diff --git "a/cost_to_push_frequency_5039/checkpoint-70000/trainer_state.json" "b/cost_to_push_frequency_5039/checkpoint-70000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/cost_to_push_frequency_5039/checkpoint-70000/trainer_state.json"
@@ -0,0 +1,10473 @@
+{
+  "best_global_step": 65000,
+  "best_metric": 3.5314598083496094,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/cost_to_push_frequency_5039/checkpoint-30000",
+  "epoch": 20.384413769002272,
+  "eval_steps": 1000,
+  "global_step": 70000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014561127613722406,
+      "grad_norm": 0.7864501476287842,
+      "learning_rate": 0.000294,
+      "loss": 8.4581,
+      "step": 50
+    },
+    {
+      "epoch": 0.029122255227444813,
+      "grad_norm": 0.5157894492149353,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7389,
+      "step": 100
+    },
+    {
+      "epoch": 0.04368338284116722,
+      "grad_norm": 0.7220619320869446,
+      "learning_rate": 0.0005998286713286713,
+      "loss": 6.358,
+      "step": 150
+    },
+    {
+      "epoch": 0.058244510454889625,
+      "grad_norm": 0.45132747292518616,
+      "learning_rate": 0.0005996538461538461,
+      "loss": 6.1209,
+      "step": 200
+    },
+    {
+      "epoch": 0.07280563806861204,
+      "grad_norm": 0.5135724544525146,
+      "learning_rate": 0.0005994790209790209,
+      "loss": 5.9889,
+      "step": 250
+    },
+    {
+      "epoch": 0.08736676568233444,
+      "grad_norm": 0.492136687040329,
+      "learning_rate": 0.0005993041958041958,
+      "loss": 5.8792,
+      "step": 300
+    },
+    {
+      "epoch": 0.10192789329605685,
+      "grad_norm": 0.4208127558231354,
+      "learning_rate": 0.0005991293706293705,
+      "loss": 5.7207,
+      "step": 350
+    },
+    {
+      "epoch": 0.11648902090977925,
+      "grad_norm": 0.41867050528526306,
+      "learning_rate": 0.0005989545454545454,
+      "loss": 5.6164,
+      "step": 400
+    },
+    {
+      "epoch": 0.13105014852350166,
+      "grad_norm": 0.5124858021736145,
+      "learning_rate": 0.0005987797202797202,
+      "loss": 5.501,
+      "step": 450
+    },
+    {
+      "epoch": 0.14561127613722408,
+      "grad_norm": 0.4207671880722046,
+      "learning_rate": 0.000598604895104895,
+      "loss": 5.406,
+      "step": 500
+    },
+    {
+      "epoch": 0.16017240375094646,
+      "grad_norm": 0.4488603472709656,
+      "learning_rate": 0.0005984300699300698,
+      "loss": 5.3207,
+      "step": 550
+    },
+    {
+      "epoch": 0.17473353136466888,
+      "grad_norm": 0.50998854637146,
+      "learning_rate": 0.0005982552447552447,
+      "loss": 5.243,
+      "step": 600
+    },
+    {
+      "epoch": 0.1892946589783913,
+      "grad_norm": 0.4845823049545288,
+      "learning_rate": 0.0005980804195804195,
+      "loss": 5.175,
+      "step": 650
+    },
+    {
+      "epoch": 0.2038557865921137,
+      "grad_norm": 0.4981078803539276,
+      "learning_rate": 0.0005979055944055943,
+      "loss": 5.1142,
+      "step": 700
+    },
+    {
+      "epoch": 0.2184169142058361,
+      "grad_norm": 0.6035723686218262,
+      "learning_rate": 0.0005977307692307691,
+      "loss": 5.0627,
+      "step": 750
+    },
+    {
+      "epoch": 0.2329780418195585,
+      "grad_norm": 0.49036705493927,
+      "learning_rate": 0.000597555944055944,
+      "loss": 5.0033,
+      "step": 800
+    },
+    {
+      "epoch": 0.24753916943328091,
+      "grad_norm": 0.4498800039291382,
+      "learning_rate": 0.0005973811188811188,
+      "loss": 4.9588,
+      "step": 850
+    },
+    {
+      "epoch": 0.2621002970470033,
+      "grad_norm": 0.44455981254577637,
+      "learning_rate": 0.0005972062937062936,
+      "loss": 4.9198,
+      "step": 900
+    },
+    {
+      "epoch": 0.27666142466072574,
+      "grad_norm": 0.40544578433036804,
+      "learning_rate": 0.0005970314685314685,
+      "loss": 4.8619,
+      "step": 950
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "grad_norm": 0.4749813377857208,
+      "learning_rate": 0.0005968566433566433,
+      "loss": 4.8278,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "eval_accuracy": 0.2567868305426714,
+      "eval_loss": 4.739474773406982,
+      "eval_runtime": 180.3909,
+      "eval_samples_per_second": 92.277,
