diff --git "a/last_to_push_frequency_2128/checkpoint-40000/trainer_state.json" "b/last_to_push_frequency_2128/checkpoint-40000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/last_to_push_frequency_2128/checkpoint-40000/trainer_state.json"
@@ -0,0 +1,6003 @@
+{
+  "best_global_step": 40000,
+  "best_metric": 3.5571908950805664,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/last_to_push_frequency_2128/checkpoint-40000",
+  "epoch": 11.648261401362921,
+  "eval_steps": 1000,
+  "global_step": 40000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014561127613722406,
+      "grad_norm": 1.4858547449111938,
+      "learning_rate": 0.000294,
+      "loss": 8.4433,
+      "step": 50
+    },
+    {
+      "epoch": 0.029122255227444813,
+      "grad_norm": 0.7950296998023987,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7515,
+      "step": 100
+    },
+    {
+      "epoch": 0.04368338284116722,
+      "grad_norm": 0.46963977813720703,
+      "learning_rate": 0.0005998286713286713,
+      "loss": 6.3451,
+      "step": 150
+    },
+    {
+      "epoch": 0.058244510454889625,
+      "grad_norm": 0.5301603674888611,
+      "learning_rate": 0.0005996538461538461,
+      "loss": 6.1351,
+      "step": 200
+    },
+    {
+      "epoch": 0.07280563806861204,
+      "grad_norm": 0.39352092146873474,
+      "learning_rate": 0.0005994790209790209,
+      "loss": 5.9881,
+      "step": 250
+    },
+    {
+      "epoch": 0.08736676568233444,
+      "grad_norm": 0.40677252411842346,
+      "learning_rate": 0.0005993041958041958,
+      "loss": 5.8428,
+      "step": 300
+    },
+    {
+      "epoch": 0.10192789329605685,
+      "grad_norm": 0.4639081358909607,
+      "learning_rate": 0.0005991293706293705,
+      "loss": 5.724,
+      "step": 350
+    },
+    {
+      "epoch": 0.11648902090977925,
+      "grad_norm": 0.4001801609992981,
+      "learning_rate": 0.0005989545454545454,
+      "loss": 5.6175,
+      "step": 400
+    },
+    {
+      "epoch": 0.13105014852350166,
+      "grad_norm": 0.48260679841041565,
+      "learning_rate": 0.0005987797202797202,
+      "loss": 5.4975,
+      "step": 450
+    },
+    {
+      "epoch": 0.14561127613722408,
+      "grad_norm": 0.5032499432563782,
+      "learning_rate": 0.000598604895104895,
+      "loss": 5.3974,
+      "step": 500
+    },
+    {
+      "epoch": 0.16017240375094646,
+      "grad_norm": 0.45361870527267456,
+      "learning_rate": 0.0005984300699300698,
+      "loss": 5.3242,
+      "step": 550
+    },
+    {
+      "epoch": 0.17473353136466888,
+      "grad_norm": 0.42623206973075867,
+      "learning_rate": 0.0005982552447552447,
+      "loss": 5.2535,
+      "step": 600
+    },
+    {
+      "epoch": 0.1892946589783913,
+      "grad_norm": 0.4469488561153412,
+      "learning_rate": 0.0005980804195804195,
+      "loss": 5.1812,
+      "step": 650
+    },
+    {
+      "epoch": 0.2038557865921137,
+      "grad_norm": 0.4211214780807495,
+      "learning_rate": 0.0005979055944055943,
+      "loss": 5.1233,
+      "step": 700
+    },
+    {
+      "epoch": 0.2184169142058361,
+      "grad_norm": 0.4314533770084381,
+      "learning_rate": 0.0005977307692307691,
+      "loss": 5.0677,
+      "step": 750
+    },
+    {
+      "epoch": 0.2329780418195585,
+      "grad_norm": 0.5419679284095764,
+      "learning_rate": 0.000597555944055944,
+      "loss": 5.0121,
+      "step": 800
+    },
+    {
+      "epoch": 0.24753916943328091,
+      "grad_norm": 0.4582156836986542,
+      "learning_rate": 0.0005973811188811188,
+      "loss": 4.9482,
+      "step": 850
+    },
+    {
+      "epoch": 0.2621002970470033,
+      "grad_norm": 0.4464901387691498,
+      "learning_rate": 0.0005972062937062936,
+      "loss": 4.9097,
+      "step": 900
+    },
+    {
+      "epoch": 0.27666142466072574,
+      "grad_norm": 0.4611063301563263,
+      "learning_rate": 0.0005970314685314685,
+      "loss": 4.867,
+      "step": 950
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "grad_norm": 0.638515830039978,
+      "learning_rate": 0.0005968566433566433,
+      "loss": 4.8082,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "eval_accuracy": 0.25687335662170213,
+      "eval_loss": 4.738000869750977,
+      "eval_runtime": 179.5194,
+      "eval_samples_per_second": 92.725,
+      "eval_steps_per_second": 5.799,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30578367988817057,
+      "grad_norm": 0.43320566415786743,
+      "learning_rate": 0.0005966818181818181,
+      "loss": 4.7777,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3203448075018929,
+      "grad_norm": 0.5117963552474976,
+      "learning_rate": 0.0005965069930069929,
+      "loss": 4.7314,
+      "step": 1100
+    },
+    {
+      "epoch": 0.33490593511561534,
+      "grad_norm": 0.43931126594543457,
+      "learning_rate": 0.0005963321678321677,
+      "loss": 4.6915,
+      "step": 1150
+    },
+    {
+      "epoch": 0.34946706272933775,
+      "grad_norm": 0.47124797105789185,
+      "learning_rate": 0.0005961573426573425,
+      "loss": 4.6574,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36402819034306017,
+      "grad_norm": 0.416259229183197,
+      "learning_rate": 0.0005959825174825174,
+      "loss": 4.6264,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3785893179567826,
+      "grad_norm": 0.46466392278671265,
+      "learning_rate": 0.0005958076923076922,
+      "loss": 4.5968,
+      "step": 1300
+    },
+    {
+      "epoch": 0.393150445570505,
+      "grad_norm": 0.4708881378173828,
+      "learning_rate": 0.000595632867132867,
+      "loss": 4.57,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4077115731842274,
+      "grad_norm": 0.48490315675735474,
+      "learning_rate": 0.0005954580419580418,
+      "loss": 4.5425,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4222727007979498,
+      "grad_norm": 0.48324763774871826,
+      "learning_rate": 0.0005952832167832168,
+      "loss": 4.5181,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4368338284116722,
+      "grad_norm": 0.39952364563941956,
+      "learning_rate": 0.0005951083916083916,
+      "loss": 4.5073,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4513949560253946,
+      "grad_norm": 0.4076775908470154,
+      "learning_rate": 0.0005949335664335664,
+      "loss": 4.4717,
+      "step": 1550
+    },
+    {
+      "epoch": 0.465956083639117,
+      "grad_norm": 0.43624168634414673,
+      "learning_rate": 0.0005947587412587413,
+      "loss": 4.4605,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4805172112528394,
+      "grad_norm": 0.41203925013542175,
+      "learning_rate": 0.0005945839160839161,
+      "loss": 4.4414,
+      "step": 1650
+    },
+    {
+      "epoch": 0.49507833886656183,
+      "grad_norm": 0.4499853849411011,
+      "learning_rate": 0.0005944090909090909,
+      "loss": 4.4138,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5096394664802842,
+      "grad_norm": 0.4240889549255371,
+      "learning_rate": 0.0005942342657342657,
+      "loss": 4.4028,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5242005940940067,
+      "grad_norm": 0.4965490698814392,
+      "learning_rate": 0.0005940594405594406,
+      "loss": 4.3865,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5387617217077291,
+      "grad_norm": 0.4231831133365631,
+      "learning_rate": 0.0005938846153846153,
+      "loss": 4.3704,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5533228493214515,
+      "grad_norm": 0.43335089087486267,
+      "learning_rate": 0.0005937097902097902,
+      "loss": 4.37,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5678839769351739,
+      "grad_norm": 0.42372700572013855,
+      "learning_rate": 0.000593534965034965,
+      "loss": 4.3537,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "grad_norm": 0.3962233364582062,
+      "learning_rate": 0.0005933601398601398,
+      "loss": 4.3388,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "eval_accuracy": 0.3000544549997378,
+      "eval_loss": 4.2761712074279785,
+      "eval_runtime": 179.7222,
+      "eval_samples_per_second": 92.621,
+      "eval_steps_per_second": 5.792,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5970062321626187,
+      "grad_norm": 0.3804718852043152,
+      "learning_rate": 0.0005931853146853146,
+      "loss": 4.3206,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6115673597763411,
+      "grad_norm": 0.36061641573905945,
+      "learning_rate": 0.0005930104895104895,
+      "loss": 4.3124,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6261284873900634,
+      "grad_norm": 0.44453078508377075,
+      "learning_rate": 0.0005928356643356643,
+      "loss": 4.2984,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6406896150037859,
+      "grad_norm": 0.381816029548645,
+      "learning_rate": 0.0005926608391608391,
+      "loss": 4.2922,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6552507426175083,
+      "grad_norm": 0.37145373225212097,
+      "learning_rate": 0.000592486013986014,
+      "loss": 4.2668,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6698118702312307,
+      "grad_norm": 0.37593939900398254,
+      "learning_rate": 0.0005923111888111888,
+      "loss": 4.2766,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6843729978449531,
+      "grad_norm": 0.40763673186302185,
+      "learning_rate": 0.0005921363636363636,
+      "loss": 4.2411,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6989341254586755,
+      "grad_norm": 0.37425413727760315,
+      "learning_rate": 0.0005919615384615384,
+      "loss": 4.2362,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7134952530723979,
+      "grad_norm": 0.46256929636001587,
+      "learning_rate": 0.0005917867132867133,
+      "loss": 4.2464,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7280563806861203,
+      "grad_norm": 0.34601518511772156,
+      "learning_rate": 0.0005916118881118881,
+      "loss": 4.2394,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7426175082998427,
+      "grad_norm": 0.4125564694404602,
+      "learning_rate": 0.0005914370629370629,
+      "loss": 4.2238,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7571786359135652,
+      "grad_norm": 0.3776606321334839,
+      "learning_rate": 0.0005912622377622377,
+      "loss": 4.2097,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7717397635272876,
+      "grad_norm": 0.3779846131801605,
+      "learning_rate": 0.0005910874125874125,
+      "loss": 4.1987,
+      "step": 2650
+    },
+    {
+      "epoch": 0.78630089114101,
+      "grad_norm": 0.3909547030925751,
+      "learning_rate": 0.0005909125874125873,
+      "loss": 4.19,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8008620187547324,
+      "grad_norm": 0.3437758982181549,
+      "learning_rate": 0.0005907377622377622,
+      "loss": 4.1919,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8154231463684548,
+      "grad_norm": 0.34142592549324036,
+      "learning_rate": 0.000590562937062937,
+      "loss": 4.175,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8299842739821772,
+      "grad_norm": 0.362221360206604,
+      "learning_rate": 0.0005903881118881118,
+      "loss": 4.1635,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8445454015958996,
+      "grad_norm": 0.37655583024024963,
+      "learning_rate": 0.0005902132867132867,
+      "loss": 4.1565,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8591065292096219,
+      "grad_norm": 0.3475290536880493,
+      "learning_rate": 0.0005900384615384615,
+      "loss": 4.1429,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "grad_norm": 0.3560430705547333,
+      "learning_rate": 0.0005898636363636363,
+      "loss": 4.1487,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "eval_accuracy": 0.31562550478444545,
+      "eval_loss": 4.094728946685791,
+      "eval_runtime": 179.6318,
+      "eval_samples_per_second": 92.667,
+      "eval_steps_per_second": 5.795,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8882287844370668,
+      "grad_norm": 0.3338482081890106,
+      "learning_rate": 0.0005896888111888111,
+      "loss": 4.1433,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9027899120507892,
+      "grad_norm": 0.3649440407752991,
+      "learning_rate": 0.000589513986013986,
+      "loss": 4.1306,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9173510396645116,
+      "grad_norm": 0.361857146024704,
+      "learning_rate": 0.0005893391608391608,
+      "loss": 4.1207,
+      "step": 3150
+    },
+    {
+      "epoch": 0.931912167278234,
+      "grad_norm": 0.34746089577674866,
+      "learning_rate": 0.0005891643356643356,
+      "loss": 4.1244,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9464732948919564,
+      "grad_norm": 0.344913512468338,
+      "learning_rate": 0.0005889895104895104,
+      "loss": 4.1127,
+      "step": 3250
+    },
+    {
+      "epoch": 0.9610344225056788,
+      "grad_norm": 0.3733556568622589,
+      "learning_rate": 0.0005888146853146853,
+      "loss": 4.1038,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9755955501194012,
+      "grad_norm": 0.38611236214637756,
+      "learning_rate": 0.00058863986013986,
+      "loss": 4.0889,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9901566777331237,
+      "grad_norm": 0.33691319823265076,
+      "learning_rate": 0.0005884650349650349,
+      "loss": 4.0955,
+      "step": 3400
+    },
+    {
+      "epoch": 1.004659560836391,
+      "grad_norm": 0.3706021010875702,
+      "learning_rate": 0.0005882902097902097,
+      "loss": 4.0726,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0192206884501136,
+      "grad_norm": 0.3402159810066223,
+      "learning_rate": 0.0005881153846153845,
+      "loss": 4.0144,
+      "step": 3500
+    },
+    {
+      "epoch": 1.033781816063836,
+      "grad_norm": 0.3579849898815155,
+      "learning_rate": 0.0005879405594405594,
+      "loss": 4.0103,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0483429436775584,
+      "grad_norm": 0.33701223134994507,
+      "learning_rate": 0.0005877657342657342,
+      "loss": 4.0348,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0629040712912807,
+      "grad_norm": 0.341237336397171,
+      "learning_rate": 0.000587590909090909,
+      "loss": 4.0054,
+      "step": 3650
+    },
+    {
+      "epoch": 1.0774651989050033,
+      "grad_norm": 0.3806729018688202,
+      "learning_rate": 0.0005874160839160838,
+      "loss": 4.0021,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0920263265187256,
+      "grad_norm": 0.33116230368614197,
+      "learning_rate": 0.0005872412587412587,
+      "loss": 4.0159,
+      "step": 3750
+    },
+    {
+      "epoch": 1.106587454132448,
+      "grad_norm": 0.35582491755485535,
+      "learning_rate": 0.0005870664335664335,
+      "loss": 4.0044,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1211485817461704,
+      "grad_norm": 0.32886144518852234,
+      "learning_rate": 0.0005868916083916083,
+      "loss": 4.0174,
+      "step": 3850
+    },
+    {
+      "epoch": 1.135709709359893,
+      "grad_norm": 0.3607575297355652,
+      "learning_rate": 0.0005867167832167831,
+      "loss": 3.9909,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1502708369736152,
+      "grad_norm": 0.3437859117984772,
+      "learning_rate": 0.000586541958041958,
+      "loss": 3.9831,
+      "step": 3950
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "grad_norm": 0.3273104727268219,
+      "learning_rate": 0.0005863671328671328,
+      "loss": 3.9789,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "eval_accuracy": 0.324832420381312,
+      "eval_loss": 3.989461660385132,
+      "eval_runtime": 179.6199,
+      "eval_samples_per_second": 92.673,
+      "eval_steps_per_second": 5.796,
+      "step": 4000
+    },
+    {
+      "epoch": 1.17939309220106,
+      "grad_norm": 0.3365177810192108,
+      "learning_rate": 0.0005861923076923076,
+      "loss": 3.9821,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1939542198147826,
+      "grad_norm": 0.34082263708114624,
+      "learning_rate": 0.0005860174825174824,
+      "loss": 3.9917,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2085153474285049,
+      "grad_norm": 0.34475305676460266,
+      "learning_rate": 0.0005858426573426573,
+      "loss": 3.9865,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2230764750422272,
+      "grad_norm": 0.31244170665740967,
+      "learning_rate": 0.000585667832167832,
+      "loss": 3.9812,
+      "step": 4200
+    },
+    {
+      "epoch": 1.2376376026559497,
+      "grad_norm": 0.36344829201698303,
+      "learning_rate": 0.000585493006993007,
+      "loss": 3.97,
+      "step": 4250
+    },
+    {
+      "epoch": 1.2521987302696722,
+      "grad_norm": 0.3585837781429291,
+      "learning_rate": 0.0005853181818181817,
+      "loss": 3.9773,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2667598578833945,
+      "grad_norm": 0.37048792839050293,
+      "learning_rate": 0.0005851433566433565,
+      "loss": 3.9808,
+      "step": 4350
+    },
+    {
+      "epoch": 1.2813209854971168,
+      "grad_norm": 0.32924020290374756,
+      "learning_rate": 0.0005849685314685315,
+      "loss": 3.9735,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2958821131108393,
+      "grad_norm": 0.5044884085655212,
+      "learning_rate": 0.0005847937062937063,
+      "loss": 3.9652,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3104432407245616,
+      "grad_norm": 0.34231820702552795,
+      "learning_rate": 0.0005846188811188811,
+      "loss": 3.9638,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3250043683382842,
+      "grad_norm": 0.34204453229904175,
+      "learning_rate": 0.0005844440559440559,
+      "loss": 3.9566,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3395654959520065,
+      "grad_norm": 0.3231236934661865,
+      "learning_rate": 0.0005842692307692308,
+      "loss": 3.9454,
+      "step": 4600
+    },
+    {
+      "epoch": 1.354126623565729,
+      "grad_norm": 0.3487296402454376,
+      "learning_rate": 0.0005840944055944056,
+      "loss": 3.9472,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3686877511794513,
+      "grad_norm": 0.3496778905391693,
+      "learning_rate": 0.0005839195804195804,
+      "loss": 3.9488,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3832488787931738,
+      "grad_norm": 0.3419700860977173,
+      "learning_rate": 0.0005837447552447552,
+      "loss": 3.953,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3978100064068961,
+      "grad_norm": 0.33043453097343445,
+      "learning_rate": 0.0005835699300699301,
+      "loss": 3.9475,
+      "step": 4800
+    },
+    {
+      "epoch": 1.4123711340206184,
+      "grad_norm": 0.316031277179718,
+      "learning_rate": 0.0005833951048951048,
+      "loss": 3.9505,
+      "step": 4850
+    },
+    {
+      "epoch": 1.426932261634341,
+      "grad_norm": 0.3300357162952423,
+      "learning_rate": 0.0005832202797202797,
+      "loss": 3.9385,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4414933892480635,
+      "grad_norm": 0.35844576358795166,
+      "learning_rate": 0.0005830454545454546,
+      "loss": 3.9406,
+      "step": 4950
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "grad_norm": 0.32424795627593994,
+      "learning_rate": 0.0005828706293706293,
+      "loss": 3.9446,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "eval_accuracy": 0.3322647284197963,
