506845,
+      "loss": 3.1907,
+      "step": 62000
+    },
+    {
+      "epoch": 18.0600093218364,
+      "eval_accuracy": 0.3722658922137352,
+      "eval_loss": 3.5532360076904297,
+      "eval_runtime": 180.9589,
+      "eval_samples_per_second": 91.96,
+      "eval_steps_per_second": 5.753,
+      "step": 62000
+    },
+    {
+      "epoch": 18.07457469121417,
+      "grad_norm": 0.3638034164905548,
+      "learning_rate": 0.0003833319731856601,
+      "loss": 3.1941,
+      "step": 62050
+    },
+    {
+      "epoch": 18.089140060591937,
+      "grad_norm": 0.3989737033843994,
+      "learning_rate": 0.00038315709705625183,
+      "loss": 3.1955,
+      "step": 62100
+    },
+    {
+      "epoch": 18.103705429969704,
+      "grad_norm": 0.37223362922668457,
+      "learning_rate": 0.00038298222092684347,
+      "loss": 3.1977,
+      "step": 62150
+    },
+    {
+      "epoch": 18.11827079934747,
+      "grad_norm": 0.37027719616889954,
+      "learning_rate": 0.0003828073447974351,
+      "loss": 3.209,
+      "step": 62200
+    },
+    {
+      "epoch": 18.132836168725238,
+      "grad_norm": 0.3753957748413086,
+      "learning_rate": 0.0003826324686680268,
+      "loss": 3.2066,
+      "step": 62250
+    },
+    {
+      "epoch": 18.147401538103008,
+      "grad_norm": 0.4126310646533966,
+      "learning_rate": 0.00038245759253861843,
+      "loss": 3.216,
+      "step": 62300
+    },
+    {
+      "epoch": 18.161966907480775,
+      "grad_norm": 0.36274200677871704,
+      "learning_rate": 0.0003822827164092101,
+      "loss": 3.2134,
+      "step": 62350
+    },
+    {
+      "epoch": 18.17653227685854,
+      "grad_norm": 0.39028841257095337,
+      "learning_rate": 0.00038210784027980176,
+      "loss": 3.2191,
+      "step": 62400
+    },
+    {
+      "epoch": 18.191097646236308,
+      "grad_norm": 0.36960962414741516,
+      "learning_rate": 0.0003819329641503934,
+      "loss": 3.203,
+      "step": 62450
+    },
+    {
+      "epoch": 18.205663015614075,
+      "grad_norm": 0.3780011236667633,
+      "learning_rate": 0.0003817580880209851,
+      "loss": 3.2187,
+      "step": 62500
+    },
+    {
+      "epoch": 18.22022838499184,
+      "grad_norm": 0.39886778593063354,
+      "learning_rate": 0.0003815832118915767,
+      "loss": 3.2112,
+      "step": 62550
+    },
+    {
+      "epoch": 18.234793754369612,
+      "grad_norm": 0.3542962372303009,
+      "learning_rate": 0.00038140833576216847,
+      "loss": 3.2226,
+      "step": 62600
+    },
+    {
+      "epoch": 18.24935912374738,
+      "grad_norm": 0.40383675694465637,
+      "learning_rate": 0.0003812334596327601,
+      "loss": 3.2237,
+      "step": 62650
+    },
+    {
+      "epoch": 18.263924493125145,
+      "grad_norm": 0.38676130771636963,
+      "learning_rate": 0.0003810585835033518,
+      "loss": 3.221,
+      "step": 62700
+    },
+    {
+      "epoch": 18.278489862502912,
+      "grad_norm": 0.37706777453422546,
+      "learning_rate": 0.00038088370737394343,
+      "loss": 3.2249,
+      "step": 62750
+    },
+    {
+      "epoch": 18.29305523188068,
+      "grad_norm": 0.38657376170158386,
+      "learning_rate": 0.00038070883124453507,
+      "loss": 3.2229,
+      "step": 62800
+    },
+    {
+      "epoch": 18.30762060125845,
+      "grad_norm": 0.38616442680358887,
+      "learning_rate": 0.00038053395511512676,
+      "loss": 3.2261,
+      "step": 62850
+    },
+    {
+      "epoch": 18.322185970636216,
+      "grad_norm": 0.3826325237751007,
+      "learning_rate": 0.0003803590789857184,
+      "loss": 3.229,
+      "step": 62900
+    },
+    {
+      "epoch": 18.336751340013983,
+      "grad_norm": 0.40408405661582947,
+      "learning_rate": 0.0003801842028563101,
+      "loss": 3.2361,
+      "step": 62950
+    },
+    {
+      "epoch": 18.35131670939175,
+      "grad_norm": 0.3674256205558777,
+      "learning_rate": 0.0003800093267269017,
+      "loss": 3.2403,
+      "step": 63000
+    },
+    {