+      "eval_steps_per_second": 5.771,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30578367988817057,
+      "grad_norm": 0.43021872639656067,
+      "learning_rate": 0.0005966818181818181,
+      "loss": 4.7606,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3203448075018929,
+      "grad_norm": 0.4772787094116211,
+      "learning_rate": 0.0005965069930069929,
+      "loss": 4.7182,
+      "step": 1100
+    },
+    {
+      "epoch": 0.33490593511561534,
+      "grad_norm": 0.4763582646846771,
+      "learning_rate": 0.0005963321678321677,
+      "loss": 4.6945,
+      "step": 1150
+    },
+    {
+      "epoch": 0.34946706272933775,
+      "grad_norm": 0.4743852913379669,
+      "learning_rate": 0.0005961573426573425,
+      "loss": 4.6563,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36402819034306017,
+      "grad_norm": 0.43077409267425537,
+      "learning_rate": 0.0005959825174825174,
+      "loss": 4.6227,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3785893179567826,
+      "grad_norm": 0.4979049563407898,
+      "learning_rate": 0.0005958076923076922,
+      "loss": 4.5896,
+      "step": 1300
+    },
+    {
+      "epoch": 0.393150445570505,
+      "grad_norm": 0.4326588809490204,
+      "learning_rate": 0.000595632867132867,
+      "loss": 4.5628,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4077115731842274,
+      "grad_norm": 0.4023474156856537,
+      "learning_rate": 0.0005954580419580418,
+      "loss": 4.5543,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4222727007979498,
+      "grad_norm": 0.4054296612739563,
+      "learning_rate": 0.0005952832167832168,
+      "loss": 4.5307,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4368338284116722,
+      "grad_norm": 0.4788004159927368,
+      "learning_rate": 0.0005951083916083916,
+      "loss": 4.4881,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4513949560253946,
+      "grad_norm": 0.46861621737480164,
+      "learning_rate": 0.0005949335664335664,
+      "loss": 4.4843,
+      "step": 1550
+    },
+    {
+      "epoch": 0.465956083639117,
+      "grad_norm": 0.4255252480506897,
+      "learning_rate": 0.0005947587412587413,
+      "loss": 4.4711,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4805172112528394,
+      "grad_norm": 0.4327350854873657,
+      "learning_rate": 0.0005945839160839161,
+      "loss": 4.4385,
+      "step": 1650
+    },
+    {
+      "epoch": 0.49507833886656183,
+      "grad_norm": 0.42307257652282715,
+      "learning_rate": 0.0005944090909090909,
+      "loss": 4.436,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5096394664802842,
+      "grad_norm": 0.4274670481681824,
+      "learning_rate": 0.0005942342657342657,
+      "loss": 4.4006,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5242005940940067,
+      "grad_norm": 0.42061132192611694,
+      "learning_rate": 0.0005940594405594406,
+      "loss": 4.3802,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5387617217077291,
+      "grad_norm": 0.4404565393924713,
+      "learning_rate": 0.0005938846153846153,
+      "loss": 4.3798,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5533228493214515,
+      "grad_norm": 0.41425594687461853,
+      "learning_rate": 0.0005937097902097902,
+      "loss": 4.3565,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5678839769351739,
+      "grad_norm": 0.39919501543045044,
+      "learning_rate": 0.000593534965034965,
+      "loss": 4.344,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "grad_norm": 0.44774821400642395,
+      "learning_rate": 0.0005933601398601398,
+      "loss": 4.3264,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "eval_accuracy": 0.30012405206330606,
+      "eval_loss": 4.277124404907227,
+      "eval_runtime": 180.5704,
+      "eval_samples_per_second": 92.186,
+      "eval_steps_per_second": 5.765,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5970062321626187,
+      "grad_norm": 0.41692787408828735,
+      "learning_rate": 0.0005931853146853146,
+      "loss": 4.3319,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6115673597763411,
+      "grad_norm": 0.37570783495903015,
+      "learning_rate": 0.0005930104895104895,
+      "loss": 4.3138,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6261284873900634,
+      "grad_norm": 0.415437787771225,
+      "learning_rate": 0.0005928356643356643,