+      "eval_loss": 3.9135546684265137,
+      "eval_runtime": 183.5994,
+      "eval_samples_per_second": 90.665,
+      "eval_steps_per_second": 5.67,
+      "step": 5000
+    },
+    {
+      "epoch": 1.470615644475508,
+      "grad_norm": 0.33048802614212036,
+      "learning_rate": 0.0005826958041958042,
+      "loss": 3.9274,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4851767720892306,
+      "grad_norm": 0.33716776967048645,
+      "learning_rate": 0.000582520979020979,
+      "loss": 3.9214,
+      "step": 5100
+    },
+    {
+      "epoch": 1.4997378997029531,
+      "grad_norm": 0.31546398997306824,
+      "learning_rate": 0.0005823461538461538,
+      "loss": 3.9127,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5142990273166754,
+      "grad_norm": 0.3368748426437378,
+      "learning_rate": 0.0005821713286713286,
+      "loss": 3.9187,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5288601549303977,
+      "grad_norm": 0.32343748211860657,
+      "learning_rate": 0.0005819965034965035,
+      "loss": 3.9177,
+      "step": 5250
+    },
+    {
+      "epoch": 1.5434212825441203,
+      "grad_norm": 0.33815833926200867,
+      "learning_rate": 0.0005818216783216783,
+      "loss": 3.9081,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5579824101578428,
+      "grad_norm": 0.32644879817962646,
+      "learning_rate": 0.0005816468531468531,
+      "loss": 3.9079,
+      "step": 5350
+    },
+    {
+      "epoch": 1.572543537771565,
+      "grad_norm": 0.32149451971054077,
+      "learning_rate": 0.0005814720279720279,
+      "loss": 3.8998,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5871046653852874,
+      "grad_norm": 0.34723344445228577,
+      "learning_rate": 0.0005812972027972028,
+      "loss": 3.9069,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6016657929990097,
+      "grad_norm": 0.32005998492240906,
+      "learning_rate": 0.0005811223776223776,
+      "loss": 3.9124,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6162269206127322,
+      "grad_norm": 0.31391555070877075,
+      "learning_rate": 0.0005809475524475524,
+      "loss": 3.8972,
+      "step": 5550
+    },
+    {
+      "epoch": 1.6307880482264547,
+      "grad_norm": 0.3457600772380829,
+      "learning_rate": 0.0005807727272727272,
+      "loss": 3.8887,
+      "step": 5600
+    },
+    {
+      "epoch": 1.645349175840177,
+      "grad_norm": 0.34775686264038086,
+      "learning_rate": 0.0005805979020979021,
+      "loss": 3.9066,
+      "step": 5650
+    },
+    {
+      "epoch": 1.6599103034538993,
+      "grad_norm": 0.3384147584438324,
+      "learning_rate": 0.0005804230769230769,
+      "loss": 3.8885,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6744714310676219,
+      "grad_norm": 0.33265241980552673,
+      "learning_rate": 0.0005802482517482517,
+      "loss": 3.8862,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6890325586813444,
+      "grad_norm": 0.33996856212615967,
+      "learning_rate": 0.0005800734265734265,
+      "loss": 3.8993,
+      "step": 5800
+    },
+    {
+      "epoch": 1.7035936862950667,
+      "grad_norm": 0.3211154043674469,
+      "learning_rate": 0.0005798986013986013,
+      "loss": 3.8839,
+      "step": 5850
+    },
+    {
+      "epoch": 1.718154813908789,
+      "grad_norm": 0.3389524817466736,
+      "learning_rate": 0.0005797237762237762,
+      "loss": 3.8848,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7327159415225115,
+      "grad_norm": 0.31527891755104065,
+      "learning_rate": 0.000579548951048951,
+      "loss": 3.8743,
+      "step": 5950
+    },
+    {
+      "epoch": 1.747277069136234,
+      "grad_norm": 0.33459725975990295,
+      "learning_rate": 0.0005793741258741258,
+      "loss": 3.8779,
+      "step": 6000
+    },
+    {
+      "epoch": 1.747277069136234,
+      "eval_accuracy": 0.33703706490372914,
+      "eval_loss": 3.856008291244507,
+      "eval_runtime": 181.6053,
+      "eval_samples_per_second": 91.66,
+      "eval_steps_per_second": 5.732,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7618381967499563,
+      "grad_norm": 0.3163743317127228,
+      "learning_rate": 0.0005791993006993006,
+      "loss": 3.8718,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7763993243636786,
+      "grad_norm": 0.3169962167739868,
+      "learning_rate": 0.0005790244755244755,
+      "loss": 3.871,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7909604519774012,
+      "grad_norm": 0.3243774175643921,
+      "learning_rate": 0.0005788496503496503,
+      "loss": 3.8746,
+      "step": 6150
+    },
+    {
+      "epoch": 1.8055215795911237,
+      "grad_norm": 0.30815380811691284,
+      "learning_rate": 0.0005786748251748251,
+      "loss": 3.8719,
+      "step": 6200
+    },
+    {
+      "epoch": 1.820082707204846,
+      "grad_norm": 0.355295866727829,
+      "learning_rate": 0.0005784999999999999,
+      "loss": 3.8605,
+      "step": 6250
+    },
+    {
+      "epoch": 1.8346438348185683,
+      "grad_norm": 0.34578466415405273,
+      "learning_rate": 0.0005783251748251748,
+      "loss": 3.8651,
+      "step": 6300
+    },
+    {
+      "epoch": 1.8492049624322906,
+      "grad_norm": 0.3516864776611328,
+      "learning_rate": 0.0005781503496503496,
+      "loss": 3.8615,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8637660900460131,
+      "grad_norm": 0.3126903772354126,
+      "learning_rate": 0.0005779755244755244,
+      "loss": 3.8609,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8783272176597356,
+      "grad_norm": 0.31201526522636414,
+      "learning_rate": 0.0005778006993006993,
+      "loss": 3.8545,
+      "step": 6450
+    },
+    {
+      "epoch": 1.892888345273458,
+      "grad_norm": 0.33601483702659607,
+      "learning_rate": 0.000577625874125874,
+      "loss": 3.8421,
+      "step": 6500
+    },
+    {
+      "epoch": 1.9074494728871803,
+      "grad_norm": 0.3365996479988098,
+      "learning_rate": 0.0005774510489510489,
+      "loss": 3.851,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9220106005009028,
+      "grad_norm": 0.30747294425964355,
+      "learning_rate": 0.0005772762237762237,
+      "loss": 3.8576,
+      "step": 6600
+    },
+    {
+      "epoch": 1.9365717281146253,
+      "grad_norm": 0.3154364228248596,
+      "learning_rate": 0.0005771013986013985,
+      "loss": 3.8576,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9511328557283476,
+      "grad_norm": 0.32481271028518677,
+      "learning_rate": 0.0005769265734265733,
+      "loss": 3.8464,
+      "step": 6700
+    },
+    {
+      "epoch": 1.96569398334207,
+      "grad_norm": 0.3115486204624176,
+      "learning_rate": 0.0005767517482517482,
+      "loss": 3.8582,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9802551109557924,
+      "grad_norm": 0.33827632665634155,
+      "learning_rate": 0.000576576923076923,
+      "loss": 3.8423,
+      "step": 6800
+    },
+    {
+      "epoch": 1.994816238569515,
+      "grad_norm": 0.31877878308296204,
+      "learning_rate": 0.0005764020979020978,
+      "loss": 3.8472,
+      "step": 6850
+    },
+    {
+      "epoch": 2.009319121672782,
+      "grad_norm": 0.31373608112335205,
+      "learning_rate": 0.0005762272727272726,
+      "loss": 3.7795,
+      "step": 6900
+    },
+    {
+      "epoch": 2.023880249286505,
+      "grad_norm": 0.3324246108531952,
+      "learning_rate": 0.0005760524475524475,
+      "loss": 3.7456,
+      "step": 6950
+    },
+    {
+      "epoch": 2.038441376900227,
+      "grad_norm": 0.33372506499290466,
+      "learning_rate": 0.0005758776223776223,
+      "loss": 3.7467,
+      "step": 7000
+    },
+    {
+      "epoch": 2.038441376900227,
+      "eval_accuracy": 0.3411688027400552,
+      "eval_loss": 3.816201686859131,
+      "eval_runtime": 179.7567,
+      "eval_samples_per_second": 92.603,
+      "eval_steps_per_second": 5.791,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0530025045139495,
+      "grad_norm": 0.32919570803642273,
+      "learning_rate": 0.0005757027972027971,
+      "loss": 3.7547,
+      "step": 7050
+    },
+    {
+      "epoch": 2.067563632127672,
+      "grad_norm": 0.35541394352912903,
+      "learning_rate": 0.000575527972027972,
+      "loss": 3.7378,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0821247597413945,
+      "grad_norm": 0.3367560803890228,
+      "learning_rate": 0.0005753531468531468,
+      "loss": 3.7486,
+      "step": 7150
+    },
+    {
+      "epoch": 2.096685887355117,
+      "grad_norm": 0.317826509475708,
+      "learning_rate": 0.0005751783216783216,
+      "loss": 3.7494,
+      "step": 7200
+    },
+    {
+      "epoch": 2.111247014968839,
+      "grad_norm": 0.31629154086112976,
+      "learning_rate": 0.0005750034965034964,
+      "loss": 3.7552,
+      "step": 7250
+    },
+    {
+      "epoch": 2.1258081425825615,
+      "grad_norm": 0.33633357286453247,
+      "learning_rate": 0.0005748286713286712,
+      "loss": 3.7447,
+      "step": 7300
+    },
+    {
+      "epoch": 2.140369270196284,
+      "grad_norm": 0.3292492628097534,
+      "learning_rate": 0.000574653846153846,
+      "loss": 3.746,
+      "step": 7350
+    },
+    {
+      "epoch": 2.1549303978100065,
+      "grad_norm": 0.3265497088432312,
+      "learning_rate": 0.000574479020979021,
+      "loss": 3.7502,
+      "step": 7400
+    },
+    {
+      "epoch": 2.169491525423729,
+      "grad_norm": 0.3585764467716217,
+      "learning_rate": 0.0005743041958041958,
+      "loss": 3.765,
+      "step": 7450
+    },
+    {
+      "epoch": 2.184052653037451,
+      "grad_norm": 0.3560110926628113,
+      "learning_rate": 0.0005741293706293706,
+      "loss": 3.7482,
+      "step": 7500
+    },
+    {
+      "epoch": 2.198613780651174,
+      "grad_norm": 0.3085196018218994,
+      "learning_rate": 0.0005739545454545454,
+      "loss": 3.7441,
+      "step": 7550
+    },
+    {
+      "epoch": 2.213174908264896,
+      "grad_norm": 0.30168864130973816,
+      "learning_rate": 0.0005737797202797203,
+      "loss": 3.7617,
+      "step": 7600
+    },
+    {
+      "epoch": 2.2277360358786185,
+      "grad_norm": 0.33070042729377747,
+      "learning_rate": 0.0005736048951048951,
+      "loss": 3.7502,
+      "step": 7650
+    },
+    {
+      "epoch": 2.2422971634923408,
+      "grad_norm": 0.3281424045562744,
+      "learning_rate": 0.0005734300699300699,
+      "loss": 3.7585,
+      "step": 7700
+    },
+    {
+      "epoch": 2.256858291106063,
+      "grad_norm": 0.311362087726593,
+      "learning_rate": 0.0005732552447552448,
+      "loss": 3.7633,
+      "step": 7750
+    },
+    {
+      "epoch": 2.271419418719786,
+      "grad_norm": 0.34092429280281067,
+      "learning_rate": 0.0005730804195804196,
+      "loss": 3.758,
+      "step": 7800
+    },
+    {
+      "epoch": 2.285980546333508,
+      "grad_norm": 0.3130812346935272,
+      "learning_rate": 0.0005729055944055944,
+      "loss": 3.7595,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3005416739472304,
+      "grad_norm": 0.3428857922554016,
+      "learning_rate": 0.0005727307692307692,
+      "loss": 3.7469,
+      "step": 7900
+    },
+    {
+      "epoch": 2.3151028015609527,
+      "grad_norm": 0.3283872604370117,
+      "learning_rate": 0.0005725559440559441,
+      "loss": 3.7548,
+      "step": 7950
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "grad_norm": 0.3140428066253662,
+      "learning_rate": 0.0005723811188811188,
+      "loss": 3.7515,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "eval_accuracy": 0.34455049114130487,
+      "eval_loss": 3.783414125442505,
+      "eval_runtime": 180.5023,
+      "eval_samples_per_second": 92.22,
+      "eval_steps_per_second": 5.767,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3442250567883978,
+      "grad_norm": 0.32429584860801697,
+      "learning_rate": 0.0005722062937062937,
+      "loss": 3.7638,
+      "step": 8050
+    },
+    {
+      "epoch": 2.35878618440212,
+      "grad_norm": 0.3218806982040405,
+      "learning_rate": 0.0005720314685314685,
+      "loss": 3.7444,
+      "step": 8100
+    },
+    {
+      "epoch": 2.3733473120158424,
+      "grad_norm": 0.33977949619293213,
+      "learning_rate": 0.0005718566433566433,
+      "loss": 3.7523,
+      "step": 8150
+    },
+    {
+      "epoch": 2.387908439629565,
+      "grad_norm": 0.3360040783882141,
+      "learning_rate": 0.0005716818181818181,
+      "loss": 3.7403,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4024695672432874,
+      "grad_norm": 0.3237862288951874,
+      "learning_rate": 0.000571506993006993,
+      "loss": 3.7531,
+      "step": 8250
+    },
+    {
+      "epoch": 2.4170306948570097,
+      "grad_norm": 0.36249440908432007,
+      "learning_rate": 0.0005713321678321678,
+      "loss": 3.7537,
+      "step": 8300
+    },
+    {
+      "epoch": 2.431591822470732,
+      "grad_norm": 0.31824633479118347,
+      "learning_rate": 0.0005711573426573426,
+      "loss": 3.7521,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4461529500844543,
+      "grad_norm": 0.31667229533195496,
+      "learning_rate": 0.0005709825174825175,
+      "loss": 3.7455,
+      "step": 8400
+    },
+    {
+      "epoch": 2.460714077698177,
+      "grad_norm": 0.3123161196708679,
+      "learning_rate": 0.0005708076923076923,
+      "loss": 3.7387,
+      "step": 8450
+    },
+    {
+      "epoch": 2.4752752053118994,
+      "grad_norm": 0.3177723288536072,
+      "learning_rate": 0.0005706328671328671,
+      "loss": 3.7467,
+      "step": 8500
+    },
+    {
+      "epoch": 2.4898363329256217,
+      "grad_norm": 0.3312056064605713,
+      "learning_rate": 0.0005704580419580419,
+      "loss": 3.7488,
+      "step": 8550
+    },
+    {
+      "epoch": 2.5043974605393444,
+      "grad_norm": 0.30859455466270447,
+      "learning_rate": 0.0005702832167832168,
+      "loss": 3.7394,
+      "step": 8600
+    },
+    {
+      "epoch": 2.5189585881530667,
+      "grad_norm": 0.30974218249320984,
+      "learning_rate": 0.0005701083916083916,
+      "loss": 3.7423,
+      "step": 8650
+    },
+    {
+      "epoch": 2.533519715766789,
+      "grad_norm": 0.3171643614768982,
+      "learning_rate": 0.0005699335664335664,
+      "loss": 3.7398,
+      "step": 8700
+    },
+    {
+      "epoch": 2.5480808433805113,
+      "grad_norm": 0.3243333101272583,
+      "learning_rate": 0.0005697587412587412,
+      "loss": 3.7445,
+      "step": 8750
+    },
+    {
+      "epoch": 2.5626419709942336,
+      "grad_norm": 0.3245783746242523,
+      "learning_rate": 0.000569583916083916,
+      "loss": 3.7514,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5772030986079564,
+      "grad_norm": 0.31804442405700684,
+      "learning_rate": 0.0005694090909090908,
+      "loss": 3.7359,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5917642262216787,
+      "grad_norm": 0.3267365097999573,
+      "learning_rate": 0.0005692342657342657,
+      "loss": 3.7395,
+      "step": 8900
+    },
+    {
+      "epoch": 2.606325353835401,
+      "grad_norm": 0.29884088039398193,
+      "learning_rate": 0.0005690594405594405,
+      "loss": 3.7334,
+      "step": 8950
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "grad_norm": 0.3067515790462494,
+      "learning_rate": 0.0005688846153846153,
+      "loss": 3.7465,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "eval_accuracy": 0.3470682119409281,
+      "eval_loss": 3.756100654602051,
+      "eval_runtime": 180.4789,
+      "eval_samples_per_second": 92.232,
+      "eval_steps_per_second": 5.768,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6354476090628456,
+      "grad_norm": 0.33908265829086304,
+      "learning_rate": 0.0005687097902097901,
+      "loss": 3.7438,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6500087366765683,
+      "grad_norm": 0.3071008026599884,
+      "learning_rate": 0.000568534965034965,
+      "loss": 3.7439,
+      "step": 9100
+    },
+    {
+      "epoch": 2.6645698642902906,
+      "grad_norm": 0.3263190686702728,
+      "learning_rate": 0.0005683601398601398,
+      "loss": 3.7297,
+      "step": 9150
+    },
+    {
+      "epoch": 2.679130991904013,
+      "grad_norm": 0.3221532702445984,
+      "learning_rate": 0.0005681853146853146,
+      "loss": 3.715,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6936921195177357,
+      "grad_norm": 0.3336532711982727,
+      "learning_rate": 0.0005680104895104895,
+      "loss": 3.7365,
+      "step": 9250
+    },
+    {
+      "epoch": 2.708253247131458,
+      "grad_norm": 0.32632067799568176,
+      "learning_rate": 0.0005678356643356643,
+      "loss": 3.7242,
+      "step": 9300
+    },
+    {
+      "epoch": 2.7228143747451803,
+      "grad_norm": 0.3240657150745392,
+      "learning_rate": 0.0005676608391608391,
+      "loss": 3.7353,
+      "step": 9350
+    },
+    {
+      "epoch": 2.7373755023589026,
+      "grad_norm": 0.3053086996078491,
+      "learning_rate": 0.0005674860139860139,
+      "loss": 3.7285,
+      "step": 9400
+    },
+    {
+      "epoch": 2.751936629972625,
+      "grad_norm": 0.33227983117103577,
+      "learning_rate": 0.0005673111888111888,
+      "loss": 3.7225,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7664977575863476,
+      "grad_norm": 0.3048139810562134,
+      "learning_rate": 0.0005671363636363635,
+      "loss": 3.7369,
+      "step": 9500
+    },
+    {
+      "epoch": 2.78105888520007,
+      "grad_norm": 0.3088410496711731,
+      "learning_rate": 0.0005669615384615384,
+      "loss": 3.7381,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7956200128137922,
+      "grad_norm": 0.311073362827301,
+      "learning_rate": 0.0005667867132867132,
+      "loss": 3.7295,
+      "step": 9600
+    },
+    {
+      "epoch": 2.8101811404275145,
+      "grad_norm": 0.3153594136238098,
+      "learning_rate": 0.000566611888111888,
+      "loss": 3.7254,
+      "step": 9650
+    },
+    {
+      "epoch": 2.824742268041237,
+      "grad_norm": 0.31726452708244324,
+      "learning_rate": 0.0005664370629370628,
+      "loss": 3.7383,
+      "step": 9700
+    },
+    {
+      "epoch": 2.8393033956549596,
+      "grad_norm": 0.35246723890304565,
+      "learning_rate": 0.0005662622377622377,
+      "loss": 3.7415,
+      "step": 9750
+    },
+    {
+      "epoch": 2.853864523268682,
+      "grad_norm": 0.34164196252822876,
+      "learning_rate": 0.0005660874125874125,
+      "loss": 3.73,
+      "step": 9800
+    },
+    {
+      "epoch": 2.868425650882404,
+      "grad_norm": 0.34838706254959106,
+      "learning_rate": 0.0005659125874125873,
+      "loss": 3.7303,
+      "step": 9850
+    },
+    {
+      "epoch": 2.882986778496127,
+      "grad_norm": 0.338086873292923,
+      "learning_rate": 0.0005657377622377622,
+      "loss": 3.7274,
+      "step": 9900
+    },
+    {
+      "epoch": 2.8975479061098492,
+      "grad_norm": 0.3079865872859955,
+      "learning_rate": 0.000565562937062937,
+      "loss": 3.7255,
+      "step": 9950
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "grad_norm": 0.3039040267467499,
+      "learning_rate": 0.0005653881118881118,
+      "loss": 3.7206,
+      "step": 10000
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "eval_accuracy": 0.3496147355793591,
+      "eval_loss": 3.727715253829956,
+      "eval_runtime": 179.8629,
+      "eval_samples_per_second": 92.548,
+      "eval_steps_per_second": 5.788,
+      "step": 10000
+    },
+    {