+      "epoch": 18.35131670939175,
+      "eval_accuracy": 0.3724525201295318,
+      "eval_loss": 3.550670623779297,
+      "eval_runtime": 180.8745,
+      "eval_samples_per_second": 92.003,
+      "eval_steps_per_second": 5.755,
+      "step": 63000
+    },
+    {
+      "epoch": 18.365882078769516,
+      "grad_norm": 0.3731068968772888,
+      "learning_rate": 0.00037983445059749335,
+      "loss": 3.2327,
+      "step": 63050
+    },
+    {
+      "epoch": 18.380447448147287,
+      "grad_norm": 0.3978533148765564,
+      "learning_rate": 0.0003796595744680851,
+      "loss": 3.2407,
+      "step": 63100
+    },
+    {
+      "epoch": 18.395012817525053,
+      "grad_norm": 0.38433948159217834,
+      "learning_rate": 0.00037948469833867674,
+      "loss": 3.2447,
+      "step": 63150
+    },
+    {
+      "epoch": 18.40957818690282,
+      "grad_norm": 0.40469464659690857,
+      "learning_rate": 0.0003793098222092684,
+      "loss": 3.2547,
+      "step": 63200
+    },
+    {
+      "epoch": 18.424143556280587,
+      "grad_norm": 0.3718130886554718,
+      "learning_rate": 0.00037913494607986006,
+      "loss": 3.2384,
+      "step": 63250
+    },
+    {
+      "epoch": 18.438708925658354,
+      "grad_norm": 0.3753129839897156,
+      "learning_rate": 0.00037896006995045175,
+      "loss": 3.2382,
+      "step": 63300
+    },
+    {
+      "epoch": 18.45327429503612,
+      "grad_norm": 0.41086849570274353,
+      "learning_rate": 0.0003787851938210434,
+      "loss": 3.2436,
+      "step": 63350
+    },
+    {
+      "epoch": 18.46783966441389,
+      "grad_norm": 0.3825591802597046,
+      "learning_rate": 0.000378610317691635,
+      "loss": 3.2377,
+      "step": 63400
+    },
+    {
+      "epoch": 18.482405033791657,
+      "grad_norm": 0.3936189115047455,
+      "learning_rate": 0.0003784354415622267,
+      "loss": 3.2444,
+      "step": 63450
+    },
+    {
+      "epoch": 18.496970403169424,
+      "grad_norm": 0.36532971262931824,
+      "learning_rate": 0.00037826056543281835,
+      "loss": 3.2549,
+      "step": 63500
+    },
+    {
+      "epoch": 18.51153577254719,
+      "grad_norm": 0.3695519268512726,
+      "learning_rate": 0.0003780856893034101,
+      "loss": 3.2458,
+      "step": 63550
+    },
+    {
+      "epoch": 18.526101141924958,
+      "grad_norm": 0.3830970525741577,
+      "learning_rate": 0.00037791081317400173,
+      "loss": 3.242,
+      "step": 63600
+    },
+    {
+      "epoch": 18.540666511302728,
+      "grad_norm": 0.36471477150917053,
+      "learning_rate": 0.00037773593704459337,
+      "loss": 3.2499,
+      "step": 63650
+    },
+    {
+      "epoch": 18.555231880680495,
+      "grad_norm": 0.3957250416278839,
+      "learning_rate": 0.00037756106091518506,
+      "loss": 3.2539,
+      "step": 63700
+    },
+    {
+      "epoch": 18.56979725005826,
+      "grad_norm": 0.36962229013442993,
+      "learning_rate": 0.0003773861847857767,
+      "loss": 3.2495,
+      "step": 63750
+    },
+    {
+      "epoch": 18.58436261943603,
+      "grad_norm": 0.3821750581264496,
+      "learning_rate": 0.0003772113086563684,
+      "loss": 3.2373,
+      "step": 63800
+    },
+    {
+      "epoch": 18.598927988813795,
+      "grad_norm": 0.39684993028640747,
+      "learning_rate": 0.00037703643252696,
+      "loss": 3.2577,
+      "step": 63850
+    },
+    {
+      "epoch": 18.613493358191565,
+      "grad_norm": 0.41482582688331604,
+      "learning_rate": 0.0003768615563975517,
+      "loss": 3.252,
+      "step": 63900
+    },
+    {
+      "epoch": 18.628058727569332,
+      "grad_norm": 0.4105575680732727,
+      "learning_rate": 0.00037668668026814335,
+      "loss": 3.2621,
+      "step": 63950
+    },
+    {
+      "epoch": 18.6426240969471,
+      "grad_norm": 0.4237818419933319,
+      "learning_rate": 0.000376511804138735,
+      "loss": 3.2635,
+      "step": 64000
+    },
+    {
+      "epoch": 18.6426240969471,
+      "eval_accuracy": 0.3728340078162641,
+      "eval_loss": 3.540757894515991,