+      "loss": 4.2858,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6406896150037859,
+      "grad_norm": 0.3619239330291748,
+      "learning_rate": 0.0005926608391608391,
+      "loss": 4.2923,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6552507426175083,
+      "grad_norm": 0.4092442989349365,
+      "learning_rate": 0.000592486013986014,
+      "loss": 4.2764,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6698118702312307,
+      "grad_norm": 0.37751713395118713,
+      "learning_rate": 0.0005923111888111888,
+      "loss": 4.2665,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6843729978449531,
+      "grad_norm": 0.41071465611457825,
+      "learning_rate": 0.0005921363636363636,
+      "loss": 4.2555,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6989341254586755,
+      "grad_norm": 0.39890703558921814,
+      "learning_rate": 0.0005919615384615384,
+      "loss": 4.2449,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7134952530723979,
+      "grad_norm": 0.36429277062416077,
+      "learning_rate": 0.0005917867132867133,
+      "loss": 4.2621,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7280563806861203,
+      "grad_norm": 0.375135213136673,
+      "learning_rate": 0.0005916118881118881,
+      "loss": 4.238,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7426175082998427,
+      "grad_norm": 0.3689658045768738,
+      "learning_rate": 0.0005914370629370629,
+      "loss": 4.2122,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7571786359135652,
+      "grad_norm": 0.3870299756526947,
+      "learning_rate": 0.0005912622377622377,
+      "loss": 4.2154,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7717397635272876,
+      "grad_norm": 0.379463255405426,
+      "learning_rate": 0.0005910874125874125,
+      "loss": 4.2048,
+      "step": 2650
+    },
+    {
+      "epoch": 0.78630089114101,
+      "grad_norm": 0.37131428718566895,
+      "learning_rate": 0.0005909125874125873,
+      "loss": 4.1908,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8008620187547324,
+      "grad_norm": 0.37358808517456055,
+      "learning_rate": 0.0005907377622377622,
+      "loss": 4.1776,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8154231463684548,
+      "grad_norm": 0.37713292241096497,
+      "learning_rate": 0.000590562937062937,
+      "loss": 4.1811,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8299842739821772,
+      "grad_norm": 0.3663122355937958,
+      "learning_rate": 0.0005903881118881118,
+      "loss": 4.1753,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8445454015958996,
+      "grad_norm": 0.40076518058776855,
+      "learning_rate": 0.0005902132867132867,
+      "loss": 4.1634,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8591065292096219,
+      "grad_norm": 0.347970187664032,
+      "learning_rate": 0.0005900384615384615,
+      "loss": 4.1627,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "grad_norm": 0.37718185782432556,
+      "learning_rate": 0.0005898636363636363,
+      "loss": 4.1401,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "eval_accuracy": 0.3151520801645312,
+      "eval_loss": 4.096925735473633,
+      "eval_runtime": 180.6379,
+      "eval_samples_per_second": 92.151,
+      "eval_steps_per_second": 5.763,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8882287844370668,
+      "grad_norm": 0.3469873070716858,
+      "learning_rate": 0.0005896888111888111,
+      "loss": 4.1311,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9027899120507892,
+      "grad_norm": 0.3316064476966858,
+      "learning_rate": 0.000589513986013986,
+      "loss": 4.1353,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9173510396645116,
+      "grad_norm": 0.362809956073761,
+      "learning_rate": 0.0005893391608391608,
+      "loss": 4.1216,
+      "step": 3150
+    },
+    {
+      "epoch": 0.931912167278234,
+      "grad_norm": 0.384818971157074,
+      "learning_rate": 0.0005891643356643356,
+      "loss": 4.129,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9464732948919564,
+      "grad_norm": 0.3387335538864136,
+      "learning_rate": 0.0005889895104895104,
+      "loss": 4.1201,
+      "step": 3250
+    },
+    {
+      "epoch": 0.9610344225056788,
+      "grad_norm": 0.33730268478393555,