+      "epoch": 2.926670161337294,
+      "grad_norm": 0.32474496960639954,
+      "learning_rate": 0.0005652132867132866,
+      "loss": 3.7148,
+      "step": 10050
+    },
+    {
+      "epoch": 2.941231288951016,
+      "grad_norm": 0.33152255415916443,
+      "learning_rate": 0.0005650384615384615,
+      "loss": 3.7307,
+      "step": 10100
+    },
+    {
+      "epoch": 2.955792416564739,
+      "grad_norm": 0.3443283438682556,
+      "learning_rate": 0.0005648636363636363,
+      "loss": 3.7031,
+      "step": 10150
+    },
+    {
+      "epoch": 2.970353544178461,
+      "grad_norm": 0.31178221106529236,
+      "learning_rate": 0.0005646888111888111,
+      "loss": 3.7225,
+      "step": 10200
+    },
+    {
+      "epoch": 2.9849146717921835,
+      "grad_norm": 0.2839028537273407,
+      "learning_rate": 0.000564513986013986,
+      "loss": 3.711,
+      "step": 10250
+    },
+    {
+      "epoch": 2.9994757994059063,
+      "grad_norm": 0.3071160912513733,
+      "learning_rate": 0.0005643391608391607,
+      "loss": 3.7179,
+      "step": 10300
+    },
+    {
+      "epoch": 3.0139786825091734,
+      "grad_norm": 0.2976529002189636,
+      "learning_rate": 0.0005641643356643355,
+      "loss": 3.6143,
+      "step": 10350
+    },
+    {
+      "epoch": 3.0285398101228957,
+      "grad_norm": 0.3535459339618683,
+      "learning_rate": 0.0005639895104895105,
+      "loss": 3.6151,
+      "step": 10400
+    },
+    {
+      "epoch": 3.0431009377366185,
+      "grad_norm": 0.3356330394744873,
+      "learning_rate": 0.0005638146853146853,
+      "loss": 3.6181,
+      "step": 10450
+    },
+    {
+      "epoch": 3.057662065350341,
+      "grad_norm": 0.3338814973831177,
+      "learning_rate": 0.0005636398601398601,
+      "loss": 3.6144,
+      "step": 10500
+    },
+    {
+      "epoch": 3.072223192964063,
+      "grad_norm": 0.30894315242767334,
+      "learning_rate": 0.000563465034965035,
+      "loss": 3.6221,
+      "step": 10550
+    },
+    {
+      "epoch": 3.0867843205777854,
+      "grad_norm": 0.3283049166202545,
+      "learning_rate": 0.0005632902097902098,
+      "loss": 3.6405,
+      "step": 10600
+    },
+    {
+      "epoch": 3.101345448191508,
+      "grad_norm": 0.31288209557533264,
+      "learning_rate": 0.0005631153846153846,
+      "loss": 3.6257,
+      "step": 10650
+    },
+    {
+      "epoch": 3.1159065758052304,
+      "grad_norm": 0.32523486018180847,
+      "learning_rate": 0.0005629405594405594,
+      "loss": 3.6256,
+      "step": 10700
+    },
+    {
+      "epoch": 3.1304677034189528,
+      "grad_norm": 0.32469362020492554,
+      "learning_rate": 0.0005627657342657343,
+      "loss": 3.6281,
+      "step": 10750
+    },
+    {
+      "epoch": 3.145028831032675,
+      "grad_norm": 0.3065453767776489,
+      "learning_rate": 0.0005625909090909091,
+      "loss": 3.6223,
+      "step": 10800
+    },
+    {
+      "epoch": 3.1595899586463974,
+      "grad_norm": 0.3205210864543915,
+      "learning_rate": 0.0005624160839160839,
+      "loss": 3.6287,
+      "step": 10850
+    },
+    {
+      "epoch": 3.17415108626012,
+      "grad_norm": 0.3646294176578522,
+      "learning_rate": 0.0005622412587412587,
+      "loss": 3.6388,
+      "step": 10900
+    },
+    {
+      "epoch": 3.1887122138738424,
+      "grad_norm": 0.31701645255088806,
+      "learning_rate": 0.0005620664335664336,
+      "loss": 3.6382,
+      "step": 10950
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "grad_norm": 0.3214119076728821,
+      "learning_rate": 0.0005618916083916083,
+      "loss": 3.6411,
+      "step": 11000
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "eval_accuracy": 0.35119712827467703,
+      "eval_loss": 3.7161929607391357,
+      "eval_runtime": 179.6418,
+      "eval_samples_per_second": 92.662,
+      "eval_steps_per_second": 5.795,
+      "step": 11000
+    },
+    {
+      "epoch": 3.217834469101287,
+      "grad_norm": 0.3275344967842102,
+      "learning_rate": 0.0005617167832167832,
+      "loss": 3.6293,
+      "step": 11050
+    },
+    {
+      "epoch": 3.2323955967150098,
+      "grad_norm": 0.3210190534591675,
+      "learning_rate": 0.000561541958041958,
+      "loss": 3.6476,
+      "step": 11100
+    },
+    {
+      "epoch": 3.246956724328732,
+      "grad_norm": 0.3466476798057556,
+      "learning_rate": 0.0005613671328671328,
+      "loss": 3.6433,
+      "step": 11150
+    },
+    {
+      "epoch": 3.2615178519424544,
+      "grad_norm": 0.3047979176044464,
+      "learning_rate": 0.0005611923076923077,
+      "loss": 3.6438,
+      "step": 11200
+    },
+    {
+      "epoch": 3.2760789795561767,
+      "grad_norm": 0.3035935163497925,
+      "learning_rate": 0.0005610174825174825,
+      "loss": 3.6351,
+      "step": 11250
+    },
+    {
+      "epoch": 3.2906401071698994,
+      "grad_norm": 0.3299195170402527,
+      "learning_rate": 0.0005608426573426573,
+      "loss": 3.6261,
+      "step": 11300
+    },
+    {
+      "epoch": 3.3052012347836217,
+      "grad_norm": 0.3068985044956207,
+      "learning_rate": 0.0005606678321678321,
+      "loss": 3.646,
+      "step": 11350
+    },
+    {
+      "epoch": 3.319762362397344,
+      "grad_norm": 0.31294307112693787,
+      "learning_rate": 0.000560493006993007,
+      "loss": 3.6391,
+      "step": 11400
+    },
+    {
+      "epoch": 3.3343234900110663,
+      "grad_norm": 0.3207429051399231,
+      "learning_rate": 0.0005603181818181818,
+      "loss": 3.6352,
+      "step": 11450
+    },
+    {
+      "epoch": 3.3488846176247886,
+      "grad_norm": 0.3064245581626892,
+      "learning_rate": 0.0005601433566433566,
+      "loss": 3.6414,
+      "step": 11500
+    },
+    {
+      "epoch": 3.3634457452385114,
+      "grad_norm": 0.32791194319725037,
+      "learning_rate": 0.0005599685314685314,
+      "loss": 3.6523,
+      "step": 11550
+    },
+    {
+      "epoch": 3.3780068728522337,
+      "grad_norm": 0.318341463804245,
+      "learning_rate": 0.0005597937062937063,
+      "loss": 3.6428,
+      "step": 11600
+    },
+    {
+      "epoch": 3.392568000465956,
+      "grad_norm": 0.30965712666511536,
+      "learning_rate": 0.0005596188811188811,
+      "loss": 3.6303,
+      "step": 11650
+    },
+    {
+      "epoch": 3.4071291280796787,
+      "grad_norm": 0.3249650001525879,
+      "learning_rate": 0.0005594440559440559,
+      "loss": 3.6413,
+      "step": 11700
+    },
+    {
+      "epoch": 3.421690255693401,
+      "grad_norm": 0.33683815598487854,
+      "learning_rate": 0.0005592692307692307,
+      "loss": 3.6245,
+      "step": 11750
+    },
+    {
+      "epoch": 3.4362513833071233,
+      "grad_norm": 0.32494112849235535,
+      "learning_rate": 0.0005590944055944055,
+      "loss": 3.6388,
+      "step": 11800
+    },
+    {
+      "epoch": 3.4508125109208456,
+      "grad_norm": 0.3111244738101959,
+      "learning_rate": 0.0005589195804195803,
+      "loss": 3.6472,
+      "step": 11850
+    },
+    {
+      "epoch": 3.465373638534568,
+      "grad_norm": 0.3108598589897156,
+      "learning_rate": 0.0005587447552447552,
+      "loss": 3.6419,
+      "step": 11900
+    },
+    {
+      "epoch": 3.4799347661482907,
+      "grad_norm": 0.32381555438041687,
+      "learning_rate": 0.00055856993006993,
+      "loss": 3.6358,
+      "step": 11950
+    },
+    {
+      "epoch": 3.494495893762013,
+      "grad_norm": 0.334005743265152,
+      "learning_rate": 0.0005583951048951048,
+      "loss": 3.6479,
+      "step": 12000
+    },
+    {
+      "epoch": 3.494495893762013,
+      "eval_accuracy": 0.3532355463240171,
+      "eval_loss": 3.6959714889526367,
+      "eval_runtime": 179.7113,
+      "eval_samples_per_second": 92.626,
+      "eval_steps_per_second": 5.793,
+      "step": 12000
+    },
+    {
+      "epoch": 3.5090570213757353,
+      "grad_norm": 0.3108784258365631,
+      "learning_rate": 0.0005582202797202797,
+      "loss": 3.6374,
+      "step": 12050
+    },
+    {
+      "epoch": 3.523618148989458,
+      "grad_norm": 0.31018486618995667,
+      "learning_rate": 0.0005580454545454545,
+      "loss": 3.6339,
+      "step": 12100
+    },
+    {
+      "epoch": 3.53817927660318,
+      "grad_norm": 0.3244931697845459,
+      "learning_rate": 0.0005578706293706293,
+      "loss": 3.6435,
+      "step": 12150
+    },
+    {
+      "epoch": 3.5527404042169026,
+      "grad_norm": 0.3162214457988739,
+      "learning_rate": 0.0005576958041958041,
+      "loss": 3.6443,
+      "step": 12200
+    },
+    {
+      "epoch": 3.567301531830625,
+      "grad_norm": 0.3021166920661926,
+      "learning_rate": 0.000557520979020979,
+      "loss": 3.653,
+      "step": 12250
+    },
+    {
+      "epoch": 3.5818626594443472,
+      "grad_norm": 0.31752142310142517,
+      "learning_rate": 0.0005573461538461538,
+      "loss": 3.6493,
+      "step": 12300
+    },
+    {
+      "epoch": 3.59642378705807,
+      "grad_norm": 0.3096977174282074,
+      "learning_rate": 0.0005571713286713286,
+      "loss": 3.647,
+      "step": 12350
+    },
+    {
+      "epoch": 3.6109849146717923,
+      "grad_norm": 0.33170512318611145,
+      "learning_rate": 0.0005569965034965034,
+      "loss": 3.641,
+      "step": 12400
+    },
+    {
+      "epoch": 3.6255460422855146,
+      "grad_norm": 0.324098140001297,
+      "learning_rate": 0.0005568216783216783,
+      "loss": 3.6424,
+      "step": 12450
+    },
+    {
+      "epoch": 3.640107169899237,
+      "grad_norm": 0.30960825085639954,
+      "learning_rate": 0.000556646853146853,
+      "loss": 3.6419,
+      "step": 12500
+    },
+    {
+      "epoch": 3.654668297512959,
+      "grad_norm": 0.3171733319759369,
+      "learning_rate": 0.0005564720279720279,
+      "loss": 3.6432,
+      "step": 12550
+    },
+    {
+      "epoch": 3.669229425126682,
+      "grad_norm": 0.3451572358608246,
+      "learning_rate": 0.0005562972027972027,
+      "loss": 3.639,
+      "step": 12600
+    },
+    {
+      "epoch": 3.6837905527404042,
+      "grad_norm": 0.32623323798179626,
+      "learning_rate": 0.0005561223776223775,
+      "loss": 3.652,
+      "step": 12650
+    },
+    {
+      "epoch": 3.6983516803541265,
+      "grad_norm": 0.3222995400428772,
+      "learning_rate": 0.0005559475524475524,
+      "loss": 3.6358,
+      "step": 12700
+    },
+    {
+      "epoch": 3.7129128079678493,
+      "grad_norm": 0.31685107946395874,
+      "learning_rate": 0.0005557727272727272,
+      "loss": 3.6335,
+      "step": 12750
+    },
+    {
+      "epoch": 3.7274739355815716,
+      "grad_norm": 0.296373188495636,
+      "learning_rate": 0.000555597902097902,
+      "loss": 3.6436,
+      "step": 12800
+    },
+    {
+      "epoch": 3.742035063195294,
+      "grad_norm": 0.3100621700286865,
+      "learning_rate": 0.0005554230769230768,
+      "loss": 3.627,
+      "step": 12850
+    },
+    {
+      "epoch": 3.756596190809016,
+      "grad_norm": 0.315127432346344,
+      "learning_rate": 0.0005552482517482517,
+      "loss": 3.6297,
+      "step": 12900
+    },
+    {
+      "epoch": 3.7711573184227385,
+      "grad_norm": 0.3003104329109192,
+      "learning_rate": 0.0005550734265734265,
+      "loss": 3.642,
+      "step": 12950
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "grad_norm": 0.31278714537620544,
+      "learning_rate": 0.0005548986013986013,
+      "loss": 3.6408,
+      "step": 13000
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "eval_accuracy": 0.3544306877906295,
+      "eval_loss": 3.6803815364837646,
+      "eval_runtime": 179.6757,
+      "eval_samples_per_second": 92.645,
+      "eval_steps_per_second": 5.794,
+      "step": 13000
+    },
+    {
+      "epoch": 3.8002795736501835,
+      "grad_norm": 0.30766820907592773,
+      "learning_rate": 0.0005547237762237761,
+      "loss": 3.6428,
+      "step": 13050
+    },
+    {
+      "epoch": 3.814840701263906,
+      "grad_norm": 0.31651240587234497,
+      "learning_rate": 0.000554548951048951,
+      "loss": 3.6435,
+      "step": 13100
+    },
+    {
+      "epoch": 3.829401828877628,
+      "grad_norm": 0.3394828736782074,
+      "learning_rate": 0.0005543741258741258,
+      "loss": 3.6463,
+      "step": 13150
+    },
+    {
+      "epoch": 3.8439629564913504,
+      "grad_norm": 0.32541000843048096,
+      "learning_rate": 0.0005541993006993006,
+      "loss": 3.64,
+      "step": 13200
+    },
+    {
+      "epoch": 3.858524084105073,
+      "grad_norm": 0.32382452487945557,
+      "learning_rate": 0.0005540244755244756,
+      "loss": 3.6398,
+      "step": 13250
+    },
+    {
+      "epoch": 3.8730852117187955,
+      "grad_norm": 0.3062750995159149,
+      "learning_rate": 0.0005538496503496502,
+      "loss": 3.6359,
+      "step": 13300
+    },
+    {
+      "epoch": 3.887646339332518,
+      "grad_norm": 0.3173046112060547,
+      "learning_rate": 0.0005536748251748252,
+      "loss": 3.6353,
+      "step": 13350
+    },
+    {
+      "epoch": 3.9022074669462405,
+      "grad_norm": 0.30907636880874634,
+      "learning_rate": 0.0005535,
+      "loss": 3.6324,
+      "step": 13400
+    },
+    {
+      "epoch": 3.916768594559963,
+      "grad_norm": 0.3197796642780304,
+      "learning_rate": 0.0005533251748251748,
+      "loss": 3.6239,
+      "step": 13450
+    },
+    {
+      "epoch": 3.931329722173685,
+      "grad_norm": 0.31007564067840576,
+      "learning_rate": 0.0005531503496503496,
+      "loss": 3.6527,
+      "step": 13500
+    },
+    {
+      "epoch": 3.9458908497874075,
+      "grad_norm": 0.3278939723968506,
+      "learning_rate": 0.0005529755244755245,
+      "loss": 3.6205,
+      "step": 13550
+    },
+    {
+      "epoch": 3.9604519774011298,
+      "grad_norm": 0.3057309091091156,
+      "learning_rate": 0.0005528006993006993,
+      "loss": 3.64,
+      "step": 13600
+    },
+    {
+      "epoch": 3.9750131050148525,
+      "grad_norm": 0.32244446873664856,
+      "learning_rate": 0.0005526258741258741,
+      "loss": 3.6376,
+      "step": 13650
+    },
+    {
+      "epoch": 3.989574232628575,
+      "grad_norm": 0.3175927400588989,
+      "learning_rate": 0.0005524510489510489,
+      "loss": 3.6282,
+      "step": 13700
+    },
+    {
+      "epoch": 4.004077115731842,
+      "grad_norm": 0.3246242105960846,
+      "learning_rate": 0.0005522762237762238,
+      "loss": 3.6123,
+      "step": 13750
+    },
+    {
+      "epoch": 4.018638243345564,
+      "grad_norm": 0.3010479211807251,
+      "learning_rate": 0.0005521013986013986,
+      "loss": 3.5191,
+      "step": 13800
+    },
+    {
+      "epoch": 4.033199370959287,
+      "grad_norm": 0.30843934416770935,
+      "learning_rate": 0.0005519265734265734,
+      "loss": 3.5335,
+      "step": 13850
+    },
+    {
+      "epoch": 4.04776049857301,
+      "grad_norm": 0.3538695275783539,
+      "learning_rate": 0.0005517517482517482,
+      "loss": 3.5185,
+      "step": 13900
+    },
+    {
+      "epoch": 4.062321626186732,
+      "grad_norm": 0.3226032257080078,
+      "learning_rate": 0.0005515769230769231,
+      "loss": 3.5432,
+      "step": 13950
+    },
+    {
+      "epoch": 4.076882753800454,
+      "grad_norm": 0.32624679803848267,
+      "learning_rate": 0.0005514020979020979,
+      "loss": 3.5399,
+      "step": 14000
+    },
+    {
+      "epoch": 4.076882753800454,
+      "eval_accuracy": 0.3558751795474921,
+      "eval_loss": 3.6733622550964355,
+      "eval_runtime": 179.6598,
+      "eval_samples_per_second": 92.653,
+      "eval_steps_per_second": 5.794,
+      "step": 14000
+    },
+    {
+      "epoch": 4.091443881414177,
+      "grad_norm": 0.3052377998828888,
+      "learning_rate": 0.0005512272727272727,
+      "loss": 3.5443,
+      "step": 14050
+    },
+    {
+      "epoch": 4.106005009027899,
+      "grad_norm": 0.32962766289711,
+      "learning_rate": 0.0005510524475524475,
+      "loss": 3.5366,
+      "step": 14100
+    },
+    {
+      "epoch": 4.120566136641622,
+      "grad_norm": 0.31939440965652466,
+      "learning_rate": 0.0005508776223776223,
+      "loss": 3.5529,
+      "step": 14150
+    },
+    {
+      "epoch": 4.135127264255344,
+      "grad_norm": 0.33249884843826294,
+      "learning_rate": 0.0005507027972027972,
+      "loss": 3.5555,
+      "step": 14200
+    },
+    {
+      "epoch": 4.149688391869066,
+      "grad_norm": 0.32969000935554504,
+      "learning_rate": 0.000550527972027972,
+      "loss": 3.5595,
+      "step": 14250
+    },
+    {
+      "epoch": 4.164249519482789,
+      "grad_norm": 0.31157979369163513,
+      "learning_rate": 0.0005503531468531468,
+      "loss": 3.5487,
+      "step": 14300
+    },
+    {
+      "epoch": 4.178810647096511,
+      "grad_norm": 0.31714004278182983,
+      "learning_rate": 0.0005501783216783216,
+      "loss": 3.5475,
+      "step": 14350
+    },
+    {
+      "epoch": 4.193371774710234,
+      "grad_norm": 0.31385311484336853,
+      "learning_rate": 0.0005500034965034965,
+      "loss": 3.5714,
+      "step": 14400
+    },
+    {
+      "epoch": 4.207932902323956,
+      "grad_norm": 0.318708598613739,
+      "learning_rate": 0.0005498286713286713,
+      "loss": 3.5619,
+      "step": 14450
+    },
+    {
+      "epoch": 4.222494029937678,
+      "grad_norm": 0.3174968957901001,
+      "learning_rate": 0.0005496538461538461,
+      "loss": 3.5536,
+      "step": 14500
+    },
+    {
+      "epoch": 4.237055157551401,
+      "grad_norm": 0.31125175952911377,
+      "learning_rate": 0.0005494790209790209,
+      "loss": 3.5567,
+      "step": 14550
+    },
+    {
+      "epoch": 4.251616285165123,
+      "grad_norm": 0.3577822744846344,
+      "learning_rate": 0.0005493041958041958,
+      "loss": 3.5653,
+      "step": 14600
+    },
+    {
+      "epoch": 4.266177412778846,
+      "grad_norm": 0.3287855386734009,
+      "learning_rate": 0.0005491293706293706,
+      "loss": 3.5722,
+      "step": 14650
+    },
+    {
+      "epoch": 4.280738540392568,
+      "grad_norm": 0.3440224528312683,
+      "learning_rate": 0.0005489545454545454,
+      "loss": 3.5661,
+      "step": 14700
+    },
+    {
+      "epoch": 4.29529966800629,
+      "grad_norm": 0.320300430059433,
+      "learning_rate": 0.0005487797202797203,
+      "loss": 3.5702,
+      "step": 14750
+    },
+    {
+      "epoch": 4.309860795620013,
+      "grad_norm": 0.31104516983032227,
+      "learning_rate": 0.000548604895104895,
+      "loss": 3.574,
+      "step": 14800
+    },
+    {
+      "epoch": 4.324421923233735,
+      "grad_norm": 0.31709253787994385,
+      "learning_rate": 0.0005484300699300699,
+      "loss": 3.5653,
+      "step": 14850
+    },
+    {
+      "epoch": 4.338983050847458,
+      "grad_norm": 0.32527729868888855,
+      "learning_rate": 0.0005482552447552447,
+      "loss": 3.5626,
+      "step": 14900
+    },
+    {
+      "epoch": 4.35354417846118,
+      "grad_norm": 0.3272615671157837,
+      "learning_rate": 0.0005480804195804195,
+      "loss": 3.5621,
+      "step": 14950
+    },
+    {
+      "epoch": 4.368105306074902,
+      "grad_norm": 0.31901460886001587,
+      "learning_rate": 0.0005479055944055943,
+      "loss": 3.5705,
+      "step": 15000
+    },
+    {
+      "epoch": 4.368105306074902,
+      "eval_accuracy": 0.3572903982151175,
+      "eval_loss": 3.659005641937256,
+      "eval_runtime": 179.666,
+      "eval_samples_per_second": 92.65,
+      "eval_steps_per_second": 5.794,
+      "step": 15000
+    },
+    {
+      "epoch": 4.382666433688625,
+      "grad_norm": 0.3244258463382721,