+      "eval_runtime": 180.8009,
+      "eval_samples_per_second": 92.04,
+      "eval_steps_per_second": 5.758,
+      "step": 64000
+    },
+    {
+      "epoch": 18.657189466324866,
+      "grad_norm": 0.39062225818634033,
+      "learning_rate": 0.00037633692800932673,
+      "loss": 3.2489,
+      "step": 64050
+    },
+    {
+      "epoch": 18.671754835702632,
+      "grad_norm": 0.36457720398902893,
+      "learning_rate": 0.00037616205187991837,
+      "loss": 3.2514,
+      "step": 64100
+    },
+    {
+      "epoch": 18.6863202050804,
+      "grad_norm": 0.3855850100517273,
+      "learning_rate": 0.00037598717575051006,
+      "loss": 3.2562,
+      "step": 64150
+    },
+    {
+      "epoch": 18.70088557445817,
+      "grad_norm": 0.3868751525878906,
+      "learning_rate": 0.0003758122996211017,
+      "loss": 3.2718,
+      "step": 64200
+    },
+    {
+      "epoch": 18.715450943835936,
+      "grad_norm": 0.3820604979991913,
+      "learning_rate": 0.00037563742349169333,
+      "loss": 3.2657,
+      "step": 64250
+    },
+    {
+      "epoch": 18.730016313213703,
+      "grad_norm": 0.4040519893169403,
+      "learning_rate": 0.000375462547362285,
+      "loss": 3.2603,
+      "step": 64300
+    },
+    {
+      "epoch": 18.74458168259147,
+      "grad_norm": 0.37921711802482605,
+      "learning_rate": 0.00037528767123287665,
+      "loss": 3.2581,
+      "step": 64350
+    },
+    {
+      "epoch": 18.759147051969236,
+      "grad_norm": 0.398656964302063,
+      "learning_rate": 0.00037511279510346834,
+      "loss": 3.2533,
+      "step": 64400
+    },
+    {
+      "epoch": 18.773712421347007,
+      "grad_norm": 0.4015568792819977,
+      "learning_rate": 0.00037493791897406,
+      "loss": 3.2628,
+      "step": 64450
+    },
+    {
+      "epoch": 18.788277790724774,
+      "grad_norm": 0.39195629954338074,
+      "learning_rate": 0.0003747630428446516,
+      "loss": 3.2504,
+      "step": 64500
+    },
+    {
+      "epoch": 18.80284316010254,
+      "grad_norm": 0.3898211121559143,
+      "learning_rate": 0.00037458816671524336,
+      "loss": 3.2767,
+      "step": 64550
+    },
+    {
+      "epoch": 18.817408529480307,
+      "grad_norm": 0.3921559751033783,
+      "learning_rate": 0.000374413290585835,
+      "loss": 3.2735,
+      "step": 64600
+    },
+    {
+      "epoch": 18.831973898858074,
+      "grad_norm": 0.3801136612892151,
+      "learning_rate": 0.0003742384144564267,
+      "loss": 3.2683,
+      "step": 64650
+    },
+    {
+      "epoch": 18.846539268235844,
+      "grad_norm": 0.3727022409439087,
+      "learning_rate": 0.0003740635383270183,
+      "loss": 3.2686,
+      "step": 64700
+    },
+    {
+      "epoch": 18.86110463761361,
+      "grad_norm": 0.3597698509693146,
+      "learning_rate": 0.00037388866219761,
+      "loss": 3.2666,
+      "step": 64750
+    },
+    {
+      "epoch": 18.875670006991378,
+      "grad_norm": 0.37198469042778015,
+      "learning_rate": 0.00037371378606820165,
+      "loss": 3.2842,
+      "step": 64800
+    },
+    {
+      "epoch": 18.890235376369144,
+      "grad_norm": 0.365289568901062,
+      "learning_rate": 0.0003735389099387933,
+      "loss": 3.263,
+      "step": 64850
+    },
+    {
+      "epoch": 18.90480074574691,
+      "grad_norm": 0.40621164441108704,
+      "learning_rate": 0.000373364033809385,
+      "loss": 3.2664,
+      "step": 64900
+    },
+    {
+      "epoch": 18.919366115124678,
+      "grad_norm": 0.37832197546958923,
+      "learning_rate": 0.0003731891576799766,
+      "loss": 3.2687,
+      "step": 64950
+    },
+    {
+      "epoch": 18.93393148450245,
+      "grad_norm": 0.4049147069454193,
+      "learning_rate": 0.00037301428155056836,
+      "loss": 3.2826,
+      "step": 65000
+    },
+    {
+      "epoch": 18.93393148450245,
+      "eval_accuracy": 0.3732434838104693,
+      "eval_loss": 3.531397819519043,
+      "eval_runtime": 181.0112,
+      "eval_samples_per_second": 91.934,