+      "learning_rate": 0.0005477307692307692,
+      "loss": 3.5795,
+      "step": 15050
+    },
+    {
+      "epoch": 4.397227561302348,
+      "grad_norm": 0.32407137751579285,
+      "learning_rate": 0.000547555944055944,
+      "loss": 3.583,
+      "step": 15100
+    },
+    {
+      "epoch": 4.41178868891607,
+      "grad_norm": 0.32865169644355774,
+      "learning_rate": 0.0005473811188811188,
+      "loss": 3.5689,
+      "step": 15150
+    },
+    {
+      "epoch": 4.426349816529792,
+      "grad_norm": 0.33799371123313904,
+      "learning_rate": 0.0005472062937062936,
+      "loss": 3.5609,
+      "step": 15200
+    },
+    {
+      "epoch": 4.440910944143514,
+      "grad_norm": 0.33456024527549744,
+      "learning_rate": 0.0005470314685314685,
+      "loss": 3.5751,
+      "step": 15250
+    },
+    {
+      "epoch": 4.455472071757237,
+      "grad_norm": 0.3036324977874756,
+      "learning_rate": 0.0005468566433566433,
+      "loss": 3.5639,
+      "step": 15300
+    },
+    {
+      "epoch": 4.47003319937096,
+      "grad_norm": 0.3154810667037964,
+      "learning_rate": 0.0005466818181818181,
+      "loss": 3.5711,
+      "step": 15350
+    },
+    {
+      "epoch": 4.4845943269846815,
+      "grad_norm": 0.33000850677490234,
+      "learning_rate": 0.000546506993006993,
+      "loss": 3.5648,
+      "step": 15400
+    },
+    {
+      "epoch": 4.499155454598404,
+      "grad_norm": 0.30540409684181213,
+      "learning_rate": 0.0005463321678321678,
+      "loss": 3.5704,
+      "step": 15450
+    },
+    {
+      "epoch": 4.513716582212126,
+      "grad_norm": 0.30891209840774536,
+      "learning_rate": 0.0005461573426573426,
+      "loss": 3.5838,
+      "step": 15500
+    },
+    {
+      "epoch": 4.528277709825849,
+      "grad_norm": 0.31514647603034973,
+      "learning_rate": 0.0005459825174825174,
+      "loss": 3.5842,
+      "step": 15550
+    },
+    {
+      "epoch": 4.542838837439572,
+      "grad_norm": 0.3225398361682892,
+      "learning_rate": 0.0005458076923076922,
+      "loss": 3.573,
+      "step": 15600
+    },
+    {
+      "epoch": 4.5573999650532935,
+      "grad_norm": 0.330959290266037,
+      "learning_rate": 0.000545632867132867,
+      "loss": 3.5666,
+      "step": 15650
+    },
+    {
+      "epoch": 4.571961092667016,
+      "grad_norm": 0.33557143807411194,
+      "learning_rate": 0.0005454580419580419,
+      "loss": 3.5644,
+      "step": 15700
+    },
+    {
+      "epoch": 4.586522220280738,
+      "grad_norm": 0.32927149534225464,
+      "learning_rate": 0.0005452832167832167,
+      "loss": 3.5739,
+      "step": 15750
+    },
+    {
+      "epoch": 4.601083347894461,
+      "grad_norm": 0.34957724809646606,
+      "learning_rate": 0.0005451083916083915,
+      "loss": 3.5869,
+      "step": 15800
+    },
+    {
+      "epoch": 4.615644475508184,
+      "grad_norm": 0.30984318256378174,
+      "learning_rate": 0.0005449335664335663,
+      "loss": 3.5735,
+      "step": 15850
+    },
+    {
+      "epoch": 4.630205603121905,
+      "grad_norm": 0.31692495942115784,
+      "learning_rate": 0.0005447587412587412,
+      "loss": 3.5705,
+      "step": 15900
+    },
+    {
+      "epoch": 4.644766730735628,
+      "grad_norm": 0.3213113248348236,
+      "learning_rate": 0.000544583916083916,
+      "loss": 3.5775,
+      "step": 15950
+    },
+    {
+      "epoch": 4.659327858349351,
+      "grad_norm": 0.31125399470329285,
+      "learning_rate": 0.0005444090909090908,
+      "loss": 3.5765,
+      "step": 16000
+    },
+    {
+      "epoch": 4.659327858349351,
+      "eval_accuracy": 0.35853856041765764,
+      "eval_loss": 3.6479623317718506,
+      "eval_runtime": 179.6168,
+      "eval_samples_per_second": 92.675,
+      "eval_steps_per_second": 5.796,
+      "step": 16000
+    },
+    {
+      "epoch": 4.673888985963073,
+      "grad_norm": 0.33686563372612,
+      "learning_rate": 0.0005442342657342657,
+      "loss": 3.5719,
+      "step": 16050
+    },
+    {
+      "epoch": 4.6884501135767955,
+      "grad_norm": 0.33824530243873596,
+      "learning_rate": 0.0005440594405594405,
+      "loss": 3.5811,
+      "step": 16100
+    },
+    {
+      "epoch": 4.703011241190518,
+      "grad_norm": 0.31893661618232727,
+      "learning_rate": 0.0005438846153846153,
+      "loss": 3.5637,
+      "step": 16150
+    },
+    {
+      "epoch": 4.71757236880424,
+      "grad_norm": 0.3194436728954315,
+      "learning_rate": 0.0005437097902097901,
+      "loss": 3.5643,
+      "step": 16200
+    },
+    {
+      "epoch": 4.732133496417963,
+      "grad_norm": 0.3170466721057892,
+      "learning_rate": 0.0005435349650349651,
+      "loss": 3.5724,
+      "step": 16250
+    },
+    {
+      "epoch": 4.746694624031685,
+      "grad_norm": 0.3126927614212036,
+      "learning_rate": 0.0005433601398601397,
+      "loss": 3.577,
+      "step": 16300
+    },
+    {
+      "epoch": 4.7612557516454075,
+      "grad_norm": 0.32292503118515015,
+      "learning_rate": 0.0005431853146853147,
+      "loss": 3.5742,
+      "step": 16350
+    },
+    {
+      "epoch": 4.77581687925913,
+      "grad_norm": 0.31847289204597473,
+      "learning_rate": 0.0005430104895104895,
+      "loss": 3.5728,
+      "step": 16400
+    },
+    {
+      "epoch": 4.790378006872852,
+      "grad_norm": 0.3244653642177582,
+      "learning_rate": 0.0005428356643356643,
+      "loss": 3.569,
+      "step": 16450
+    },
+    {
+      "epoch": 4.804939134486575,
+      "grad_norm": 0.3252134323120117,
+      "learning_rate": 0.0005426608391608391,
+      "loss": 3.578,
+      "step": 16500
+    },
+    {
+      "epoch": 4.819500262100297,
+      "grad_norm": 0.3335552215576172,
+      "learning_rate": 0.000542486013986014,
+      "loss": 3.5879,
+      "step": 16550
+    },
+    {
+      "epoch": 4.834061389714019,
+      "grad_norm": 0.32028138637542725,
+      "learning_rate": 0.0005423111888111888,
+      "loss": 3.5684,
+      "step": 16600
+    },
+    {
+      "epoch": 4.848622517327742,
+      "grad_norm": 0.35907861590385437,
+      "learning_rate": 0.0005421363636363636,
+      "loss": 3.5713,
+      "step": 16650
+    },
+    {
+      "epoch": 4.863183644941464,
+      "grad_norm": 0.3116554617881775,
+      "learning_rate": 0.0005419615384615385,
+      "loss": 3.569,
+      "step": 16700
+    },
+    {
+      "epoch": 4.877744772555187,
+      "grad_norm": 0.32443204522132874,
+      "learning_rate": 0.0005417867132867133,
+      "loss": 3.5691,
+      "step": 16750
+    },
+    {
+      "epoch": 4.892305900168909,
+      "grad_norm": 0.32515308260917664,
+      "learning_rate": 0.0005416118881118881,
+      "loss": 3.5714,
+      "step": 16800
+    },
+    {
+      "epoch": 4.906867027782631,
+      "grad_norm": 0.3074589669704437,
+      "learning_rate": 0.0005414370629370629,
+      "loss": 3.5773,
+      "step": 16850
+    },
+    {
+      "epoch": 4.921428155396354,
+      "grad_norm": 0.32288840413093567,
+      "learning_rate": 0.0005412622377622378,
+      "loss": 3.5677,
+      "step": 16900
+    },
+    {
+      "epoch": 4.935989283010076,
+      "grad_norm": 0.314647912979126,
+      "learning_rate": 0.0005410874125874126,
+      "loss": 3.5722,
+      "step": 16950
+    },
+    {
+      "epoch": 4.950550410623799,
+      "grad_norm": 0.33100083470344543,
+      "learning_rate": 0.0005409125874125874,
+      "loss": 3.5786,
+      "step": 17000
+    },
+    {
+      "epoch": 4.950550410623799,
+      "eval_accuracy": 0.35935820691630227,
+      "eval_loss": 3.6380977630615234,
+      "eval_runtime": 179.6744,
+      "eval_samples_per_second": 92.645,
+      "eval_steps_per_second": 5.794,
+      "step": 17000
+    },
+    {
+      "epoch": 4.9651115382375215,
+      "grad_norm": 0.30516111850738525,
+      "learning_rate": 0.0005407377622377622,
+      "loss": 3.5811,
+      "step": 17050
+    },
+    {
+      "epoch": 4.979672665851243,
+      "grad_norm": 0.32663649320602417,
+      "learning_rate": 0.000540562937062937,
+      "loss": 3.5675,
+      "step": 17100
+    },
+    {
+      "epoch": 4.994233793464966,
+      "grad_norm": 0.3119376599788666,
+      "learning_rate": 0.0005403881118881118,
+      "loss": 3.5709,
+      "step": 17150
+    },
+    {
+      "epoch": 5.008736676568233,
+      "grad_norm": 0.32153835892677307,
+      "learning_rate": 0.0005402132867132867,
+      "loss": 3.5135,
+      "step": 17200
+    },
+    {
+      "epoch": 5.023297804181956,
+      "grad_norm": 0.3149344325065613,
+      "learning_rate": 0.0005400384615384615,
+      "loss": 3.4688,
+      "step": 17250
+    },
+    {
+      "epoch": 5.037858931795678,
+      "grad_norm": 0.31834840774536133,
+      "learning_rate": 0.0005398636363636363,
+      "loss": 3.4763,
+      "step": 17300
+    },
+    {
+      "epoch": 5.052420059409401,
+      "grad_norm": 0.31698572635650635,
+      "learning_rate": 0.0005396888111888111,
+      "loss": 3.4748,
+      "step": 17350
+    },
+    {
+      "epoch": 5.066981187023123,
+      "grad_norm": 0.33012357354164124,
+      "learning_rate": 0.000539513986013986,
+      "loss": 3.4676,
+      "step": 17400
+    },
+    {
+      "epoch": 5.081542314636845,
+      "grad_norm": 0.3190668821334839,
+      "learning_rate": 0.0005393391608391608,
+      "loss": 3.4737,
+      "step": 17450
+    },
+    {
+      "epoch": 5.096103442250568,
+      "grad_norm": 0.3394056260585785,
+      "learning_rate": 0.0005391643356643356,
+      "loss": 3.4838,
+      "step": 17500
+    },
+    {
+      "epoch": 5.110664569864291,
+      "grad_norm": 0.33012598752975464,
+      "learning_rate": 0.0005389895104895105,
+      "loss": 3.4785,
+      "step": 17550
+    },
+    {
+      "epoch": 5.125225697478013,
+      "grad_norm": 0.32027074694633484,
+      "learning_rate": 0.0005388146853146853,
+      "loss": 3.4866,
+      "step": 17600
+    },
+    {
+      "epoch": 5.139786825091735,
+      "grad_norm": 0.34201139211654663,
+      "learning_rate": 0.0005386398601398601,
+      "loss": 3.4902,
+      "step": 17650
+    },
+    {
+      "epoch": 5.154347952705457,
+      "grad_norm": 0.3230908513069153,
+      "learning_rate": 0.0005384650349650349,
+      "loss": 3.4993,
+      "step": 17700
+    },
+    {
+      "epoch": 5.16890908031918,
+      "grad_norm": 0.34298035502433777,
+      "learning_rate": 0.0005382902097902098,
+      "loss": 3.4852,
+      "step": 17750
+    },
+    {
+      "epoch": 5.183470207932903,
+      "grad_norm": 0.32764288783073425,
+      "learning_rate": 0.0005381153846153845,
+      "loss": 3.492,
+      "step": 17800
+    },
+    {
+      "epoch": 5.1980313355466246,
+      "grad_norm": 0.3107599914073944,
+      "learning_rate": 0.0005379405594405594,
+      "loss": 3.5035,
+      "step": 17850
+    },
+    {
+      "epoch": 5.212592463160347,
+      "grad_norm": 0.3222023844718933,
+      "learning_rate": 0.0005377657342657342,
+      "loss": 3.5083,
+      "step": 17900
+    },
+    {
+      "epoch": 5.227153590774069,
+      "grad_norm": 0.328563928604126,
+      "learning_rate": 0.000537590909090909,
+      "loss": 3.5112,
+      "step": 17950
+    },
+    {
+      "epoch": 5.241714718387792,
+      "grad_norm": 0.31996095180511475,
+      "learning_rate": 0.0005374160839160838,
+      "loss": 3.4988,
+      "step": 18000
+    },
+    {
+      "epoch": 5.241714718387792,
+      "eval_accuracy": 0.3599712959137824,
+      "eval_loss": 3.6374216079711914,
+      "eval_runtime": 179.7436,
+      "eval_samples_per_second": 92.61,
+      "eval_steps_per_second": 5.792,
+      "step": 18000
+    },
+    {
+      "epoch": 5.256275846001515,
+      "grad_norm": 0.3201790750026703,
+      "learning_rate": 0.0005372412587412587,
+      "loss": 3.5086,
+      "step": 18050
+    },
+    {
+      "epoch": 5.2708369736152365,
+      "grad_norm": 0.33598074316978455,
+      "learning_rate": 0.0005370664335664335,
+      "loss": 3.5094,
+      "step": 18100
+    },
+    {
+      "epoch": 5.285398101228959,
+      "grad_norm": 0.3486185073852539,
+      "learning_rate": 0.0005368916083916083,
+      "loss": 3.5194,
+      "step": 18150
+    },
+    {
+      "epoch": 5.299959228842681,
+      "grad_norm": 0.3399152159690857,
+      "learning_rate": 0.0005367167832167832,
+      "loss": 3.5113,
+      "step": 18200
+    },
+    {
+      "epoch": 5.314520356456404,
+      "grad_norm": 0.3166239559650421,
+      "learning_rate": 0.000536541958041958,
+      "loss": 3.5098,
+      "step": 18250
+    },
+    {
+      "epoch": 5.329081484070127,
+      "grad_norm": 0.33888429403305054,
+      "learning_rate": 0.0005363671328671328,
+      "loss": 3.5138,
+      "step": 18300
+    },
+    {
+      "epoch": 5.3436426116838485,
+      "grad_norm": 0.3163507878780365,
+      "learning_rate": 0.0005361923076923076,
+      "loss": 3.5154,
+      "step": 18350
+    },
+    {
+      "epoch": 5.358203739297571,
+      "grad_norm": 0.3177511990070343,
+      "learning_rate": 0.0005360174825174825,
+      "loss": 3.5125,
+      "step": 18400
+    },
+    {
+      "epoch": 5.372764866911294,
+      "grad_norm": 0.30713725090026855,
+      "learning_rate": 0.0005358426573426573,
+      "loss": 3.5124,
+      "step": 18450
+    },
+    {
+      "epoch": 5.387325994525016,
+      "grad_norm": 0.3410167694091797,
+      "learning_rate": 0.0005356678321678321,
+      "loss": 3.5107,
+      "step": 18500
+    },
+    {
+      "epoch": 5.401887122138739,
+      "grad_norm": 0.32011181116104126,
+      "learning_rate": 0.0005354930069930069,
+      "loss": 3.5171,
+      "step": 18550
+    },
+    {
+      "epoch": 5.41644824975246,
+      "grad_norm": 0.33928465843200684,
+      "learning_rate": 0.0005353181818181817,
+      "loss": 3.5235,
+      "step": 18600
+    },
+    {
+      "epoch": 5.431009377366183,
+      "grad_norm": 0.3426041901111603,
+      "learning_rate": 0.0005351433566433565,
+      "loss": 3.5077,
+      "step": 18650
+    },
+    {
+      "epoch": 5.445570504979906,
+      "grad_norm": 0.3379696309566498,
+      "learning_rate": 0.0005349685314685314,
+      "loss": 3.515,
+      "step": 18700
+    },
+    {
+      "epoch": 5.460131632593628,
+      "grad_norm": 0.3430345356464386,
+      "learning_rate": 0.0005347937062937062,
+      "loss": 3.5124,
+      "step": 18750
+    },
+    {
+      "epoch": 5.4746927602073505,
+      "grad_norm": 0.3233330547809601,
+      "learning_rate": 0.000534618881118881,
+      "loss": 3.5201,
+      "step": 18800
+    },
+    {
+      "epoch": 5.489253887821073,
+      "grad_norm": 0.353807270526886,
+      "learning_rate": 0.0005344440559440559,
+      "loss": 3.5234,
+      "step": 18850
+    },
+    {
+      "epoch": 5.503815015434795,
+      "grad_norm": 0.3057951033115387,
+      "learning_rate": 0.0005342692307692307,
+      "loss": 3.5253,
+      "step": 18900
+    },
+    {
+      "epoch": 5.518376143048518,
+      "grad_norm": 0.3360867202281952,
+      "learning_rate": 0.0005340944055944055,
+      "loss": 3.5197,
+      "step": 18950
+    },
+    {
+      "epoch": 5.53293727066224,
+      "grad_norm": 0.31620731949806213,
+      "learning_rate": 0.0005339195804195803,
+      "loss": 3.513,
+      "step": 19000
+    },
+    {
+      "epoch": 5.53293727066224,
+      "eval_accuracy": 0.36111988258787275,
+      "eval_loss": 3.6275086402893066,
+      "eval_runtime": 179.6266,
+      "eval_samples_per_second": 92.67,
+      "eval_steps_per_second": 5.795,
+      "step": 19000
+    },
+    {
+      "epoch": 5.5474983982759625,
+      "grad_norm": 0.3172198235988617,
+      "learning_rate": 0.0005337447552447552,
+      "loss": 3.5143,
+      "step": 19050
+    },
+    {
+      "epoch": 5.562059525889685,
+      "grad_norm": 0.30855488777160645,
+      "learning_rate": 0.00053356993006993,
+      "loss": 3.5163,
+      "step": 19100
+    },
+    {
+      "epoch": 5.576620653503407,
+      "grad_norm": 0.320773720741272,
+      "learning_rate": 0.0005333951048951048,
+      "loss": 3.5108,
+      "step": 19150
+    },
+    {
+      "epoch": 5.59118178111713,
+      "grad_norm": 0.33892980217933655,
+      "learning_rate": 0.0005332202797202796,
+      "loss": 3.5252,
+      "step": 19200
+    },
+    {
+      "epoch": 5.605742908730852,
+      "grad_norm": 0.34134751558303833,
+      "learning_rate": 0.0005330454545454546,
+      "loss": 3.5172,
+      "step": 19250
+    },
+    {
+      "epoch": 5.620304036344574,
+      "grad_norm": 0.3078143000602722,
+      "learning_rate": 0.0005328706293706292,
+      "loss": 3.5308,
+      "step": 19300
+    },
+    {
+      "epoch": 5.634865163958297,
+      "grad_norm": 0.34207335114479065,
+      "learning_rate": 0.0005326958041958042,
+      "loss": 3.5397,
+      "step": 19350
+    },
+    {
+      "epoch": 5.649426291572019,
+      "grad_norm": 0.32775601744651794,
+      "learning_rate": 0.000532520979020979,
+      "loss": 3.5251,
+      "step": 19400
+    },
+    {
+      "epoch": 5.663987419185742,
+      "grad_norm": 0.30578893423080444,
+      "learning_rate": 0.0005323461538461538,
+      "loss": 3.5262,
+      "step": 19450
+    },
+    {
+      "epoch": 5.6785485467994645,
+      "grad_norm": 0.3056182563304901,
+      "learning_rate": 0.0005321713286713287,
+      "loss": 3.525,
+      "step": 19500
+    },
+    {
+      "epoch": 5.693109674413186,
+      "grad_norm": 0.35645192861557007,
+      "learning_rate": 0.0005319965034965035,
+      "loss": 3.5154,
+      "step": 19550
+    },
+    {
+      "epoch": 5.707670802026909,
+      "grad_norm": 0.32187339663505554,
+      "learning_rate": 0.0005318216783216783,
+      "loss": 3.5408,
+      "step": 19600
+    },
+    {
+      "epoch": 5.722231929640631,
+      "grad_norm": 0.33344683051109314,
+      "learning_rate": 0.0005316468531468531,
+      "loss": 3.5218,
+      "step": 19650
+    },
+    {
+      "epoch": 5.736793057254354,
+      "grad_norm": 0.30798256397247314,
+      "learning_rate": 0.000531472027972028,
+      "loss": 3.5235,
+      "step": 19700
+    },
+    {
+      "epoch": 5.7513541848680765,
+      "grad_norm": 0.34034493565559387,
+      "learning_rate": 0.0005312972027972028,
+      "loss": 3.5312,
+      "step": 19750
+    },
+    {
+      "epoch": 5.765915312481798,
+      "grad_norm": 0.335997611284256,
+      "learning_rate": 0.0005311223776223776,
+      "loss": 3.5245,
+      "step": 19800
+    },
+    {
+      "epoch": 5.780476440095521,
+      "grad_norm": 0.34608331322669983,
+      "learning_rate": 0.0005309475524475524,
+      "loss": 3.5313,
+      "step": 19850
+    },
+    {
+      "epoch": 5.795037567709244,
+      "grad_norm": 0.31513699889183044,
+      "learning_rate": 0.0005307727272727273,
+      "loss": 3.5271,
+      "step": 19900
+    },
+    {
+      "epoch": 5.809598695322966,
+      "grad_norm": 0.3229629695415497,
+      "learning_rate": 0.0005305979020979021,
+      "loss": 3.5301,
+      "step": 19950
+    },
+    {
+      "epoch": 5.824159822936688,
+      "grad_norm": 0.3298615515232086,
+      "learning_rate": 0.0005304230769230769,
+      "loss": 3.5213,
+      "step": 20000
+    },
+    {
+      "epoch": 5.824159822936688,
+      "eval_accuracy": 0.3618738115889927,
+      "eval_loss": 3.6149404048919678,
+      "eval_runtime": 179.7404,
+      "eval_samples_per_second": 92.611,
+      "eval_steps_per_second": 5.792,
+      "step": 20000
+    },
+    {
+      "epoch": 5.83872095055041,
+      "grad_norm": 0.33221063017845154,
+      "learning_rate": 0.0005302482517482517,
+      "loss": 3.5375,
+      "step": 20050
+    },
+    {
+      "epoch": 5.853282078164133,
+      "grad_norm": 0.31011486053466797,
+      "learning_rate": 0.0005300734265734265,
+      "loss": 3.5339,
+      "step": 20100
+    },
+    {
+      "epoch": 5.867843205777856,
+      "grad_norm": 0.31870928406715393,
+      "learning_rate": 0.0005298986013986013,
+      "loss": 3.5193,
+      "step": 20150
+    },
+    {
+      "epoch": 5.882404333391578,
+      "grad_norm": 0.31526702642440796,
+      "learning_rate": 0.0005297237762237762,
+      "loss": 3.526,
+      "step": 20200
+    },
+    {
+      "epoch": 5.8969654610053,
+      "grad_norm": 0.33031392097473145,
+      "learning_rate": 0.000529548951048951,
+      "loss": 3.5219,
+      "step": 20250
+    },
+    {
+      "epoch": 5.911526588619022,
+      "grad_norm": 0.29852011799812317,
+      "learning_rate": 0.0005293741258741258,
+      "loss": 3.5221,
+      "step": 20300
+    },
+    {
+      "epoch": 5.926087716232745,
+      "grad_norm": 0.32039979100227356,
+      "learning_rate": 0.0005291993006993007,
+      "loss": 3.5277,
+      "step": 20350
+    },
+    {
+      "epoch": 5.940648843846468,
+      "grad_norm": 0.3331441879272461,
+      "learning_rate": 0.0005290244755244755,
+      "loss": 3.5332,
+      "step": 20400
+    },
+    {
+      "epoch": 5.95520997146019,
+      "grad_norm": 0.3218717873096466,
+      "learning_rate": 0.0005288496503496503,
+      "loss": 3.5284,
+      "step": 20450
+    },
+    {
+      "epoch": 5.969771099073912,
+      "grad_norm": 0.322635680437088,
+      "learning_rate": 0.0005286748251748251,
+      "loss": 3.5293,
+      "step": 20500
+    },
+    {
+      "epoch": 5.984332226687634,
+      "grad_norm": 0.3240005075931549,
+      "learning_rate": 0.0005285,
+      "loss": 3.5176,
+      "step": 20550
+    },
+    {
+      "epoch": 5.998893354301357,
+      "grad_norm": 0.3016450107097626,
+      "learning_rate": 0.0005283251748251748,
+      "loss": 3.5365,
+      "step": 20600
+    },
+    {
+      "epoch": 6.013396237404625,
+      "grad_norm": 0.32817426323890686,
+      "learning_rate": 0.0005281503496503496,
+      "loss": 3.4283,
+      "step": 20650
+    },
+    {
+      "epoch": 6.027957365018347,
+      "grad_norm": 0.35756346583366394,
+      "learning_rate": 0.0005279755244755244,
+      "loss": 3.4236,
+      "step": 20700
+    },
+    {
+      "epoch": 6.04251849263207,
+      "grad_norm": 0.32357922196388245,
+      "learning_rate": 0.0005278006993006993,
+      "loss": 3.4285,
+      "step": 20750
+    },
+    {
+      "epoch": 6.0570796202457915,
+      "grad_norm": 0.3198466897010803,
+      "learning_rate": 0.000527625874125874,
+      "loss": 3.4119,
+      "step": 20800
+    },
+    {
+      "epoch": 6.071640747859514,
+      "grad_norm": 0.3190059959888458,
+      "learning_rate": 0.0005274510489510489,
+      "loss": 3.4369,
+      "step": 20850
+    },
+    {
+      "epoch": 6.086201875473237,
+      "grad_norm": 0.33216336369514465,
+      "learning_rate": 0.0005272762237762238,
+      "loss": 3.4332,
+      "step": 20900
+    },
+    {
+      "epoch": 6.100763003086959,
+      "grad_norm": 0.3313469886779785,
+      "learning_rate": 0.0005271013986013985,
+      "loss": 3.4451,
+      "step": 20950
+    },
+    {
+      "epoch": 6.115324130700682,
+      "grad_norm": 0.3283461630344391,
+      "learning_rate": 0.0005269265734265734,
+      "loss": 3.4434,
+      "step": 21000
+    },
+    {
+      "epoch": 6.115324130700682,
+      "eval_accuracy": 0.36220145857575725,
+      "eval_loss": 3.6175012588500977,
+      "eval_runtime": 179.6715,
+      "eval_samples_per_second": 92.647,
+      "eval_steps_per_second": 5.794,
+      "step": 21000
+    },
+    {
+      "epoch": 6.1298852583144035,
+      "grad_norm": 0.3373625576496124,
+      "learning_rate": 0.0005267517482517482,
+      "loss": 3.4382,
+      "step": 21050
+    },
+    {
+      "epoch": 6.144446385928126,
+      "grad_norm": 0.3465948700904846,
+      "learning_rate": 0.000526576923076923,
+      "loss": 3.4579,
+      "step": 21100
+    },
+    {
+      "epoch": 6.159007513541849,
+      "grad_norm": 0.32088997960090637,
+      "learning_rate": 0.0005264020979020978,
+      "loss": 3.4415,
+      "step": 21150
+    },
+    {
+      "epoch": 6.173568641155571,
+      "grad_norm": 0.34192368388175964,
+      "learning_rate": 0.0005262272727272727,
+      "loss": 3.4381,
+      "step": 21200
+    },
+    {
+      "epoch": 6.1881297687692935,
+      "grad_norm": 0.3320399224758148,
+      "learning_rate": 0.0005260524475524475,
+      "loss": 3.4543,
+      "step": 21250
+    },
+    {
+      "epoch": 6.202690896383016,
+      "grad_norm": 0.3265776038169861,
+      "learning_rate": 0.0005258776223776223,
+      "loss": 3.4382,
+      "step": 21300
+    },
+    {
+      "epoch": 6.217252023996738,
+      "grad_norm": 0.32715266942977905,
+      "learning_rate": 0.0005257027972027971,
+      "loss": 3.4466,
+      "step": 21350
+    },
+    {
+      "epoch": 6.231813151610461,
+      "grad_norm": 0.342515766620636,
+      "learning_rate": 0.000525527972027972,
+      "loss": 3.4544,
+      "step": 21400
+    },
+    {
+      "epoch": 6.246374279224183,
+      "grad_norm": 0.32371506094932556,
+      "learning_rate": 0.0005253531468531468,
+      "loss": 3.4531,
+      "step": 21450
+    },
+    {
+      "epoch": 6.2609354068379055,
+      "grad_norm": 0.3336334228515625,
+      "learning_rate": 0.0005251783216783216,
+      "loss": 3.448,
+      "step": 21500
+    },
+    {
+      "epoch": 6.275496534451628,
+      "grad_norm": 0.3309522867202759,
+      "learning_rate": 0.0005250034965034965,
+      "loss": 3.4578,
+      "step": 21550
+    },
+    {
+      "epoch": 6.29005766206535,
+      "grad_norm": 0.3392053544521332,
+      "learning_rate": 0.0005248286713286712,
+      "loss": 3.4488,
+      "step": 21600
+    },
+    {
+      "epoch": 6.304618789679073,
+      "grad_norm": 0.319116473197937,
+      "learning_rate": 0.0005246538461538461,
+      "loss": 3.4704,
+      "step": 21650
+    },
+    {
+      "epoch": 6.319179917292795,
+      "grad_norm": 0.34017646312713623,
+      "learning_rate": 0.0005244790209790209,
+      "loss": 3.4707,
+      "step": 21700
+    },
+    {
+      "epoch": 6.3337410449065175,
+      "grad_norm": 0.3570762872695923,
+      "learning_rate": 0.0005243041958041957,
+      "loss": 3.4818,
+      "step": 21750
+    },
+    {
+      "epoch": 6.34830217252024,
+      "grad_norm": 0.32639503479003906,
+      "learning_rate": 0.0005241293706293705,
+      "loss": 3.4707,
+      "step": 21800
+    },
+    {
+      "epoch": 6.362863300133962,
+      "grad_norm": 0.3118131458759308,
+      "learning_rate": 0.0005239545454545454,
+      "loss": 3.474,
+      "step": 21850
+    },
+    {
+      "epoch": 6.377424427747685,
+      "grad_norm": 0.334991991519928,
+      "learning_rate": 0.0005237797202797202,
+      "loss": 3.4688,
+      "step": 21900
+    },
+    {
+      "epoch": 6.391985555361408,
+      "grad_norm": 0.3370341956615448,
+      "learning_rate": 0.000523604895104895,
+      "loss": 3.4606,
+      "step": 21950
+    },
+    {
+      "epoch": 6.406546682975129,
+      "grad_norm": 0.30481281876564026,
+      "learning_rate": 0.0005234300699300698,
+      "loss": 3.4748,
+      "step": 22000
+    },
+    {
+      "epoch": 6.406546682975129,
+      "eval_accuracy": 0.3628913159558557,
+      "eval_loss": 3.6097097396850586,
+      "eval_runtime": 179.7178,
+      "eval_samples_per_second": 92.623,
+      "eval_steps_per_second": 5.792,
+      "step": 22000
+    },
+    {
+      "epoch": 6.421107810588852,
+      "grad_norm": 0.31929466128349304,
+      "learning_rate": 0.0005232552447552447,
+      "loss": 3.4819,
+      "step": 22050
+    },
+    {
+      "epoch": 6.435668938202574,
+      "grad_norm": 0.3431412875652313,
+      "learning_rate": 0.0005230804195804195,
+      "loss": 3.4821,
+      "step": 22100
+    },
+    {
+      "epoch": 6.450230065816297,
+      "grad_norm": 0.31847938895225525,
+      "learning_rate": 0.0005229055944055943,
+      "loss": 3.4796,
+      "step": 22150
+    },
+    {
+      "epoch": 6.4647911934300195,
+      "grad_norm": 0.33600088953971863,
+      "learning_rate": 0.0005227307692307691,
+      "loss": 3.4894,
+      "step": 22200
+    },
+    {
+      "epoch": 6.479352321043741,
+      "grad_norm": 0.3299783766269684,
+      "learning_rate": 0.0005225559440559441,
+      "loss": 3.4861,
+      "step": 22250
+    },
+    {
+      "epoch": 6.493913448657464,
+      "grad_norm": 0.3157522976398468,
+      "learning_rate": 0.0005223811188811189,
+      "loss": 3.4758,
+      "step": 22300
+    },
+    {
+      "epoch": 6.508474576271187,
+      "grad_norm": 0.3057360053062439,
+      "learning_rate": 0.0005222062937062937,
+      "loss": 3.4781,
+      "step": 22350
+    },
+    {
+      "epoch": 6.523035703884909,
+      "grad_norm": 0.32586047053337097,
+      "learning_rate": 0.0005220314685314686,
+      "loss": 3.4823,
+      "step": 22400
+    },
+    {
+      "epoch": 6.5375968314986315,
+      "grad_norm": 0.3262418210506439,
+      "learning_rate": 0.0005218566433566433,
+      "loss": 3.4934,
+      "step": 22450
+    },
+    {
+      "epoch": 6.552157959112353,
+      "grad_norm": 0.33487650752067566,
+      "learning_rate": 0.0005216818181818182,
+      "loss": 3.4761,
+      "step": 22500
+    },
+    {
+      "epoch": 6.566719086726076,
+      "grad_norm": 0.342220664024353,
+      "learning_rate": 0.000521506993006993,
+      "loss": 3.4971,
+      "step": 22550
+    },
+    {
+      "epoch": 6.581280214339799,
+      "grad_norm": 0.32479947805404663,
+      "learning_rate": 0.0005213321678321678,
+      "loss": 3.492,
+      "step": 22600
+    },
+    {
+      "epoch": 6.595841341953521,
+      "grad_norm": 0.3203052580356598,
+      "learning_rate": 0.0005211573426573426,
+      "loss": 3.4777,
+      "step": 22650
+    },
+    {
+      "epoch": 6.610402469567243,
+      "grad_norm": 0.3330882787704468,
+      "learning_rate": 0.0005209825174825175,
+      "loss": 3.4797,
+      "step": 22700
+    },
+    {
+      "epoch": 6.624963597180965,
+      "grad_norm": 0.3708348572254181,
+      "learning_rate": 0.0005208076923076923,
+      "loss": 3.4816,
+      "step": 22750
+    },
+    {
+      "epoch": 6.639524724794688,
+      "grad_norm": 0.3509019613265991,
+      "learning_rate": 0.0005206328671328671,
+      "loss": 3.4941,
+      "step": 22800
+    },
+    {
+      "epoch": 6.654085852408411,
+      "grad_norm": 0.3052864968776703,
+      "learning_rate": 0.0005204580419580419,
+      "loss": 3.4806,
+      "step": 22850
+    },
+    {
+      "epoch": 6.668646980022133,
+      "grad_norm": 0.309030681848526,
+      "learning_rate": 0.0005202832167832168,
+      "loss": 3.4757,
+      "step": 22900
+    },
+    {
+      "epoch": 6.683208107635855,
+      "grad_norm": 0.3289431035518646,
+      "learning_rate": 0.0005201083916083916,
+      "loss": 3.4787,
+      "step": 22950
+    },
+    {
+      "epoch": 6.697769235249577,
+      "grad_norm": 0.3410831391811371,
+      "learning_rate": 0.0005199335664335664,
+      "loss": 3.4856,
+      "step": 23000
+    },
+    {
+      "epoch": 6.697769235249577,
+      "eval_accuracy": 0.3638117136090239,
+      "eval_loss": 3.60054349899292,
+      "eval_runtime": 179.6205,
+      "eval_samples_per_second": 92.673,
+      "eval_steps_per_second": 5.796,
+      "step": 23000
+    },
+    {
+      "epoch": 6.7123303628633,
+      "grad_norm": 0.3080759346485138,
+      "learning_rate": 0.0005197587412587413,
+      "loss": 3.4842,
+      "step": 23050
+    },
+    {
+      "epoch": 6.726891490477023,
+      "grad_norm": 0.3141343295574188,
+      "learning_rate": 0.0005195839160839161,
+      "loss": 3.4951,
+      "step": 23100
+    },
+    {
+      "epoch": 6.741452618090745,
+      "grad_norm": 0.332527220249176,
+      "learning_rate": 0.0005194090909090909,
+      "loss": 3.4855,
+      "step": 23150
+    },
+    {
+      "epoch": 6.756013745704467,
+      "grad_norm": 0.32234665751457214,
+      "learning_rate": 0.0005192342657342657,
+      "loss": 3.4815,
+      "step": 23200
+    },
+    {
+      "epoch": 6.77057487331819,
+      "grad_norm": 0.33322620391845703,
+      "learning_rate": 0.0005190594405594405,
+      "loss": 3.4962,
+      "step": 23250
+    },
+    {
+      "epoch": 6.785136000931912,
+      "grad_norm": 0.337287575006485,
+      "learning_rate": 0.0005188846153846153,
+      "loss": 3.4865,
+      "step": 23300
+    },
+    {
+      "epoch": 6.799697128545635,
+      "grad_norm": 0.3224928379058838,
+      "learning_rate": 0.0005187097902097902,
+      "loss": 3.4865,
+      "step": 23350
+    },
+    {
+      "epoch": 6.814258256159357,
+      "grad_norm": 0.32035595178604126,
+      "learning_rate": 0.000518534965034965,
+      "loss": 3.4969,
+      "step": 23400
+    },
+    {
+      "epoch": 6.828819383773079,
+      "grad_norm": 0.343305379152298,
+      "learning_rate": 0.0005183601398601398,
+      "loss": 3.4836,
+      "step": 23450
+    },
+    {
+      "epoch": 6.843380511386802,
+      "grad_norm": 0.3211219012737274,
+      "learning_rate": 0.0005181853146853146,
+      "loss": 3.4926,
+      "step": 23500
+    },
+    {
+      "epoch": 6.857941639000524,
+      "grad_norm": 0.31823667883872986,
+      "learning_rate": 0.0005180104895104895,
+      "loss": 3.49,
+      "step": 23550
+    },
+    {
+      "epoch": 6.872502766614247,
+      "grad_norm": 0.31736093759536743,
+      "learning_rate": 0.0005178356643356643,
+      "loss": 3.497,
+      "step": 23600
+    },
+    {
+      "epoch": 6.887063894227969,
+      "grad_norm": 0.3398060202598572,
+      "learning_rate": 0.0005176608391608391,
+      "loss": 3.4832,
+      "step": 23650
+    },
+    {
+      "epoch": 6.901625021841691,
+      "grad_norm": 0.37513840198516846,
+      "learning_rate": 0.000517486013986014,
+      "loss": 3.4865,
+      "step": 23700
+    },
+    {
+      "epoch": 6.916186149455414,
+      "grad_norm": 0.3081456422805786,
+      "learning_rate": 0.0005173111888111888,
+      "loss": 3.4894,
+      "step": 23750
+    },
+    {
+      "epoch": 6.930747277069136,
+      "grad_norm": 0.31910234689712524,
+      "learning_rate": 0.0005171363636363636,
+      "loss": 3.4849,
+      "step": 23800
+    },
+    {
+      "epoch": 6.945308404682859,
+      "grad_norm": 0.3732234537601471,
+      "learning_rate": 0.0005169615384615384,
+      "loss": 3.4812,
+      "step": 23850
+    },
+    {
+      "epoch": 6.959869532296581,
+      "grad_norm": 0.31321918964385986,
+      "learning_rate": 0.0005167867132867133,
+      "loss": 3.4905,
+      "step": 23900
+    },
+    {
+      "epoch": 6.974430659910303,
+      "grad_norm": 0.3159354627132416,
+      "learning_rate": 0.000516611888111888,
+      "loss": 3.4908,
+      "step": 23950
+    },
+    {
+      "epoch": 6.988991787524026,
+      "grad_norm": 0.324131041765213,
+      "learning_rate": 0.0005164370629370629,
+      "loss": 3.499,
+      "step": 24000
+    },
+    {
+      "epoch": 6.988991787524026,
+      "eval_accuracy": 0.3645026290525888,
+      "eval_loss": 3.5911543369293213,
+      "eval_runtime": 179.6242,
+      "eval_samples_per_second": 92.671,
+      "eval_steps_per_second": 5.795,
+      "step": 24000
+    },
+    {
+      "epoch": 7.003494670627293,
+      "grad_norm": 0.3455987274646759,
+      "learning_rate": 0.0005162622377622377,
+      "loss": 3.468,
+      "step": 24050
+    },
+    {
+      "epoch": 7.018055798241016,
+      "grad_norm": 0.32421788573265076,
+      "learning_rate": 0.0005160874125874125,
+      "loss": 3.3858,
+      "step": 24100
+    },
+    {
+      "epoch": 7.032616925854738,
+      "grad_norm": 0.3512269854545593,
+      "learning_rate": 0.0005159125874125873,
+      "loss": 3.3879,
+      "step": 24150
+    },
+    {
+      "epoch": 7.0471780534684605,
+      "grad_norm": 0.3255019783973694,
+      "learning_rate": 0.0005157377622377622,
+      "loss": 3.3864,
+      "step": 24200
+    },
+    {
+      "epoch": 7.061739181082183,
+      "grad_norm": 0.3749047517776489,
+      "learning_rate": 0.000515562937062937,
+      "loss": 3.3933,
+      "step": 24250
+    },
+    {
+      "epoch": 7.076300308695905,
+      "grad_norm": 0.33762863278388977,
+      "learning_rate": 0.0005153881118881118,
+      "loss": 3.406,
+      "step": 24300
+    },
+    {
+      "epoch": 7.090861436309628,
+      "grad_norm": 0.34979817271232605,
+      "learning_rate": 0.0005152132867132867,
+      "loss": 3.4044,
+      "step": 24350
+    },
+    {
+      "epoch": 7.105422563923351,
+      "grad_norm": 0.3524860739707947,
+      "learning_rate": 0.0005150384615384615,
+      "loss": 3.3998,
+      "step": 24400
+    },
+    {
+      "epoch": 7.1199836915370724,
+      "grad_norm": 0.3533429503440857,
+      "learning_rate": 0.0005148636363636363,
+      "loss": 3.3967,
+      "step": 24450
+    },
+    {
+      "epoch": 7.134544819150795,
+      "grad_norm": 0.33228060603141785,
+      "learning_rate": 0.0005146888111888111,
+      "loss": 3.4,
+      "step": 24500
+    },
+    {
+      "epoch": 7.149105946764517,
+      "grad_norm": 0.33005383610725403,
+      "learning_rate": 0.000514513986013986,
+      "loss": 3.4111,
+      "step": 24550
+    },
+    {
+      "epoch": 7.16366707437824,
+      "grad_norm": 0.32785147428512573,
+      "learning_rate": 0.0005143391608391608,
+      "loss": 3.4146,
+      "step": 24600
+    },
+    {
+      "epoch": 7.1782282019919625,
+      "grad_norm": 0.34615424275398254,
+      "learning_rate": 0.0005141643356643356,
+      "loss": 3.4083,
+      "step": 24650
+    },
+    {
+      "epoch": 7.192789329605684,
+      "grad_norm": 0.31757858395576477,
+      "learning_rate": 0.0005139895104895104,
+      "loss": 3.4192,
+      "step": 24700
+    },
+    {
+      "epoch": 7.207350457219407,
+      "grad_norm": 0.3305107355117798,
+      "learning_rate": 0.0005138146853146852,
+      "loss": 3.4174,
+      "step": 24750
+    },
+    {
+      "epoch": 7.22191158483313,
+      "grad_norm": 0.3234144151210785,
+      "learning_rate": 0.00051363986013986,
+      "loss": 3.4124,
+      "step": 24800
+    },
+    {
+      "epoch": 7.236472712446852,
+      "grad_norm": 0.3411427438259125,
+      "learning_rate": 0.0005134650349650349,
+      "loss": 3.4177,
+      "step": 24850
+    },
+    {
+      "epoch": 7.2510338400605745,
+      "grad_norm": 0.3230850398540497,
+      "learning_rate": 0.0005132902097902097,
+      "loss": 3.4201,
+      "step": 24900
+    },
+    {
+      "epoch": 7.265594967674296,
+      "grad_norm": 0.3260183036327362,
+      "learning_rate": 0.0005131153846153845,
+      "loss": 3.4257,
+      "step": 24950
+    },
+    {
+      "epoch": 7.280156095288019,
+      "grad_norm": 0.3340539038181305,
+      "learning_rate": 0.0005129405594405594,
+      "loss": 3.4248,
+      "step": 25000
+    },
+    {
+      "epoch": 7.280156095288019,
+      "eval_accuracy": 0.3648761254562311,
+      "eval_loss": 3.597316026687622,
+      "eval_runtime": 179.6106,
+      "eval_samples_per_second": 92.678,
+      "eval_steps_per_second": 5.796,
+      "step": 25000
+    },
+    {
+      "epoch": 7.294717222901742,
+      "grad_norm": 0.3235202133655548,
+      "learning_rate": 0.0005127657342657342,
+      "loss": 3.4151,
+      "step": 25050
+    },
+    {