+      "eval_steps_per_second": 5.751,
+      "step": 65000
+    },
+    {
+      "epoch": 18.948496853880215,
+      "grad_norm": 0.38759538531303406,
+      "learning_rate": 0.00037283940542116,
+      "loss": 3.2799,
+      "step": 65050
+    },
+    {
+      "epoch": 18.96306222325798,
+      "grad_norm": 0.3717094659805298,
+      "learning_rate": 0.00037266452929175163,
+      "loss": 3.2782,
+      "step": 65100
+    },
+    {
+      "epoch": 18.97762759263575,
+      "grad_norm": 0.34802091121673584,
+      "learning_rate": 0.0003724896531623433,
+      "loss": 3.2772,
+      "step": 65150
+    },
+    {
+      "epoch": 18.992192962013515,
+      "grad_norm": 0.3782845735549927,
+      "learning_rate": 0.00037231477703293496,
+      "loss": 3.2588,
+      "step": 65200
+    },
+    {
+      "epoch": 19.006700069913773,
+      "grad_norm": 0.40761205554008484,
+      "learning_rate": 0.00037213990090352665,
+      "loss": 3.2356,
+      "step": 65250
+    },
+    {
+      "epoch": 19.02126543929154,
+      "grad_norm": 0.4476456940174103,
+      "learning_rate": 0.0003719650247741183,
+      "loss": 3.1615,
+      "step": 65300
+    },
+    {
+      "epoch": 19.035830808669306,
+      "grad_norm": 0.3935956358909607,
+      "learning_rate": 0.00037179014864471,
+      "loss": 3.1767,
+      "step": 65350
+    },
+    {
+      "epoch": 19.050396178047077,
+      "grad_norm": 0.3989955484867096,
+      "learning_rate": 0.0003716152725153016,
+      "loss": 3.1669,
+      "step": 65400
+    },
+    {
+      "epoch": 19.064961547424844,
+      "grad_norm": 0.39710503816604614,
+      "learning_rate": 0.00037144039638589325,
+      "loss": 3.1738,
+      "step": 65450
+    },
+    {
+      "epoch": 19.07952691680261,
+      "grad_norm": 0.4290755093097687,
+      "learning_rate": 0.000371265520256485,
+      "loss": 3.1737,
+      "step": 65500
+    },
+    {
+      "epoch": 19.094092286180377,
+      "grad_norm": 0.36990752816200256,
+      "learning_rate": 0.00037109064412707663,
+      "loss": 3.1868,
+      "step": 65550
+    },
+    {
+      "epoch": 19.108657655558144,
+      "grad_norm": 0.3792154788970947,
+      "learning_rate": 0.0003709157679976683,
+      "loss": 3.1943,
+      "step": 65600
+    },
+    {
+      "epoch": 19.123223024935914,
+      "grad_norm": 0.3826157748699188,
+      "learning_rate": 0.00037074089186825995,
+      "loss": 3.1908,
+      "step": 65650
+    },
+    {
+      "epoch": 19.13778839431368,
+      "grad_norm": 0.399487167596817,
+      "learning_rate": 0.0003705660157388516,
+      "loss": 3.1772,
+      "step": 65700
+    },
+    {
+      "epoch": 19.152353763691448,
+      "grad_norm": 0.3746923506259918,
+      "learning_rate": 0.0003703911396094433,
+      "loss": 3.2043,
+      "step": 65750
+    },
+    {
+      "epoch": 19.166919133069214,
+      "grad_norm": 0.39206549525260925,
+      "learning_rate": 0.0003702162634800349,
+      "loss": 3.1959,
+      "step": 65800
+    },
+    {
+      "epoch": 19.18148450244698,
+      "grad_norm": 0.4111844003200531,
+      "learning_rate": 0.0003700413873506266,
+      "loss": 3.2045,
+      "step": 65850
+    },
+    {
+      "epoch": 19.196049871824748,
+      "grad_norm": 0.36120426654815674,
+      "learning_rate": 0.00036986651122121824,
+      "loss": 3.1998,
+      "step": 65900
+    },
+    {
+      "epoch": 19.210615241202518,
+      "grad_norm": 0.4161994457244873,
+      "learning_rate": 0.00036969163509181,
+      "loss": 3.2078,
+      "step": 65950
+    },
+    {
+      "epoch": 19.225180610580285,
+      "grad_norm": 0.37533485889434814,
+      "learning_rate": 0.0003695167589624016,
+      "loss": 3.2112,
+      "step": 66000
+    },
+    {
+      "epoch": 19.225180610580285,
+      "eval_accuracy": 0.3726248010978002,
+      "eval_loss": 3.549912452697754,
+      "eval_runtime": 180.9848,
+      "eval_samples_per_second": 91.947,
+      "eval_steps_per_second": 5.752,
+      "step": 66000
+    },
+    {