+      "epoch": 7.309278350515464,
+      "grad_norm": 0.3350401818752289,
+      "learning_rate": 0.000512590909090909,
+      "loss": 3.4264,
+      "step": 25100
+    },
+    {
+      "epoch": 7.3238394781291865,
+      "grad_norm": 0.3224312663078308,
+      "learning_rate": 0.0005124160839160838,
+      "loss": 3.4351,
+      "step": 25150
+    },
+    {
+      "epoch": 7.338400605742908,
+      "grad_norm": 0.34827619791030884,
+      "learning_rate": 0.0005122412587412588,
+      "loss": 3.419,
+      "step": 25200
+    },
+    {
+      "epoch": 7.352961733356631,
+      "grad_norm": 0.3302023708820343,
+      "learning_rate": 0.0005120664335664336,
+      "loss": 3.4288,
+      "step": 25250
+    },
+    {
+      "epoch": 7.367522860970354,
+      "grad_norm": 0.32455217838287354,
+      "learning_rate": 0.0005118916083916084,
+      "loss": 3.4304,
+      "step": 25300
+    },
+    {
+      "epoch": 7.382083988584076,
+      "grad_norm": 0.3269766867160797,
+      "learning_rate": 0.0005117167832167832,
+      "loss": 3.4261,
+      "step": 25350
+    },
+    {
+      "epoch": 7.396645116197798,
+      "grad_norm": 0.33805644512176514,
+      "learning_rate": 0.0005115419580419581,
+      "loss": 3.4302,
+      "step": 25400
+    },
+    {
+      "epoch": 7.411206243811521,
+      "grad_norm": 0.30504998564720154,
+      "learning_rate": 0.0005113671328671328,
+      "loss": 3.4436,
+      "step": 25450
+    },
+    {
+      "epoch": 7.425767371425243,
+      "grad_norm": 0.3210298717021942,
+      "learning_rate": 0.0005111923076923077,
+      "loss": 3.4379,
+      "step": 25500
+    },
+    {
+      "epoch": 7.440328499038966,
+      "grad_norm": 0.3328574597835541,
+      "learning_rate": 0.0005110174825174825,
+      "loss": 3.4364,
+      "step": 25550
+    },
+    {
+      "epoch": 7.454889626652688,
+      "grad_norm": 0.3316180408000946,
+      "learning_rate": 0.0005108426573426573,
+      "loss": 3.4421,
+      "step": 25600
+    },
+    {
+      "epoch": 7.46945075426641,
+      "grad_norm": 0.32503074407577515,
+      "learning_rate": 0.0005106678321678321,
+      "loss": 3.4373,
+      "step": 25650
+    },
+    {
+      "epoch": 7.484011881880133,
+      "grad_norm": 0.34754231572151184,
+      "learning_rate": 0.000510493006993007,
+      "loss": 3.4424,
+      "step": 25700
+    },
+    {
+      "epoch": 7.498573009493855,
+      "grad_norm": 0.3636908531188965,
+      "learning_rate": 0.0005103181818181818,
+      "loss": 3.4421,
+      "step": 25750
+    },
+    {
+      "epoch": 7.513134137107578,
+      "grad_norm": 0.32700809836387634,
+      "learning_rate": 0.0005101433566433566,
+      "loss": 3.4452,
+      "step": 25800
+    },
+    {
+      "epoch": 7.5276952647213005,
+      "grad_norm": 0.34218472242355347,
+      "learning_rate": 0.0005099685314685315,
+      "loss": 3.4455,
+      "step": 25850
+    },
+    {
+      "epoch": 7.542256392335022,
+      "grad_norm": 0.3296022117137909,
+      "learning_rate": 0.0005097937062937063,
+      "loss": 3.4512,
+      "step": 25900
+    },
+    {
+      "epoch": 7.556817519948745,
+      "grad_norm": 0.3350162208080292,
+      "learning_rate": 0.0005096188811188811,
+      "loss": 3.4474,
+      "step": 25950
+    },
+    {
+      "epoch": 7.571378647562467,
+      "grad_norm": 0.30475959181785583,
+      "learning_rate": 0.0005094440559440559,
+      "loss": 3.452,
+      "step": 26000
+    },
+    {
+      "epoch": 7.571378647562467,
+      "eval_accuracy": 0.36505987581156407,
+      "eval_loss": 3.592822790145874,
+      "eval_runtime": 179.6236,
+      "eval_samples_per_second": 92.672,
+      "eval_steps_per_second": 5.795,
+      "step": 26000
+    },
+    {
+      "epoch": 7.58593977517619,
+      "grad_norm": 0.34838974475860596,
+      "learning_rate": 0.0005092692307692308,
+      "loss": 3.4448,
+      "step": 26050
+    },
+    {
+      "epoch": 7.600500902789912,
+      "grad_norm": 0.3396695852279663,
+      "learning_rate": 0.0005090944055944056,
+      "loss": 3.4505,
+      "step": 26100
+    },
+    {
+      "epoch": 7.615062030403634,
+      "grad_norm": 0.3172091543674469,
+      "learning_rate": 0.0005089195804195804,
+      "loss": 3.4539,
+      "step": 26150
+    },
+    {
+      "epoch": 7.629623158017357,
+      "grad_norm": 0.3365660607814789,
+      "learning_rate": 0.0005087447552447552,
+      "loss": 3.4448,
+      "step": 26200
+    },
+    {
+      "epoch": 7.644184285631079,
+      "grad_norm": 0.3183286488056183,
+      "learning_rate": 0.00050856993006993,
+      "loss": 3.4447,
+      "step": 26250
+    },
+    {
+      "epoch": 7.658745413244802,
+      "grad_norm": 0.32615023851394653,
+      "learning_rate": 0.0005083951048951048,
+      "loss": 3.4541,
+      "step": 26300
+    },
+    {
+      "epoch": 7.673306540858524,
+      "grad_norm": 0.33578434586524963,
+      "learning_rate": 0.0005082202797202797,
+      "loss": 3.4546,
+      "step": 26350
+    },
+    {
+      "epoch": 7.687867668472246,
+      "grad_norm": 0.3528318405151367,
+      "learning_rate": 0.0005080454545454545,
+      "loss": 3.4597,
+      "step": 26400
+    },
+    {
+      "epoch": 7.702428796085969,
+      "grad_norm": 0.33206814527511597,
+      "learning_rate": 0.0005078706293706293,
+      "loss": 3.4576,
+      "step": 26450
+    },
+    {
+      "epoch": 7.716989923699691,
+      "grad_norm": 0.35905346274375916,
+      "learning_rate": 0.0005076958041958042,
+      "loss": 3.4657,
+      "step": 26500
+    },
+    {
+      "epoch": 7.731551051313414,
+      "grad_norm": 0.32397958636283875,
+      "learning_rate": 0.000507520979020979,
+      "loss": 3.4592,
+      "step": 26550
+    },
+    {
+      "epoch": 7.746112178927136,
+      "grad_norm": 0.3245496451854706,
+      "learning_rate": 0.0005073461538461538,
+      "loss": 3.4444,
+      "step": 26600
+    },
+    {
+      "epoch": 7.760673306540858,
+      "grad_norm": 0.3393577039241791,
+      "learning_rate": 0.0005071713286713286,
+      "loss": 3.455,
+      "step": 26650
+    },
+    {
+      "epoch": 7.775234434154581,
+      "grad_norm": 0.3460133373737335,
+      "learning_rate": 0.0005069965034965035,
+      "loss": 3.4519,
+      "step": 26700
+    },
+    {
+      "epoch": 7.789795561768304,
+      "grad_norm": 0.3299025893211365,
+      "learning_rate": 0.0005068216783216783,
+      "loss": 3.4521,
+      "step": 26750
+    },
+    {
+      "epoch": 7.8043566893820255,
+      "grad_norm": 0.32487180829048157,
+      "learning_rate": 0.0005066468531468531,
+      "loss": 3.4672,
+      "step": 26800
+    },
+    {
+      "epoch": 7.818917816995748,
+      "grad_norm": 0.3425562381744385,
+      "learning_rate": 0.0005064720279720279,
+      "loss": 3.4408,
+      "step": 26850
+    },
+    {
+      "epoch": 7.833478944609471,
+      "grad_norm": 0.2968789339065552,
+      "learning_rate": 0.0005062972027972028,
+      "loss": 3.4553,
+      "step": 26900
+    },
+    {
+      "epoch": 7.848040072223193,
+      "grad_norm": 0.35579219460487366,
+      "learning_rate": 0.0005061223776223775,
+      "loss": 3.4581,
+      "step": 26950
+    },
+    {
+      "epoch": 7.862601199836916,
+      "grad_norm": 0.32660984992980957,
+      "learning_rate": 0.0005059475524475524,
+      "loss": 3.4628,
+      "step": 27000
+    },
+    {
+      "epoch": 7.862601199836916,
+      "eval_accuracy": 0.36588081549888984,
+      "eval_loss": 3.5821452140808105,
+      "eval_runtime": 179.7196,
+      "eval_samples_per_second": 92.622,
+      "eval_steps_per_second": 5.792,
+      "step": 27000
+    },
+    {
+      "epoch": 7.8771623274506375,
+      "grad_norm": 0.34840381145477295,
+      "learning_rate": 0.0005057727272727272,
+      "loss": 3.4578,
+      "step": 27050
+    },
+    {
+      "epoch": 7.89172345506436,
+      "grad_norm": 0.3340885043144226,
+      "learning_rate": 0.000505597902097902,
+      "loss": 3.4542,
+      "step": 27100
+    },
+    {
+      "epoch": 7.906284582678083,
+      "grad_norm": 0.3486022651195526,
+      "learning_rate": 0.0005054230769230769,
+      "loss": 3.4558,
+      "step": 27150
+    },
+    {
+      "epoch": 7.920845710291805,
+      "grad_norm": 0.342074453830719,
+      "learning_rate": 0.0005052482517482517,
+      "loss": 3.4689,
+      "step": 27200
+    },
+    {
+      "epoch": 7.935406837905528,
+      "grad_norm": 0.32947641611099243,
+      "learning_rate": 0.0005050734265734265,
+      "loss": 3.4728,
+      "step": 27250
+    },
+    {
+      "epoch": 7.9499679655192494,
+      "grad_norm": 0.31653088331222534,
+      "learning_rate": 0.0005048986013986013,
+      "loss": 3.4685,
+      "step": 27300
+    },
+    {
+      "epoch": 7.964529093132972,
+      "grad_norm": 0.34901005029678345,
+      "learning_rate": 0.0005047237762237762,
+      "loss": 3.4647,
+      "step": 27350
+    },
+    {
+      "epoch": 7.979090220746695,
+      "grad_norm": 0.3230254352092743,
+      "learning_rate": 0.000504548951048951,
+      "loss": 3.4553,
+      "step": 27400
+    },
+    {
+      "epoch": 7.993651348360417,
+      "grad_norm": 0.3505324125289917,
+      "learning_rate": 0.0005043741258741258,
+      "loss": 3.4651,
+      "step": 27450
+    },
+    {
+      "epoch": 8.008154231463685,
+      "grad_norm": 0.33360299468040466,
+      "learning_rate": 0.0005041993006993006,
+      "loss": 3.4055,
+      "step": 27500
+    },
+    {
+      "epoch": 8.022715359077408,
+      "grad_norm": 0.34768131375312805,
+      "learning_rate": 0.0005040244755244755,
+      "loss": 3.3498,
+      "step": 27550
+    },
+    {
+      "epoch": 8.037276486691129,
+      "grad_norm": 0.3448757827281952,
+      "learning_rate": 0.0005038496503496503,
+      "loss": 3.3425,
+      "step": 27600
+    },
+    {
+      "epoch": 8.051837614304851,
+      "grad_norm": 0.3235391676425934,
+      "learning_rate": 0.0005036748251748251,
+      "loss": 3.3552,
+      "step": 27650
+    },
+    {
+      "epoch": 8.066398741918574,
+      "grad_norm": 0.3475419580936432,
+      "learning_rate": 0.0005034999999999999,
+      "loss": 3.3667,
+      "step": 27700
+    },
+    {
+      "epoch": 8.080959869532297,
+      "grad_norm": 0.3373063802719116,
+      "learning_rate": 0.0005033251748251747,
+      "loss": 3.3604,
+      "step": 27750
+    },
+    {
+      "epoch": 8.09552099714602,
+      "grad_norm": 0.33979612588882446,
+      "learning_rate": 0.0005031503496503496,
+      "loss": 3.3733,
+      "step": 27800
+    },
+    {
+      "epoch": 8.11008212475974,
+      "grad_norm": 0.3554824888706207,
+      "learning_rate": 0.0005029755244755244,
+      "loss": 3.3742,
+      "step": 27850
+    },
+    {
+      "epoch": 8.124643252373463,
+      "grad_norm": 0.3337261378765106,
+      "learning_rate": 0.0005028006993006992,
+      "loss": 3.3687,
+      "step": 27900
+    },
+    {
+      "epoch": 8.139204379987186,
+      "grad_norm": 0.3323470652103424,
+      "learning_rate": 0.000502625874125874,
+      "loss": 3.3727,
+      "step": 27950
+    },
+    {
+      "epoch": 8.153765507600909,
+      "grad_norm": 0.3721840977668762,
+      "learning_rate": 0.000502451048951049,
+      "loss": 3.3875,
+      "step": 28000
+    },
+    {
+      "epoch": 8.153765507600909,
+      "eval_accuracy": 0.36575702207332006,
+      "eval_loss": 3.587688684463501,
+      "eval_runtime": 179.6153,
+      "eval_samples_per_second": 92.676,
+      "eval_steps_per_second": 5.796,
+      "step": 28000
+    },
+    {
+      "epoch": 8.168326635214632,
+      "grad_norm": 0.32385772466659546,
+      "learning_rate": 0.0005022762237762237,
+      "loss": 3.377,
+      "step": 28050
+    },
+    {
+      "epoch": 8.182887762828354,
+      "grad_norm": 0.34891006350517273,
+      "learning_rate": 0.0005021013986013985,
+      "loss": 3.3799,
+      "step": 28100
+    },
+    {
+      "epoch": 8.197448890442075,
+      "grad_norm": 0.34678199887275696,
+      "learning_rate": 0.0005019265734265733,
+      "loss": 3.3868,
+      "step": 28150
+    },
+    {
+      "epoch": 8.212010018055798,
+      "grad_norm": 0.3331698477268219,
+      "learning_rate": 0.0005017517482517483,
+      "loss": 3.3834,
+      "step": 28200
+    },
+    {
+      "epoch": 8.22657114566952,
+      "grad_norm": 0.3234923779964447,
+      "learning_rate": 0.0005015769230769231,
+      "loss": 3.3933,
+      "step": 28250
+    },
+    {
+      "epoch": 8.241132273283243,
+      "grad_norm": 0.33203065395355225,
+      "learning_rate": 0.0005014020979020979,
+      "loss": 3.3947,
+      "step": 28300
+    },
+    {
+      "epoch": 8.255693400896966,
+      "grad_norm": 0.3614442050457001,
+      "learning_rate": 0.0005012272727272727,
+      "loss": 3.396,
+      "step": 28350
+    },
+    {
+      "epoch": 8.270254528510687,
+      "grad_norm": 0.3333202302455902,
+      "learning_rate": 0.0005010524475524476,
+      "loss": 3.3944,
+      "step": 28400
+    },
+    {
+      "epoch": 8.28481565612441,
+      "grad_norm": 0.36438530683517456,
+      "learning_rate": 0.0005008776223776223,
+      "loss": 3.3969,
+      "step": 28450
+    },
+    {
+      "epoch": 8.299376783738133,
+      "grad_norm": 0.3521031439304352,
+      "learning_rate": 0.0005007027972027972,
+      "loss": 3.4055,
+      "step": 28500
+    },
+    {
+      "epoch": 8.313937911351855,
+      "grad_norm": 0.3472055494785309,
+      "learning_rate": 0.000500527972027972,
+      "loss": 3.4046,
+      "step": 28550
+    },
+    {
+      "epoch": 8.328499038965578,
+      "grad_norm": 0.3367118239402771,
+      "learning_rate": 0.0005003531468531468,
+      "loss": 3.3965,
+      "step": 28600
+    },
+    {
+      "epoch": 8.3430601665793,
+      "grad_norm": 0.32049131393432617,
+      "learning_rate": 0.0005001783216783217,
+      "loss": 3.4007,
+      "step": 28650
+    },
+    {
+      "epoch": 8.357621294193022,
+      "grad_norm": 0.36562150716781616,
+      "learning_rate": 0.0005000034965034965,
+      "loss": 3.398,
+      "step": 28700
+    },
+    {
+      "epoch": 8.372182421806745,
+      "grad_norm": 0.3249284327030182,
+      "learning_rate": 0.0004998286713286713,
+      "loss": 3.3956,
+      "step": 28750
+    },
+    {
+      "epoch": 8.386743549420467,
+      "grad_norm": 0.34902429580688477,
+      "learning_rate": 0.0004996538461538461,
+      "loss": 3.4019,
+      "step": 28800
+    },
+    {
+      "epoch": 8.40130467703419,
+      "grad_norm": 0.35250189900398254,
+      "learning_rate": 0.000499479020979021,
+      "loss": 3.396,
+      "step": 28850
+    },
+    {
+      "epoch": 8.415865804647911,
+      "grad_norm": 0.3263583481311798,
+      "learning_rate": 0.0004993041958041958,
+      "loss": 3.4019,
+      "step": 28900
+    },
+    {
+      "epoch": 8.430426932261634,
+      "grad_norm": 0.3295762836933136,
+      "learning_rate": 0.0004991293706293706,
+      "loss": 3.4142,
+      "step": 28950
+    },
+    {
+      "epoch": 8.444988059875357,
+      "grad_norm": 0.3347003161907196,
+      "learning_rate": 0.0004989545454545454,
+      "loss": 3.4067,
+      "step": 29000
+    },
+    {
+      "epoch": 8.444988059875357,
+      "eval_accuracy": 0.36621951337074804,
+      "eval_loss": 3.5861732959747314,
+      "eval_runtime": 179.638,
+      "eval_samples_per_second": 92.664,
+      "eval_steps_per_second": 5.795,
+      "step": 29000
+    },
+    {
+      "epoch": 8.45954918748908,
+      "grad_norm": 0.3246140778064728,
+      "learning_rate": 0.0004987797202797203,
+      "loss": 3.4132,
+      "step": 29050
+    },
+    {
+      "epoch": 8.474110315102802,
+      "grad_norm": 0.3662111163139343,
+      "learning_rate": 0.0004986048951048951,
+      "loss": 3.4005,
+      "step": 29100
+    },
+    {
+      "epoch": 8.488671442716523,
+      "grad_norm": 0.3711738884449005,
+      "learning_rate": 0.0004984300699300699,
+      "loss": 3.4136,
+      "step": 29150
+    },
+    {
+      "epoch": 8.503232570330246,
+      "grad_norm": 0.3224983513355255,
+      "learning_rate": 0.0004982552447552448,
+      "loss": 3.4193,
+      "step": 29200
+    },
+    {
+      "epoch": 8.517793697943969,
+      "grad_norm": 0.34007972478866577,
+      "learning_rate": 0.0004980804195804195,
+      "loss": 3.4033,
+      "step": 29250
+    },
+    {
+      "epoch": 8.532354825557691,
+      "grad_norm": 0.34795621037483215,
+      "learning_rate": 0.0004979055944055944,
+      "loss": 3.4279,
+      "step": 29300
+    },
+    {
+      "epoch": 8.546915953171414,
+      "grad_norm": 0.3562271296977997,
+      "learning_rate": 0.0004977307692307692,
+      "loss": 3.4191,
+      "step": 29350
+    },
+    {
+      "epoch": 8.561477080785137,
+      "grad_norm": 0.33655011653900146,
+      "learning_rate": 0.000497555944055944,
+      "loss": 3.4124,
+      "step": 29400
+    },
+    {
+      "epoch": 8.576038208398858,
+      "grad_norm": 0.3349234461784363,
+      "learning_rate": 0.0004973811188811188,
+      "loss": 3.4197,
+      "step": 29450
+    },
+    {
+      "epoch": 8.59059933601258,
+      "grad_norm": 0.3217172622680664,
+      "learning_rate": 0.0004972062937062937,
+      "loss": 3.4254,
+      "step": 29500
+    },
+    {
+      "epoch": 8.605160463626303,
+      "grad_norm": 0.33423301577568054,
+      "learning_rate": 0.0004970314685314685,
+      "loss": 3.4127,
+      "step": 29550
+    },
+    {
+      "epoch": 8.619721591240026,
+      "grad_norm": 0.34116896986961365,
+      "learning_rate": 0.0004968566433566433,
+      "loss": 3.4258,
+      "step": 29600
+    },
+    {
+      "epoch": 8.634282718853749,
+      "grad_norm": 0.3287065029144287,
+      "learning_rate": 0.0004966818181818181,
+      "loss": 3.4198,
+      "step": 29650
+    },
+    {
+      "epoch": 8.64884384646747,
+      "grad_norm": 0.35336413979530334,
+      "learning_rate": 0.000496506993006993,
+      "loss": 3.4168,
+      "step": 29700
+    },
+    {
+      "epoch": 8.663404974081192,
+      "grad_norm": 0.34202465415000916,
+      "learning_rate": 0.0004963321678321678,
+      "loss": 3.4277,
+      "step": 29750
+    },
+    {
+      "epoch": 8.677966101694915,
+      "grad_norm": 0.3359808921813965,
+      "learning_rate": 0.0004961573426573426,
+      "loss": 3.428,
+      "step": 29800
+    },
+    {
+      "epoch": 8.692527229308638,
+      "grad_norm": 0.33416661620140076,
+      "learning_rate": 0.0004959825174825175,
+      "loss": 3.4254,
+      "step": 29850
+    },
+    {
+      "epoch": 8.70708835692236,
+      "grad_norm": 0.34198299050331116,
+      "learning_rate": 0.0004958076923076923,
+      "loss": 3.4263,
+      "step": 29900
+    },
+    {
+      "epoch": 8.721649484536082,
+      "grad_norm": 0.359495609998703,
+      "learning_rate": 0.0004956328671328671,
+      "loss": 3.4389,
+      "step": 29950
+    },
+    {
+      "epoch": 8.736210612149804,
+      "grad_norm": 0.3388700485229492,
+      "learning_rate": 0.0004954580419580419,
+      "loss": 3.4324,
+      "step": 30000
+    },
+    {
+      "epoch": 8.736210612149804,
+      "eval_accuracy": 0.3668087371589303,
+      "eval_loss": 3.5739517211914062,
+      "eval_runtime": 179.5999,
+      "eval_samples_per_second": 92.684,
+      "eval_steps_per_second": 5.796,
+      "step": 30000
+    },
+    {
+      "epoch": 8.750771739763527,
+      "grad_norm": 0.3278059959411621,
+      "learning_rate": 0.0004952832167832167,
+      "loss": 3.4248,
+      "step": 30050
+    },
+    {
+      "epoch": 8.76533286737725,
+      "grad_norm": 0.34186989068984985,
+      "learning_rate": 0.0004951083916083915,
+      "loss": 3.4251,
+      "step": 30100
+    },
+    {
+      "epoch": 8.779893994990973,
+      "grad_norm": 0.3263101577758789,
+      "learning_rate": 0.0004949335664335664,
+      "loss": 3.4302,
+      "step": 30150
+    },
+    {
+      "epoch": 8.794455122604695,
+      "grad_norm": 0.354638010263443,
+      "learning_rate": 0.0004947587412587412,
+      "loss": 3.416,
+      "step": 30200
+    },
+    {
+      "epoch": 8.809016250218416,
+      "grad_norm": 0.3320951759815216,
+      "learning_rate": 0.000494583916083916,
+      "loss": 3.4236,
+      "step": 30250
+    },
+    {
+      "epoch": 8.82357737783214,
+      "grad_norm": 0.3258265554904938,
+      "learning_rate": 0.0004944090909090908,
+      "loss": 3.4288,
+      "step": 30300
+    },
+    {
+      "epoch": 8.838138505445862,
+      "grad_norm": 0.3345946967601776,
+      "learning_rate": 0.0004942342657342657,
+      "loss": 3.4451,
+      "step": 30350
+    },
+    {
+      "epoch": 8.852699633059585,
+      "grad_norm": 0.32482603192329407,
+      "learning_rate": 0.0004940594405594405,
+      "loss": 3.4356,
+      "step": 30400
+    },
+    {
+      "epoch": 8.867260760673307,
+      "grad_norm": 0.3448339104652405,
+      "learning_rate": 0.0004938846153846153,
+      "loss": 3.4376,
+      "step": 30450
+    },
+    {
+      "epoch": 8.881821888287028,
+      "grad_norm": 0.33671805262565613,
+      "learning_rate": 0.0004937097902097901,
+      "loss": 3.4342,
+      "step": 30500
+    },
+    {
+      "epoch": 8.896383015900751,
+      "grad_norm": 0.33519789576530457,
+      "learning_rate": 0.000493534965034965,
+      "loss": 3.4395,
+      "step": 30550
+    },
+    {
+      "epoch": 8.910944143514474,
+      "grad_norm": 0.3589082658290863,
+      "learning_rate": 0.0004933601398601398,
+      "loss": 3.4191,
+      "step": 30600
+    },
+    {
+      "epoch": 8.925505271128197,
+      "grad_norm": 0.3367752730846405,
+      "learning_rate": 0.0004931853146853146,
+      "loss": 3.4361,
+      "step": 30650
+    },
+    {
+      "epoch": 8.94006639874192,
+      "grad_norm": 0.31563615798950195,
+      "learning_rate": 0.0004930104895104895,
+      "loss": 3.4309,
+      "step": 30700
+    },
+    {
+      "epoch": 8.95462752635564,
+      "grad_norm": 0.32066795229911804,
+      "learning_rate": 0.0004928356643356642,
+      "loss": 3.4346,
+      "step": 30750
+    },
+    {
+      "epoch": 8.969188653969363,
+      "grad_norm": 0.33618220686912537,
+      "learning_rate": 0.0004926608391608391,
+      "loss": 3.4448,
+      "step": 30800
+    },
+    {
+      "epoch": 8.983749781583086,
+      "grad_norm": 0.3660648465156555,
+      "learning_rate": 0.0004924860139860139,
+      "loss": 3.4283,
+      "step": 30850
+    },
+    {
+      "epoch": 8.998310909196809,
+      "grad_norm": 0.363942414522171,
+      "learning_rate": 0.0004923111888111887,
+      "loss": 3.442,
+      "step": 30900
+    },
+    {
+      "epoch": 9.012813792300076,
+      "grad_norm": 0.31260979175567627,
+      "learning_rate": 0.0004921363636363635,
+      "loss": 3.3326,
+      "step": 30950
+    },
+    {
+      "epoch": 9.027374919913798,
+      "grad_norm": 0.3426477909088135,
+      "learning_rate": 0.0004919615384615384,
+      "loss": 3.3223,
+      "step": 31000
+    },
+    {
+      "epoch": 9.027374919913798,
+      "eval_accuracy": 0.36718293893821685,
+      "eval_loss": 3.578636646270752,
+      "eval_runtime": 179.7856,
+      "eval_samples_per_second": 92.588,
+      "eval_steps_per_second": 5.79,
+      "step": 31000
+    },
+    {
+      "epoch": 9.041936047527521,
+      "grad_norm": 0.3265765309333801,
+      "learning_rate": 0.0004917867132867132,
+      "loss": 3.3151,
+      "step": 31050
+    },
+    {
+      "epoch": 9.056497175141242,
+      "grad_norm": 0.341594934463501,
+      "learning_rate": 0.000491611888111888,
+      "loss": 3.3359,
+      "step": 31100
+    },
+    {
+      "epoch": 9.071058302754965,
+      "grad_norm": 0.3292291462421417,
+      "learning_rate": 0.0004914370629370628,
+      "loss": 3.3368,
+      "step": 31150
+    },
+    {
+      "epoch": 9.085619430368688,
+      "grad_norm": 0.3459334671497345,
+      "learning_rate": 0.0004912622377622378,
+      "loss": 3.3496,
+      "step": 31200
+    },
+    {
+      "epoch": 9.10018055798241,
+      "grad_norm": 0.3272961974143982,
+      "learning_rate": 0.0004910874125874126,
+      "loss": 3.3405,
+      "step": 31250
+    },
+    {
+      "epoch": 9.114741685596133,
+      "grad_norm": 0.32643869519233704,
+      "learning_rate": 0.0004909125874125874,
+      "loss": 3.3479,
+      "step": 31300
+    },
+    {
+      "epoch": 9.129302813209854,
+      "grad_norm": 0.3543800413608551,
+      "learning_rate": 0.0004907377622377623,
+      "loss": 3.3437,
+      "step": 31350
+    },
+    {
+      "epoch": 9.143863940823577,
+      "grad_norm": 0.3529180586338043,
+      "learning_rate": 0.0004905629370629371,
+      "loss": 3.3537,
+      "step": 31400
+    },
+    {
+      "epoch": 9.1584250684373,
+      "grad_norm": 0.3241512179374695,
+      "learning_rate": 0.0004903881118881119,
+      "loss": 3.3549,
+      "step": 31450
+    },
+    {
+      "epoch": 9.172986196051022,
+      "grad_norm": 0.3228810429573059,
+      "learning_rate": 0.0004902132867132867,
+      "loss": 3.3541,
+      "step": 31500
+    },
+    {
+      "epoch": 9.187547323664745,
+      "grad_norm": 0.350665420293808,
+      "learning_rate": 0.0004900384615384615,
+      "loss": 3.3433,
+      "step": 31550
+    },
+    {
+      "epoch": 9.202108451278466,
+      "grad_norm": 0.3541180491447449,
+      "learning_rate": 0.0004898636363636363,
+      "loss": 3.3573,
+      "step": 31600
+    },
+    {
+      "epoch": 9.216669578892189,
+      "grad_norm": 0.33510008454322815,
+      "learning_rate": 0.0004896888111888112,
+      "loss": 3.363,
+      "step": 31650
+    },
+    {
+      "epoch": 9.231230706505912,
+      "grad_norm": 0.3581586182117462,
+      "learning_rate": 0.000489513986013986,
+      "loss": 3.363,
+      "step": 31700
+    },
+    {
+      "epoch": 9.245791834119634,
+      "grad_norm": 0.33881622552871704,
+      "learning_rate": 0.0004893391608391608,
+      "loss": 3.3719,
+      "step": 31750
+    },
+    {
+      "epoch": 9.260352961733357,
+      "grad_norm": 0.34674209356307983,
+      "learning_rate": 0.0004891643356643356,
+      "loss": 3.3701,
+      "step": 31800
+    },
+    {
+      "epoch": 9.27491408934708,
+      "grad_norm": 0.3241184949874878,
+      "learning_rate": 0.0004889895104895105,
+      "loss": 3.3741,
+      "step": 31850
+    },
+    {
+      "epoch": 9.2894752169608,
+      "grad_norm": 0.354744017124176,
+      "learning_rate": 0.0004888146853146853,
+      "loss": 3.3578,
+      "step": 31900
+    },
+    {
+      "epoch": 9.304036344574524,
+      "grad_norm": 0.31170573830604553,
+      "learning_rate": 0.0004886398601398601,
+      "loss": 3.37,
+      "step": 31950
+    },
+    {
+      "epoch": 9.318597472188246,
+      "grad_norm": 0.341860830783844,
+      "learning_rate": 0.000488465034965035,
+      "loss": 3.3788,
+      "step": 32000
+    },
+    {
+      "epoch": 9.318597472188246,
+      "eval_accuracy": 0.36685105969758663,
+      "eval_loss": 3.579946279525757,
+      "eval_runtime": 179.6871,
+      "eval_samples_per_second": 92.639,
+      "eval_steps_per_second": 5.793,
+      "step": 32000
+    },
+    {
+      "epoch": 9.333158599801969,
+      "grad_norm": 0.33674827218055725,
+      "learning_rate": 0.0004882902097902098,
+      "loss": 3.3761,
+      "step": 32050
+    },
+    {
+      "epoch": 9.347719727415692,
+      "grad_norm": 0.34504297375679016,
+      "learning_rate": 0.0004881153846153846,
+      "loss": 3.3931,
+      "step": 32100
+    },
+    {
+      "epoch": 9.362280855029413,
+      "grad_norm": 0.334306538105011,
+      "learning_rate": 0.0004879405594405594,
+      "loss": 3.3882,
+      "step": 32150
+    },
+    {
+      "epoch": 9.376841982643136,
+      "grad_norm": 0.3271053433418274,
+      "learning_rate": 0.00048776573426573424,
+      "loss": 3.3648,
+      "step": 32200
+    },
+    {
+      "epoch": 9.391403110256858,
+      "grad_norm": 0.3338295519351959,
+      "learning_rate": 0.00048759090909090904,
+      "loss": 3.3715,
+      "step": 32250
+    },
+    {
+      "epoch": 9.405964237870581,
+      "grad_norm": 0.34873339533805847,
+      "learning_rate": 0.0004874160839160839,
+      "loss": 3.3766,
+      "step": 32300
+    },
+    {
+      "epoch": 9.420525365484304,
+      "grad_norm": 0.36661872267723083,
+      "learning_rate": 0.0004872412587412587,
+      "loss": 3.3698,
+      "step": 32350
+    },
+    {
+      "epoch": 9.435086493098025,
+      "grad_norm": 0.3396371006965637,
+      "learning_rate": 0.00048706643356643354,
+      "loss": 3.3939,
+      "step": 32400
+    },
+    {
+      "epoch": 9.449647620711747,
+      "grad_norm": 0.34849968552589417,
+      "learning_rate": 0.00048689160839160834,
+      "loss": 3.3774,
+      "step": 32450
+    },
+    {
+      "epoch": 9.46420874832547,
+      "grad_norm": 0.3258074223995209,
+      "learning_rate": 0.0004867167832167832,
+      "loss": 3.3934,
+      "step": 32500
+    },
+    {
+      "epoch": 9.478769875939193,
+      "grad_norm": 0.35088443756103516,
+      "learning_rate": 0.00048654195804195794,
+      "loss": 3.379,
+      "step": 32550
+    },
+    {
+      "epoch": 9.493331003552916,
+      "grad_norm": 0.3309447765350342,
+      "learning_rate": 0.00048636713286713285,
+      "loss": 3.3924,
+      "step": 32600
+    },
+    {
+      "epoch": 9.507892131166638,
+      "grad_norm": 0.3345271944999695,
+      "learning_rate": 0.0004861923076923077,
+      "loss": 3.3948,
+      "step": 32650
+    },
+    {
+      "epoch": 9.52245325878036,
+      "grad_norm": 0.3319976031780243,
+      "learning_rate": 0.00048601748251748245,
+      "loss": 3.3951,
+      "step": 32700
+    },
+    {
+      "epoch": 9.537014386394082,
+      "grad_norm": 0.3754766583442688,
+      "learning_rate": 0.0004858426573426573,
+      "loss": 3.3805,
+      "step": 32750
+    },
+    {
+      "epoch": 9.551575514007805,
+      "grad_norm": 0.34478750824928284,
+      "learning_rate": 0.0004856678321678321,
+      "loss": 3.3952,
+      "step": 32800
+    },
+    {
+      "epoch": 9.566136641621528,
+      "grad_norm": 0.32059377431869507,
+      "learning_rate": 0.00048549300699300696,
+      "loss": 3.3938,
+      "step": 32850
+    },
+    {
+      "epoch": 9.58069776923525,
+      "grad_norm": 0.3344727158546448,
+      "learning_rate": 0.00048531818181818176,
+      "loss": 3.4108,
+      "step": 32900
+    },
+    {
+      "epoch": 9.595258896848971,
+      "grad_norm": 0.3274308443069458,
+      "learning_rate": 0.0004851433566433566,
+      "loss": 3.3958,
+      "step": 32950
+    },
+    {
+      "epoch": 9.609820024462694,
+      "grad_norm": 0.37887972593307495,
+      "learning_rate": 0.0004849685314685314,
+      "loss": 3.3937,
+      "step": 33000
+    },
+    {
+      "epoch": 9.609820024462694,
+      "eval_accuracy": 0.36769280796641846,
+      "eval_loss": 3.572748899459839,
+      "eval_runtime": 179.797,
+      "eval_samples_per_second": 92.582,
+      "eval_steps_per_second": 5.79,
+      "step": 33000
+    },
+    {
+      "epoch": 9.624381152076417,
+      "grad_norm": 0.32813072204589844,
+      "learning_rate": 0.00048479370629370627,
+      "loss": 3.3986,
+      "step": 33050
+    },
+    {
+      "epoch": 9.63894227969014,
+      "grad_norm": 0.3385777771472931,
+      "learning_rate": 0.00048461888111888106,
+      "loss": 3.4083,
+      "step": 33100
+    },
+    {
+      "epoch": 9.653503407303862,
+      "grad_norm": 0.34293919801712036,
+      "learning_rate": 0.0004844440559440559,
+      "loss": 3.4105,
+      "step": 33150
+    },
+    {
+      "epoch": 9.668064534917583,
+      "grad_norm": 0.34296914935112,
+      "learning_rate": 0.0004842692307692307,
+      "loss": 3.387,
+      "step": 33200
+    },
+    {
+      "epoch": 9.682625662531306,
+      "grad_norm": 0.3283473551273346,
+      "learning_rate": 0.00048409440559440557,
+      "loss": 3.3968,
+      "step": 33250
+    },
+    {
+      "epoch": 9.697186790145029,
+      "grad_norm": 0.3375408351421356,
+      "learning_rate": 0.0004839195804195803,
+      "loss": 3.4106,
+      "step": 33300
+    },
+    {
+      "epoch": 9.711747917758752,
+      "grad_norm": 0.38011592626571655,
+      "learning_rate": 0.0004837447552447552,
+      "loss": 3.3923,
+      "step": 33350
+    },
+    {
+      "epoch": 9.726309045372474,
+      "grad_norm": 0.3376014232635498,
+      "learning_rate": 0.0004835699300699301,
+      "loss": 3.4004,
+      "step": 33400
+    },
+    {
+      "epoch": 9.740870172986195,
+      "grad_norm": 0.34998080134391785,
+      "learning_rate": 0.0004833951048951048,
+      "loss": 3.415,
+      "step": 33450
+    },
+    {
+      "epoch": 9.755431300599918,
+      "grad_norm": 0.33272120356559753,
+      "learning_rate": 0.0004832202797202797,
+      "loss": 3.4076,
+      "step": 33500
+    },
+    {
+      "epoch": 9.76999242821364,
+      "grad_norm": 0.3352126181125641,
+      "learning_rate": 0.0004830454545454545,
+      "loss": 3.3951,
+      "step": 33550
+    },
+    {
+      "epoch": 9.784553555827364,
+      "grad_norm": 0.32453832030296326,
+      "learning_rate": 0.00048287062937062933,
+      "loss": 3.4116,
+      "step": 33600
+    },
+    {
+      "epoch": 9.799114683441086,
+      "grad_norm": 0.35308536887168884,
+      "learning_rate": 0.00048269580419580413,
+      "loss": 3.4049,
+      "step": 33650
+    },
+    {
+      "epoch": 9.813675811054807,
+      "grad_norm": 0.36469438672065735,
+      "learning_rate": 0.000482520979020979,
+      "loss": 3.4074,
+      "step": 33700
+    },
+    {
+      "epoch": 9.82823693866853,
+      "grad_norm": 0.312566339969635,
+      "learning_rate": 0.0004823461538461538,
+      "loss": 3.3985,
+      "step": 33750
+    },
+    {
+      "epoch": 9.842798066282253,
+      "grad_norm": 0.34336328506469727,
+      "learning_rate": 0.00048217132867132864,
+      "loss": 3.3979,
+      "step": 33800
+    },
+    {
+      "epoch": 9.857359193895975,
+      "grad_norm": 0.35954421758651733,
+      "learning_rate": 0.00048199650349650344,
+      "loss": 3.3987,
+      "step": 33850
+    },
+    {
+      "epoch": 9.871920321509698,
+      "grad_norm": 0.3485148847103119,
+      "learning_rate": 0.0004818216783216783,
+      "loss": 3.3936,
+      "step": 33900
+    },
+    {
+      "epoch": 9.88648144912342,
+      "grad_norm": 0.3321332633495331,
+      "learning_rate": 0.0004816468531468531,
+      "loss": 3.4155,
+      "step": 33950
+    },
+    {
+      "epoch": 9.901042576737142,
+      "grad_norm": 0.34463703632354736,
+      "learning_rate": 0.00048147202797202795,
+      "loss": 3.4155,
+      "step": 34000
+    },
+    {
+      "epoch": 9.901042576737142,
+      "eval_accuracy": 0.3683254123567235,
+      "eval_loss": 3.5605504512786865,
+      "eval_runtime": 179.6436,
+      "eval_samples_per_second": 92.661,
+      "eval_steps_per_second": 5.795,
+      "step": 34000
+    },
+    {
+      "epoch": 9.915603704350865,
+      "grad_norm": 0.33447834849357605,
+      "learning_rate": 0.0004812972027972028,
+      "loss": 3.407,
+      "step": 34050
+    },
+    {
+      "epoch": 9.930164831964587,
+      "grad_norm": 0.3506791889667511,
+      "learning_rate": 0.0004811223776223776,
+      "loss": 3.4097,
+      "step": 34100
+    },
+    {
+      "epoch": 9.94472595957831,
+      "grad_norm": 0.33796730637550354,
+      "learning_rate": 0.00048094755244755245,
+      "loss": 3.4044,
+      "step": 34150
+    },
+    {
+      "epoch": 9.959287087192033,
+      "grad_norm": 0.3389424979686737,
+      "learning_rate": 0.0004807727272727272,
+      "loss": 3.4017,
+      "step": 34200
+    },
+    {
+      "epoch": 9.973848214805754,
+      "grad_norm": 0.32537972927093506,
+      "learning_rate": 0.00048059790209790205,
+      "loss": 3.4185,
+      "step": 34250
+    },
+    {
+      "epoch": 9.988409342419477,
+      "grad_norm": 0.349602073431015,
+      "learning_rate": 0.00048042307692307685,
+      "loss": 3.4123,
+      "step": 34300
+    },
+    {
+      "epoch": 10.002912225522744,
+      "grad_norm": 0.3237372040748596,
+      "learning_rate": 0.0004802482517482517,
+      "loss": 3.3901,
+      "step": 34350
+    },
+    {
+      "epoch": 10.017473353136467,
+      "grad_norm": 0.35387957096099854,
+      "learning_rate": 0.0004800734265734265,
+      "loss": 3.2917,
+      "step": 34400
+    },
+    {
+      "epoch": 10.03203448075019,
+      "grad_norm": 0.35952070355415344,
+      "learning_rate": 0.00047989860139860136,
+      "loss": 3.2902,
+      "step": 34450
+    },
+    {
+      "epoch": 10.046595608363912,
+      "grad_norm": 0.3631179928779602,
+      "learning_rate": 0.00047972377622377616,
+      "loss": 3.2917,
+      "step": 34500
+    },
+    {
+      "epoch": 10.061156735977635,
+      "grad_norm": 0.38791096210479736,
+      "learning_rate": 0.000479548951048951,
+      "loss": 3.3143,
+      "step": 34550
+    },
+    {
+      "epoch": 10.075717863591356,
+      "grad_norm": 0.3530723750591278,
+      "learning_rate": 0.0004793741258741258,
+      "loss": 3.3145,
+      "step": 34600
+    },
+    {
+      "epoch": 10.090278991205079,
+      "grad_norm": 0.3433317542076111,
+      "learning_rate": 0.00047919930069930067,
+      "loss": 3.3156,
+      "step": 34650
+    },
+    {
+      "epoch": 10.104840118818801,
+      "grad_norm": 0.31986525654792786,
+      "learning_rate": 0.0004790244755244755,
+      "loss": 3.3162,
+      "step": 34700
+    },
+    {
+      "epoch": 10.119401246432524,
+      "grad_norm": 0.3420194089412689,
+      "learning_rate": 0.0004788496503496503,
+      "loss": 3.31,
+      "step": 34750
+    },
+    {
+      "epoch": 10.133962374046247,
+      "grad_norm": 0.35869932174682617,
+      "learning_rate": 0.0004786748251748252,
+      "loss": 3.3294,
+      "step": 34800
+    },
+    {
+      "epoch": 10.148523501659968,
+      "grad_norm": 0.3446190655231476,
+      "learning_rate": 0.0004785,
+      "loss": 3.3318,
+      "step": 34850
+    },
+    {
+      "epoch": 10.16308462927369,
+      "grad_norm": 0.3328004777431488,
+      "learning_rate": 0.00047832517482517483,
+      "loss": 3.3373,
+      "step": 34900
+    },
+    {
+      "epoch": 10.177645756887413,
+      "grad_norm": 0.37485411763191223,
+      "learning_rate": 0.0004781503496503496,
+      "loss": 3.3202,
+      "step": 34950
+    },
+    {
+      "epoch": 10.192206884501136,
+      "grad_norm": 0.36680352687835693,
+      "learning_rate": 0.00047797552447552443,
+      "loss": 3.34,
+      "step": 35000
+    },
+    {
+      "epoch": 10.192206884501136,
+      "eval_accuracy": 0.3679084177883511,
+      "eval_loss": 3.5739078521728516,
+      "eval_runtime": 231.8779,
+      "eval_samples_per_second": 71.788,
+      "eval_steps_per_second": 4.489,
+      "step": 35000
+    },
+    {
+      "epoch": 10.206768012114859,
+      "grad_norm": 0.3383858799934387,
+      "learning_rate": 0.00047780069930069923,
+      "loss": 3.3325,
+      "step": 35050
+    },
+    {
+      "epoch": 10.221329139728581,
+      "grad_norm": 0.3608876168727875,
+      "learning_rate": 0.0004776258741258741,