+      "epoch": 19.23974597995805,
+      "grad_norm": 0.38279658555984497,
+      "learning_rate": 0.00036934188283299326,
+      "loss": 3.21,
+      "step": 66050
+    },
+    {
+      "epoch": 19.25431134933582,
+      "grad_norm": 0.4204421043395996,
+      "learning_rate": 0.00036916700670358495,
+      "loss": 3.2106,
+      "step": 66100
+    },
+    {
+      "epoch": 19.268876718713585,
+      "grad_norm": 0.3843953311443329,
+      "learning_rate": 0.0003689921305741766,
+      "loss": 3.2139,
+      "step": 66150
+    },
+    {
+      "epoch": 19.283442088091356,
+      "grad_norm": 0.4173610508441925,
+      "learning_rate": 0.0003688172544447683,
+      "loss": 3.2154,
+      "step": 66200
+    },
+    {
+      "epoch": 19.298007457469122,
+      "grad_norm": 0.4082426428794861,
+      "learning_rate": 0.0003686423783153599,
+      "loss": 3.2327,
+      "step": 66250
+    },
+    {
+      "epoch": 19.31257282684689,
+      "grad_norm": 0.42148110270500183,
+      "learning_rate": 0.00036846750218595155,
+      "loss": 3.2225,
+      "step": 66300
+    },
+    {
+      "epoch": 19.327138196224656,
+      "grad_norm": 0.3666995167732239,
+      "learning_rate": 0.00036829262605654324,
+      "loss": 3.2239,
+      "step": 66350
+    },
+    {
+      "epoch": 19.341703565602423,
+      "grad_norm": 0.39264219999313354,
+      "learning_rate": 0.0003681177499271349,
+      "loss": 3.2201,
+      "step": 66400
+    },
+    {
+      "epoch": 19.356268934980193,
+      "grad_norm": 0.3931173086166382,
+      "learning_rate": 0.0003679428737977266,
+      "loss": 3.2255,
+      "step": 66450
+    },
+    {
+      "epoch": 19.37083430435796,
+      "grad_norm": 0.4180058538913727,
+      "learning_rate": 0.00036776799766831826,
+      "loss": 3.2224,
+      "step": 66500
+    },
+    {
+      "epoch": 19.385399673735726,
+      "grad_norm": 0.400886595249176,
+      "learning_rate": 0.0003675931215389099,
+      "loss": 3.2338,
+      "step": 66550
+    },
+    {
+      "epoch": 19.399965043113493,
+      "grad_norm": 0.3798990249633789,
+      "learning_rate": 0.0003674182454095016,
+      "loss": 3.2232,
+      "step": 66600
+    },
+    {
+      "epoch": 19.41453041249126,
+      "grad_norm": 0.3792427182197571,
+      "learning_rate": 0.0003672433692800932,
+      "loss": 3.2285,
+      "step": 66650
+    },
+    {
+      "epoch": 19.429095781869027,
+      "grad_norm": 0.3846825957298279,
+      "learning_rate": 0.0003670684931506849,
+      "loss": 3.2213,
+      "step": 66700
+    },
+    {
+      "epoch": 19.443661151246797,
+      "grad_norm": 0.39196518063545227,
+      "learning_rate": 0.00036689361702127655,
+      "loss": 3.2473,
+      "step": 66750
+    },
+    {
+      "epoch": 19.458226520624564,
+      "grad_norm": 0.4084039628505707,
+      "learning_rate": 0.00036671874089186824,
+      "loss": 3.2204,
+      "step": 66800
+    },
+    {
+      "epoch": 19.47279189000233,
+      "grad_norm": 0.3979856073856354,
+      "learning_rate": 0.00036654386476245987,
+      "loss": 3.2488,
+      "step": 66850
+    },
+    {
+      "epoch": 19.487357259380097,
+      "grad_norm": 0.415708065032959,
+      "learning_rate": 0.0003663689886330515,
+      "loss": 3.2268,
+      "step": 66900
+    },
+    {
+      "epoch": 19.501922628757864,
+      "grad_norm": 0.39281144738197327,
+      "learning_rate": 0.00036619411250364325,
+      "loss": 3.2434,
+      "step": 66950
+    },
+    {
+      "epoch": 19.516487998135634,
+      "grad_norm": 0.3792383074760437,
+      "learning_rate": 0.0003660192363742349,
+      "loss": 3.2371,
+      "step": 67000
+    },
+    {
+      "epoch": 19.516487998135634,
+      "eval_accuracy": 0.3729391403661835,
+      "eval_loss": 3.542454957962036,
+      "eval_runtime": 181.2242,
+      "eval_samples_per_second": 91.825,
+      "eval_steps_per_second": 5.744,
+      "step": 67000
+    },
+    {
+      "epoch": 19.5310533675134,
+      "grad_norm": 0.4281834363937378,