+      "loss": 3.3455,
+      "step": 35100
+    },
+    {
+      "epoch": 10.235890267342302,
+      "grad_norm": 0.38920077681541443,
+      "learning_rate": 0.0004774510489510489,
+      "loss": 3.3508,
+      "step": 35150
+    },
+    {
+      "epoch": 10.250451394956025,
+      "grad_norm": 0.3486644923686981,
+      "learning_rate": 0.00047727622377622374,
+      "loss": 3.3526,
+      "step": 35200
+    },
+    {
+      "epoch": 10.265012522569748,
+      "grad_norm": 0.3848312199115753,
+      "learning_rate": 0.00047710139860139854,
+      "loss": 3.3389,
+      "step": 35250
+    },
+    {
+      "epoch": 10.27957365018347,
+      "grad_norm": 0.34103095531463623,
+      "learning_rate": 0.0004769265734265734,
+      "loss": 3.336,
+      "step": 35300
+    },
+    {
+      "epoch": 10.294134777797193,
+      "grad_norm": 0.3553299605846405,
+      "learning_rate": 0.0004767517482517482,
+      "loss": 3.3438,
+      "step": 35350
+    },
+    {
+      "epoch": 10.308695905410914,
+      "grad_norm": 0.3677619695663452,
+      "learning_rate": 0.00047657692307692304,
+      "loss": 3.3505,
+      "step": 35400
+    },
+    {
+      "epoch": 10.323257033024637,
+      "grad_norm": 0.3474290370941162,
+      "learning_rate": 0.0004764020979020979,
+      "loss": 3.3535,
+      "step": 35450
+    },
+    {
+      "epoch": 10.33781816063836,
+      "grad_norm": 0.35205620527267456,
+      "learning_rate": 0.0004762272727272727,
+      "loss": 3.3448,
+      "step": 35500
+    },
+    {
+      "epoch": 10.352379288252083,
+      "grad_norm": 0.3427030146121979,
+      "learning_rate": 0.00047605244755244755,
+      "loss": 3.3546,
+      "step": 35550
+    },
+    {
+      "epoch": 10.366940415865805,
+      "grad_norm": 0.36262500286102295,
+      "learning_rate": 0.00047587762237762235,
+      "loss": 3.3639,
+      "step": 35600
+    },
+    {
+      "epoch": 10.381501543479526,
+      "grad_norm": 0.37175849080085754,
+      "learning_rate": 0.0004757027972027972,
+      "loss": 3.3623,
+      "step": 35650
+    },
+    {
+      "epoch": 10.396062671093249,
+      "grad_norm": 0.32053637504577637,
+      "learning_rate": 0.00047552797202797195,
+      "loss": 3.3639,
+      "step": 35700
+    },
+    {
+      "epoch": 10.410623798706972,
+      "grad_norm": 0.3394221365451813,
+      "learning_rate": 0.0004753531468531468,
+      "loss": 3.3564,
+      "step": 35750
+    },
+    {
+      "epoch": 10.425184926320695,
+      "grad_norm": 0.33870938420295715,
+      "learning_rate": 0.0004751783216783216,
+      "loss": 3.3586,
+      "step": 35800
+    },
+    {
+      "epoch": 10.439746053934417,
+      "grad_norm": 0.3561233878135681,
+      "learning_rate": 0.00047500349650349646,
+      "loss": 3.3674,
+      "step": 35850
+    },
+    {
+      "epoch": 10.454307181548138,
+      "grad_norm": 0.3519926071166992,
+      "learning_rate": 0.00047482867132867126,
+      "loss": 3.3557,
+      "step": 35900
+    },
+    {
+      "epoch": 10.468868309161861,
+      "grad_norm": 0.33235129714012146,
+      "learning_rate": 0.0004746538461538461,
+      "loss": 3.3529,
+      "step": 35950
+    },
+    {
+      "epoch": 10.483429436775584,
+      "grad_norm": 0.3245657682418823,
+      "learning_rate": 0.0004744790209790209,
+      "loss": 3.3755,
+      "step": 36000
+    },
+    {
+      "epoch": 10.483429436775584,
+      "eval_accuracy": 0.3681230871094247,
+      "eval_loss": 3.5692830085754395,
+      "eval_runtime": 179.6694,
+      "eval_samples_per_second": 92.648,
+      "eval_steps_per_second": 5.794,
+      "step": 36000
+    },
+    {
+      "epoch": 10.497990564389307,
+      "grad_norm": 0.33548834919929504,
+      "learning_rate": 0.00047430419580419576,
+      "loss": 3.3667,
+      "step": 36050
+    },
+    {
+      "epoch": 10.51255169200303,
+      "grad_norm": 0.3279496133327484,
+      "learning_rate": 0.0004741293706293706,
+      "loss": 3.3668,
+      "step": 36100
+    },
+    {
+      "epoch": 10.52711281961675,
+      "grad_norm": 0.3511890470981598,
+      "learning_rate": 0.0004739545454545454,
+      "loss": 3.3761,
+      "step": 36150
+    },
+    {
+      "epoch": 10.541673947230473,
+      "grad_norm": 0.34296634793281555,
+      "learning_rate": 0.00047377972027972027,
+      "loss": 3.3659,
+      "step": 36200
+    },
+    {
+      "epoch": 10.556235074844196,
+      "grad_norm": 0.39123594760894775,
+      "learning_rate": 0.00047360489510489507,
+      "loss": 3.3614,
+      "step": 36250
+    },
+    {
+      "epoch": 10.570796202457919,
+      "grad_norm": 0.3517671227455139,
+      "learning_rate": 0.0004734300699300699,
+      "loss": 3.3887,
+      "step": 36300
+    },
+    {
+      "epoch": 10.585357330071641,
+      "grad_norm": 0.31128349900245667,
+      "learning_rate": 0.0004732552447552447,
+      "loss": 3.3766,
+      "step": 36350
+    },
+    {
+      "epoch": 10.599918457685362,
+      "grad_norm": 0.3317906856536865,
+      "learning_rate": 0.0004730804195804196,
+      "loss": 3.3748,
+      "step": 36400
+    },
+    {
+      "epoch": 10.614479585299085,
+      "grad_norm": 0.33843693137168884,
+      "learning_rate": 0.0004729055944055943,
+      "loss": 3.3758,
+      "step": 36450
+    },
+    {
+      "epoch": 10.629040712912808,
+      "grad_norm": 0.3605981767177582,
+      "learning_rate": 0.0004727307692307692,
+      "loss": 3.3822,
+      "step": 36500
+    },
+    {
+      "epoch": 10.64360184052653,
+      "grad_norm": 0.36126452684402466,
+      "learning_rate": 0.000472555944055944,
+      "loss": 3.3727,
+      "step": 36550
+    },
+    {
+      "epoch": 10.658162968140253,
+      "grad_norm": 0.3546067476272583,
+      "learning_rate": 0.00047238111888111883,
+      "loss": 3.365,
+      "step": 36600
+    },
+    {
+      "epoch": 10.672724095753976,
+      "grad_norm": 0.322816401720047,
+      "learning_rate": 0.00047220629370629363,
+      "loss": 3.3828,
+      "step": 36650
+    },
+    {
+      "epoch": 10.687285223367697,
+      "grad_norm": 0.3517470061779022,
+      "learning_rate": 0.0004720314685314685,
+      "loss": 3.365,
+      "step": 36700
+    },
+    {
+      "epoch": 10.70184635098142,
+      "grad_norm": 0.3618316948413849,
+      "learning_rate": 0.0004718566433566433,
+      "loss": 3.3713,
+      "step": 36750
+    },
+    {
+      "epoch": 10.716407478595142,
+      "grad_norm": 0.3243914544582367,
+      "learning_rate": 0.00047168181818181814,
+      "loss": 3.3811,
+      "step": 36800
+    },
+    {
+      "epoch": 10.730968606208865,
+      "grad_norm": 0.3605787456035614,
+      "learning_rate": 0.000471506993006993,
+      "loss": 3.3727,
+      "step": 36850
+    },
+    {
+      "epoch": 10.745529733822588,
+      "grad_norm": 0.3196176588535309,
+      "learning_rate": 0.0004713321678321678,
+      "loss": 3.3806,
+      "step": 36900
+    },
+    {
+      "epoch": 10.760090861436309,
+      "grad_norm": 0.35002508759498596,
+      "learning_rate": 0.00047115734265734265,
+      "loss": 3.4017,
+      "step": 36950
+    },
+    {
+      "epoch": 10.774651989050032,
+      "grad_norm": 0.35589703917503357,
+      "learning_rate": 0.00047098251748251745,
+      "loss": 3.3982,
+      "step": 37000
+    },
+    {
+      "epoch": 10.774651989050032,
+      "eval_accuracy": 0.3692364050013014,
+      "eval_loss": 3.5608346462249756,
+      "eval_runtime": 179.5409,
+      "eval_samples_per_second": 92.714,
+      "eval_steps_per_second": 5.798,
+      "step": 37000
+    },
+    {
+      "epoch": 10.789213116663754,
+      "grad_norm": 0.34006816148757935,
+      "learning_rate": 0.0004708076923076923,
+      "loss": 3.3882,
+      "step": 37050
+    },
+    {
+      "epoch": 10.803774244277477,
+      "grad_norm": 0.34267741441726685,
+      "learning_rate": 0.0004706328671328671,
+      "loss": 3.3936,
+      "step": 37100
+    },
+    {
+      "epoch": 10.8183353718912,
+      "grad_norm": 0.3394986391067505,
+      "learning_rate": 0.00047045804195804195,
+      "loss": 3.3949,
+      "step": 37150
+    },
+    {
+      "epoch": 10.83289649950492,
+      "grad_norm": 0.35784146189689636,
+      "learning_rate": 0.0004702832167832167,
+      "loss": 3.386,
+      "step": 37200
+    },
+    {
+      "epoch": 10.847457627118644,
+      "grad_norm": 0.3268081247806549,
+      "learning_rate": 0.00047010839160839155,
+      "loss": 3.3931,
+      "step": 37250
+    },
+    {
+      "epoch": 10.862018754732366,
+      "grad_norm": 0.34974247217178345,
+      "learning_rate": 0.00046993356643356635,
+      "loss": 3.3942,
+      "step": 37300
+    },
+    {
+      "epoch": 10.876579882346089,
+      "grad_norm": 0.3425455093383789,
+      "learning_rate": 0.0004697587412587412,
+      "loss": 3.3862,
+      "step": 37350
+    },
+    {
+      "epoch": 10.891141009959812,
+      "grad_norm": 0.32915806770324707,
+      "learning_rate": 0.000469583916083916,
+      "loss": 3.3955,
+      "step": 37400
+    },
+    {
+      "epoch": 10.905702137573535,
+      "grad_norm": 0.33617880940437317,
+      "learning_rate": 0.00046940909090909086,
+      "loss": 3.3878,
+      "step": 37450
+    },
+    {
+      "epoch": 10.920263265187256,
+      "grad_norm": 0.3757038414478302,
+      "learning_rate": 0.0004692342657342657,
+      "loss": 3.3805,
+      "step": 37500
+    },
+    {
+      "epoch": 10.934824392800978,
+      "grad_norm": 0.34372201561927795,
+      "learning_rate": 0.0004690594405594405,
+      "loss": 3.387,
+      "step": 37550
+    },
+    {
+      "epoch": 10.949385520414701,
+      "grad_norm": 0.33192873001098633,
+      "learning_rate": 0.00046888461538461537,
+      "loss": 3.3873,
+      "step": 37600
+    },
+    {
+      "epoch": 10.963946648028424,
+      "grad_norm": 0.38427871465682983,
+      "learning_rate": 0.00046870979020979017,
+      "loss": 3.3847,
+      "step": 37650
+    },
+    {
+      "epoch": 10.978507775642147,
+      "grad_norm": 0.3522765338420868,
+      "learning_rate": 0.000468534965034965,
+      "loss": 3.3893,
+      "step": 37700
+    },
+    {
+      "epoch": 10.993068903255867,
+      "grad_norm": 0.34591537714004517,
+      "learning_rate": 0.0004683601398601398,
+      "loss": 3.3719,
+      "step": 37750
+    },
+    {
+      "epoch": 11.007571786359136,
+      "grad_norm": 0.3357653319835663,
+      "learning_rate": 0.0004681853146853147,
+      "loss": 3.3227,
+      "step": 37800
+    },
+    {
+      "epoch": 11.022132913972857,
+      "grad_norm": 0.3487904369831085,
+      "learning_rate": 0.0004680104895104895,
+      "loss": 3.2621,
+      "step": 37850
+    },
+    {
+      "epoch": 11.03669404158658,
+      "grad_norm": 0.34650513529777527,
+      "learning_rate": 0.00046783566433566433,
+      "loss": 3.2841,
+      "step": 37900
+    },
+    {
+      "epoch": 11.051255169200303,
+      "grad_norm": 0.3532591760158539,
+      "learning_rate": 0.0004676608391608391,
+      "loss": 3.3019,
+      "step": 37950
+    },
+    {
+      "epoch": 11.065816296814026,
+      "grad_norm": 0.3463428020477295,
+      "learning_rate": 0.00046748601398601393,
+      "loss": 3.2846,
+      "step": 38000
+    },
+    {
+      "epoch": 11.065816296814026,
+      "eval_accuracy": 0.3690252625584492,
+      "eval_loss": 3.5694565773010254,
+      "eval_runtime": 179.6506,
+      "eval_samples_per_second": 92.658,
+      "eval_steps_per_second": 5.795,
+      "step": 38000
+    },
+    {
+      "epoch": 11.080377424427748,
+      "grad_norm": 0.34090572595596313,
+      "learning_rate": 0.00046731118881118873,
+      "loss": 3.2824,
+      "step": 38050
+    },
+    {
+      "epoch": 11.09493855204147,
+      "grad_norm": 0.342952162027359,
+      "learning_rate": 0.0004671363636363636,
+      "loss": 3.2853,
+      "step": 38100
+    },
+    {
+      "epoch": 11.109499679655192,
+      "grad_norm": 0.34960615634918213,
+      "learning_rate": 0.00046696153846153844,
+      "loss": 3.306,
+      "step": 38150
+    },
+    {
+      "epoch": 11.124060807268915,
+      "grad_norm": 0.3609829545021057,
+      "learning_rate": 0.00046678671328671324,
+      "loss": 3.3002,
+      "step": 38200
+    },
+    {
+      "epoch": 11.138621934882638,
+      "grad_norm": 0.36214137077331543,
+      "learning_rate": 0.0004666118881118881,
+      "loss": 3.301,
+      "step": 38250
+    },
+    {
+      "epoch": 11.15318306249636,
+      "grad_norm": 0.34393003582954407,
+      "learning_rate": 0.0004664370629370629,
+      "loss": 3.2942,
+      "step": 38300
+    },
+    {
+      "epoch": 11.167744190110081,
+      "grad_norm": 0.34303128719329834,
+      "learning_rate": 0.00046626223776223774,
+      "loss": 3.3214,
+      "step": 38350
+    },
+    {
+      "epoch": 11.182305317723804,
+      "grad_norm": 0.328708291053772,
+      "learning_rate": 0.00046608741258741254,
+      "loss": 3.3219,
+      "step": 38400
+    },
+    {
+      "epoch": 11.196866445337527,
+      "grad_norm": 0.3446009159088135,
+      "learning_rate": 0.0004659125874125874,
+      "loss": 3.3167,
+      "step": 38450
+    },
+    {
+      "epoch": 11.21142757295125,
+      "grad_norm": 0.34211987257003784,
+      "learning_rate": 0.0004657377622377622,
+      "loss": 3.3174,
+      "step": 38500
+    },
+    {
+      "epoch": 11.225988700564972,
+      "grad_norm": 0.3242914378643036,
+      "learning_rate": 0.00046556293706293705,
+      "loss": 3.3241,
+      "step": 38550
+    },
+    {
+      "epoch": 11.240549828178693,
+      "grad_norm": 0.3906053304672241,
+      "learning_rate": 0.00046538811188811185,
+      "loss": 3.3242,
+      "step": 38600
+    },
+    {
+      "epoch": 11.255110955792416,
+      "grad_norm": 0.35929811000823975,
+      "learning_rate": 0.0004652132867132867,
+      "loss": 3.3162,
+      "step": 38650
+    },
+    {
+      "epoch": 11.269672083406139,
+      "grad_norm": 0.362967312335968,
+      "learning_rate": 0.00046503846153846145,
+      "loss": 3.3231,
+      "step": 38700
+    },
+    {
+      "epoch": 11.284233211019862,
+      "grad_norm": 0.3617202639579773,
+      "learning_rate": 0.0004648636363636363,
+      "loss": 3.3383,
+      "step": 38750
+    },
+    {
+      "epoch": 11.298794338633584,
+      "grad_norm": 0.3506316840648651,
+      "learning_rate": 0.0004646888111888111,
+      "loss": 3.3331,
+      "step": 38800
+    },
+    {
+      "epoch": 11.313355466247307,
+      "grad_norm": 0.35573744773864746,
+      "learning_rate": 0.00046451398601398596,
+      "loss": 3.3353,
+      "step": 38850
+    },
+    {
+      "epoch": 11.327916593861028,
+      "grad_norm": 0.3686698377132416,
+      "learning_rate": 0.0004643391608391608,
+      "loss": 3.331,
+      "step": 38900
+    },
+    {
+      "epoch": 11.34247772147475,
+      "grad_norm": 0.3425818681716919,
+      "learning_rate": 0.0004641643356643356,
+      "loss": 3.3356,
+      "step": 38950
+    },
+    {
+      "epoch": 11.357038849088473,
+      "grad_norm": 0.3510589003562927,
+      "learning_rate": 0.00046398951048951046,
+      "loss": 3.3348,
+      "step": 39000
+    },
+    {
+      "epoch": 11.357038849088473,
+      "eval_accuracy": 0.3690062174160538,
+      "eval_loss": 3.5675673484802246,
+      "eval_runtime": 179.7196,
+      "eval_samples_per_second": 92.622,
+      "eval_steps_per_second": 5.792,
+      "step": 39000
+    },
+    {
+      "epoch": 11.371599976702196,
+      "grad_norm": 0.33674928545951843,
+      "learning_rate": 0.00046381468531468526,
+      "loss": 3.3398,
+      "step": 39050
+    },
+    {
+      "epoch": 11.386161104315919,
+      "grad_norm": 0.3514574468135834,
+      "learning_rate": 0.0004636398601398601,
+      "loss": 3.3288,
+      "step": 39100
+    },
+    {
+      "epoch": 11.40072223192964,
+      "grad_norm": 0.3495193123817444,
+      "learning_rate": 0.0004634650349650349,
+      "loss": 3.3363,
+      "step": 39150
+    },
+    {
+      "epoch": 11.415283359543363,
+      "grad_norm": 0.32252052426338196,
+      "learning_rate": 0.00046329020979020977,
+      "loss": 3.339,
+      "step": 39200
+    },
+    {
+      "epoch": 11.429844487157085,
+      "grad_norm": 0.35050782561302185,
+      "learning_rate": 0.00046311538461538457,
+      "loss": 3.3347,
+      "step": 39250
+    },
+    {
+      "epoch": 11.444405614770808,
+      "grad_norm": 0.3583682179450989,
+      "learning_rate": 0.0004629405594405594,
+      "loss": 3.3401,
+      "step": 39300
+    },
+    {
+      "epoch": 11.458966742384531,
+      "grad_norm": 0.34238868951797485,
+      "learning_rate": 0.0004627657342657342,
+      "loss": 3.3355,
+      "step": 39350
+    },
+    {
+      "epoch": 11.473527869998252,
+      "grad_norm": 0.37667524814605713,
+      "learning_rate": 0.0004625909090909091,
+      "loss": 3.3388,
+      "step": 39400
+    },
+    {
+      "epoch": 11.488088997611975,
+      "grad_norm": 0.3632262349128723,
+      "learning_rate": 0.0004624160839160838,
+      "loss": 3.3425,
+      "step": 39450
+    },
+    {
+      "epoch": 11.502650125225697,
+      "grad_norm": 0.35607948899269104,
+      "learning_rate": 0.0004622412587412587,
+      "loss": 3.351,
+      "step": 39500
+    },
+    {
+      "epoch": 11.51721125283942,
+      "grad_norm": 0.37387025356292725,
+      "learning_rate": 0.00046206643356643353,
+      "loss": 3.3431,
+      "step": 39550
+    },
+    {
+      "epoch": 11.531772380453143,
+      "grad_norm": 0.33748266100883484,
+      "learning_rate": 0.00046189160839160833,
+      "loss": 3.3548,
+      "step": 39600
+    },
+    {
+      "epoch": 11.546333508066864,
+      "grad_norm": 0.35886332392692566,
+      "learning_rate": 0.0004617167832167832,
+      "loss": 3.3526,
+      "step": 39650
+    },
+    {
+      "epoch": 11.560894635680587,
+      "grad_norm": 0.34527456760406494,
+      "learning_rate": 0.000461541958041958,
+      "loss": 3.3471,
+      "step": 39700
+    },
+    {
+      "epoch": 11.57545576329431,
+      "grad_norm": 0.3368580639362335,
+      "learning_rate": 0.00046136713286713284,
+      "loss": 3.355,
+      "step": 39750
+    },
+    {
+      "epoch": 11.590016890908032,
+      "grad_norm": 0.3541625738143921,
+      "learning_rate": 0.00046119230769230764,
+      "loss": 3.3553,
+      "step": 39800
+    },
+    {
+      "epoch": 11.604578018521755,
+      "grad_norm": 0.3746795058250427,
+      "learning_rate": 0.0004610174825174825,
+      "loss": 3.3639,
+      "step": 39850
+    },
+    {
+      "epoch": 11.619139146135478,
+      "grad_norm": 0.3315316438674927,
+      "learning_rate": 0.0004608426573426573,
+      "loss": 3.3647,
+      "step": 39900
+    },
+    {
+      "epoch": 11.633700273749199,
+      "grad_norm": 0.3384944796562195,
+      "learning_rate": 0.00046066783216783215,
+      "loss": 3.3451,
+      "step": 39950
+    },
+    {
+      "epoch": 11.648261401362921,
+      "grad_norm": 0.3402671813964844,
+      "learning_rate": 0.00046049300699300695,
+      "loss": 3.3578,
+      "step": 40000
+    },
+    {
+      "epoch": 11.648261401362921,
+      "eval_accuracy": 0.3696004963963534,
+      "eval_loss": 3.5571908950805664,
+      "eval_runtime": 179.721,
+      "eval_samples_per_second": 92.621,
+      "eval_steps_per_second": 5.792,
+      "step": 40000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 171700,
+  "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": 0
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.3607701815296e+17,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}