+      "learning_rate": 0.0003658443602448266,
+      "loss": 3.2322,
+      "step": 67050
+    },
+    {
+      "epoch": 19.545618736891168,
+      "grad_norm": 0.38975411653518677,
+      "learning_rate": 0.0003656694841154182,
+      "loss": 3.2313,
+      "step": 67100
+    },
+    {
+      "epoch": 19.560184106268935,
+      "grad_norm": 0.38386330008506775,
+      "learning_rate": 0.00036549460798600985,
+      "loss": 3.2512,
+      "step": 67150
+    },
+    {
+      "epoch": 19.5747494756467,
+      "grad_norm": 0.4082994759082794,
+      "learning_rate": 0.00036531973185660154,
+      "loss": 3.2442,
+      "step": 67200
+    },
+    {
+      "epoch": 19.589314845024468,
+      "grad_norm": 0.4414058029651642,
+      "learning_rate": 0.0003651448557271932,
+      "loss": 3.2379,
+      "step": 67250
+    },
+    {
+      "epoch": 19.60388021440224,
+      "grad_norm": 0.3642562925815582,
+      "learning_rate": 0.00036496997959778487,
+      "loss": 3.2414,
+      "step": 67300
+    },
+    {
+      "epoch": 19.618445583780005,
+      "grad_norm": 0.3908953070640564,
+      "learning_rate": 0.0003647951034683765,
+      "loss": 3.2359,
+      "step": 67350
+    },
+    {
+      "epoch": 19.633010953157772,
+      "grad_norm": 0.38066744804382324,
+      "learning_rate": 0.00036462022733896825,
+      "loss": 3.2482,
+      "step": 67400
+    },
+    {
+      "epoch": 19.64757632253554,
+      "grad_norm": 0.4061073064804077,
+      "learning_rate": 0.0003644453512095599,
+      "loss": 3.2442,
+      "step": 67450
+    },
+    {
+      "epoch": 19.662141691913305,
+      "grad_norm": 0.381572961807251,
+      "learning_rate": 0.0003642704750801515,
+      "loss": 3.2389,
+      "step": 67500
+    },
+    {
+      "epoch": 19.676707061291076,
+      "grad_norm": 0.37178897857666016,
+      "learning_rate": 0.0003640955989507432,
+      "loss": 3.2413,
+      "step": 67550
+    },
+    {
+      "epoch": 19.691272430668842,
+      "grad_norm": 0.38903379440307617,
+      "learning_rate": 0.00036392072282133485,
+      "loss": 3.2522,
+      "step": 67600
+    },
+    {
+      "epoch": 19.70583780004661,
+      "grad_norm": 0.3859827518463135,
+      "learning_rate": 0.00036374584669192654,
+      "loss": 3.2528,
+      "step": 67650
+    },
+    {
+      "epoch": 19.720403169424376,
+      "grad_norm": 0.3851865530014038,
+      "learning_rate": 0.0003635709705625182,
+      "loss": 3.2517,
+      "step": 67700
+    },
+    {
+      "epoch": 19.734968538802143,
+      "grad_norm": 0.40094801783561707,
+      "learning_rate": 0.0003633960944331098,
+      "loss": 3.2451,
+      "step": 67750
+    },
+    {
+      "epoch": 19.749533908179913,
+      "grad_norm": 0.3775973916053772,
+      "learning_rate": 0.0003632212183037015,
+      "loss": 3.2532,
+      "step": 67800
+    },
+    {
+      "epoch": 19.76409927755768,
+      "grad_norm": 0.3796077072620392,
+      "learning_rate": 0.00036304634217429314,
+      "loss": 3.2554,
+      "step": 67850
+    },
+    {
+      "epoch": 19.778664646935447,
+      "grad_norm": 0.3908114433288574,
+      "learning_rate": 0.0003628714660448849,
+      "loss": 3.2595,
+      "step": 67900
+    },
+    {
+      "epoch": 19.793230016313213,
+      "grad_norm": 0.3649216294288635,
+      "learning_rate": 0.0003626965899154765,
+      "loss": 3.2491,
+      "step": 67950
+    },
+    {
+      "epoch": 19.80779538569098,
+      "grad_norm": 0.35740792751312256,
+      "learning_rate": 0.0003625217137860682,
+      "loss": 3.2569,
+      "step": 68000
+    },
+    {
+      "epoch": 19.80779538569098,
+      "eval_accuracy": 0.373764207446983,
+      "eval_loss": 3.5353307723999023,
+      "eval_runtime": 181.1426,
+      "eval_samples_per_second": 91.867,
+      "eval_steps_per_second": 5.747,
+      "step": 68000
+    },
+    {
+      "epoch": 19.822360755068747,
+      "grad_norm": 0.4066384732723236,
+      "learning_rate": 0.00036234683765665985,
+      "loss": 3.2602,
+      "step": 68050
+    },
+    {
+      "epoch": 19.836926124446517,
+      "grad_norm": 0.3879859745502472,
+      "learning_rate": 0.0003621719615272515,
+      "loss": 3.2572,
+      "step": 68100
+    },
+    {
+      "epoch": 19.851491493824284,
+      "grad_norm": 0.4153246581554413,
+      "learning_rate": 0.00036199708539784317,
+      "loss": 3.2463,
+      "step": 68150
+    },
+    {
+      "epoch": 19.86605686320205,
+      "grad_norm": 0.3946223855018616,
+      "learning_rate": 0.0003618222092684348,
+      "loss": 3.2382,
+      "step": 68200
+    },
+    {
+      "epoch": 19.880622232579817,
+      "grad_norm": 0.4290454387664795,
+      "learning_rate": 0.0003616473331390265,
+      "loss": 3.2647,
+      "step": 68250
+    },
+    {
+      "epoch": 19.895187601957584,
+      "grad_norm": 0.3885732591152191,
+      "learning_rate": 0.00036147245700961813,
+      "loss": 3.2541,
+      "step": 68300
+    },
+    {
+      "epoch": 19.909752971335354,
+      "grad_norm": 0.3529806435108185,
+      "learning_rate": 0.00036129758088020977,
+      "loss": 3.2569,
+      "step": 68350
+    },
+    {
+      "epoch": 19.92431834071312,
+      "grad_norm": 0.38532155752182007,
+      "learning_rate": 0.0003611227047508015,
+      "loss": 3.2563,
+      "step": 68400
+    },
+    {
+      "epoch": 19.938883710090888,
+      "grad_norm": 0.3666004538536072,
+      "learning_rate": 0.00036094782862139315,
+      "loss": 3.2643,
+      "step": 68450
+    },
+    {
+      "epoch": 19.953449079468655,
+      "grad_norm": 0.3638163208961487,
+      "learning_rate": 0.00036077295249198484,
+      "loss": 3.2688,
+      "step": 68500
+    },
+    {
+      "epoch": 19.96801444884642,
+      "grad_norm": 0.4050072729587555,
+      "learning_rate": 0.0003605980763625765,
+      "loss": 3.2769,
+      "step": 68550
+    },
+    {
+      "epoch": 19.982579818224192,
+      "grad_norm": 0.43740108609199524,
+      "learning_rate": 0.0003604232002331681,
+      "loss": 3.2813,
+      "step": 68600
+    },
+    {
+      "epoch": 19.99714518760196,
+      "grad_norm": 0.37149035930633545,
+      "learning_rate": 0.0003602483241037598,
+      "loss": 3.2609,
+      "step": 68650
+    },
+    {
+      "epoch": 20.011652295502213,
+      "grad_norm": 0.4246836304664612,
+      "learning_rate": 0.00036007344797435144,
+      "loss": 3.1749,
+      "step": 68700
+    },
+    {
+      "epoch": 20.026217664879983,
+      "grad_norm": 0.4009047746658325,
+      "learning_rate": 0.00035989857184494313,
+      "loss": 3.1502,
+      "step": 68750
+    },
+    {
+      "epoch": 20.04078303425775,
+      "grad_norm": 0.3864176273345947,
+      "learning_rate": 0.00035972369571553477,
+      "loss": 3.1587,
+      "step": 68800
+    },
+    {
+      "epoch": 20.055348403635517,
+      "grad_norm": 0.43269333243370056,
+      "learning_rate": 0.0003595488195861265,
+      "loss": 3.1566,
+      "step": 68850
+    },
+    {
+      "epoch": 20.069913773013283,
+      "grad_norm": 0.3805122375488281,
+      "learning_rate": 0.00035937394345671815,
+      "loss": 3.1739,
+      "step": 68900
+    },
+    {
+      "epoch": 20.08447914239105,
+      "grad_norm": 0.39742806553840637,
+      "learning_rate": 0.0003591990673273098,
+      "loss": 3.1673,
+      "step": 68950
+    },
+    {
+      "epoch": 20.099044511768817,
+      "grad_norm": 0.3912816643714905,
+      "learning_rate": 0.0003590241911979015,
+      "loss": 3.1833,
+      "step": 69000
+    },
+    {
+      "epoch": 20.099044511768817,
+      "eval_accuracy": 0.3728480019700005,
+      "eval_loss": 3.5502328872680664,
+      "eval_runtime": 181.0296,
+      "eval_samples_per_second": 91.924,
+      "eval_steps_per_second": 5.75,
+      "step": 69000
+    },
+    {
+      "epoch": 20.113609881146587,
+      "grad_norm": 0.3727506101131439,
+      "learning_rate": 0.0003588493150684931,
+      "loss": 3.185,
+      "step": 69050
+    },
+    {
+      "epoch": 20.128175250524354,
+      "grad_norm": 0.39985391497612,
+      "learning_rate": 0.0003586744389390848,
+      "loss": 3.1799,
+      "step": 69100
+    },
+    {
+      "epoch": 20.14274061990212,
+      "grad_norm": 0.3829278349876404,
+      "learning_rate": 0.00035849956280967644,
+      "loss": 3.1729,
+      "step": 69150
+    },
+    {
+      "epoch": 20.157305989279887,
+      "grad_norm": 0.40887129306793213,
+      "learning_rate": 0.0003583246866802681,
+      "loss": 3.1763,
+      "step": 69200
+    },
+    {
+      "epoch": 20.171871358657654,
+      "grad_norm": 0.38781973719596863,
+      "learning_rate": 0.00035814981055085976,
+      "loss": 3.1868,
+      "step": 69250
+    },
+    {
+      "epoch": 20.186436728035424,
+      "grad_norm": 0.3758205771446228,
+      "learning_rate": 0.0003579749344214514,
+      "loss": 3.1901,
+      "step": 69300
+    },
+    {
+      "epoch": 20.20100209741319,
+      "grad_norm": 0.43557459115982056,
+      "learning_rate": 0.00035780005829204315,
+      "loss": 3.197,
+      "step": 69350
+    },
+    {
+      "epoch": 20.215567466790958,
+      "grad_norm": 0.4066002666950226,
+      "learning_rate": 0.0003576251821626348,
+      "loss": 3.1904,
+      "step": 69400
+    },
+    {
+      "epoch": 20.230132836168725,
+      "grad_norm": 0.4399625062942505,
+      "learning_rate": 0.00035745030603322647,
+      "loss": 3.2151,
+      "step": 69450
+    },
+    {
+      "epoch": 20.24469820554649,
+      "grad_norm": 0.4101237952709198,
+      "learning_rate": 0.0003572754299038181,
+      "loss": 3.1968,
+      "step": 69500
+    },
+    {
+      "epoch": 20.25926357492426,
+      "grad_norm": 0.3889336884021759,
+      "learning_rate": 0.00035710055377440974,
+      "loss": 3.212,
+      "step": 69550
+    },
+    {
+      "epoch": 20.27382894430203,
+      "grad_norm": 0.364547997713089,
+      "learning_rate": 0.00035692567764500143,
+      "loss": 3.2036,
+      "step": 69600
+    },
+    {
+      "epoch": 20.288394313679795,
+      "grad_norm": 0.4031059741973877,
+      "learning_rate": 0.00035675080151559307,
+      "loss": 3.2007,
+      "step": 69650
+    },
+    {
+      "epoch": 20.302959683057562,
+      "grad_norm": 0.392978698015213,
+      "learning_rate": 0.00035657592538618476,
+      "loss": 3.2054,
+      "step": 69700
+    },
+    {
+      "epoch": 20.31752505243533,
+      "grad_norm": 0.43729880452156067,
+      "learning_rate": 0.0003564010492567764,
+      "loss": 3.2135,
+      "step": 69750
+    },
+    {
+      "epoch": 20.332090421813096,
+      "grad_norm": 0.4184466004371643,
+      "learning_rate": 0.00035622617312736803,
+      "loss": 3.2051,
+      "step": 69800
+    },
+    {
+      "epoch": 20.346655791190866,
+      "grad_norm": 0.4293440878391266,
+      "learning_rate": 0.0003560512969979598,
+      "loss": 3.208,
+      "step": 69850
+    },
+    {
+      "epoch": 20.361221160568633,
+      "grad_norm": 0.39345675706863403,
+      "learning_rate": 0.0003558764208685514,
+      "loss": 3.2162,
+      "step": 69900
+    },
+    {
+      "epoch": 20.3757865299464,
+      "grad_norm": 0.41667041182518005,
+      "learning_rate": 0.0003557015447391431,
+      "loss": 3.21,
+      "step": 69950
+    },
+    {
+      "epoch": 20.390351899324166,
+      "grad_norm": 0.3816865086555481,
+      "learning_rate": 0.00035552666860973474,
+      "loss": 3.2127,
+      "step": 70000
+    },
+    {
+      "epoch": 20.390351899324166,
+      "eval_accuracy": 0.3731536389915225,
+      "eval_loss": 3.544055461883545,
+      "eval_runtime": 181.1128,
+      "eval_samples_per_second": 91.882,
+      "eval_steps_per_second": 5.748,
+      "step": 70000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 171650,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 50,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 20,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 5
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.46311518486528e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}