+      "learning_rate": 0.0005888146853146853,
+      "loss": 4.1139,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9755955501194012,
+      "grad_norm": 0.3745166063308716,
+      "learning_rate": 0.00058863986013986,
+      "loss": 4.0941,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9901566777331237,
+      "grad_norm": 0.35486307740211487,
+      "learning_rate": 0.0005884650349650349,
+      "loss": 4.088,
+      "step": 3400
+    },
+    {
+      "epoch": 1.004659560836391,
+      "grad_norm": 0.36032548546791077,
+      "learning_rate": 0.0005882902097902097,
+      "loss": 4.0745,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0192206884501136,
+      "grad_norm": 0.38221433758735657,
+      "learning_rate": 0.0005881153846153845,
+      "loss": 4.0227,
+      "step": 3500
+    },
+    {
+      "epoch": 1.033781816063836,
+      "grad_norm": 0.35653796792030334,
+      "learning_rate": 0.0005879405594405594,
+      "loss": 4.018,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0483429436775584,
+      "grad_norm": 0.3454371392726898,
+      "learning_rate": 0.0005877657342657342,
+      "loss": 4.0198,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0629040712912807,
+      "grad_norm": 0.3450075387954712,
+      "learning_rate": 0.000587590909090909,
+      "loss": 4.0057,
+      "step": 3650
+    },
+    {
+      "epoch": 1.0774651989050033,
+      "grad_norm": 0.3462330996990204,
+      "learning_rate": 0.0005874160839160838,
+      "loss": 3.9938,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0920263265187256,
+      "grad_norm": 0.35249096155166626,
+      "learning_rate": 0.0005872412587412587,
+      "loss": 4.0156,
+      "step": 3750
+    },
+    {
+      "epoch": 1.106587454132448,
+      "grad_norm": 0.32855677604675293,
+      "learning_rate": 0.0005870664335664335,
+      "loss": 4.013,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1211485817461704,
+      "grad_norm": 0.33263683319091797,
+      "learning_rate": 0.0005868916083916083,
+      "loss": 4.0015,
+      "step": 3850
+    },
+    {
+      "epoch": 1.135709709359893,
+      "grad_norm": 0.3342099189758301,
+      "learning_rate": 0.0005867167832167831,
+      "loss": 4.0121,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1502708369736152,
+      "grad_norm": 0.3328862190246582,
+      "learning_rate": 0.000586541958041958,
+      "loss": 4.0018,
+      "step": 3950
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "grad_norm": 0.32106339931488037,
+      "learning_rate": 0.0005863671328671328,
+      "loss": 3.9957,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "eval_accuracy": 0.32516712112451923,
+      "eval_loss": 3.989675521850586,
+      "eval_runtime": 180.2628,
+      "eval_samples_per_second": 92.343,
+      "eval_steps_per_second": 5.775,
+      "step": 4000
+    },
+    {
+      "epoch": 1.17939309220106,
+      "grad_norm": 0.328213632106781,
+      "learning_rate": 0.0005861923076923076,
+      "loss": 3.9967,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1939542198147826,
+      "grad_norm": 0.32576414942741394,
+      "learning_rate": 0.0005860174825174824,
+      "loss": 3.9813,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2085153474285049,
+      "grad_norm": 0.34450486302375793,
+      "learning_rate": 0.0005858426573426573,
+      "loss": 3.9912,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2230764750422272,
+      "grad_norm": 0.3487413823604584,
+      "learning_rate": 0.000585667832167832,
+      "loss": 3.9733,
+      "step": 4200
+    },
+    {
+      "epoch": 1.2376376026559497,
+      "grad_norm": 0.34708768129348755,
+      "learning_rate": 0.000585493006993007,
+      "loss": 3.9726,
+      "step": 4250
+    },
+    {
+      "epoch": 1.2521987302696722,
+      "grad_norm": 0.3304347097873688,
+      "learning_rate": 0.0005853181818181817,
+      "loss": 3.9653,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2667598578833945,
+      "grad_norm": 0.3297647535800934,
+      "learning_rate": 0.0005851433566433565,
+      "loss": 3.9575,
+      "step": 4350
+    },
+    {
+      "epoch": 1.2813209854971168,
+      "grad_norm": 0.33607807755470276,
+      "learning_rate": 0.0005849685314685315,
+      "loss": 3.97,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2958821131108393,
+      "grad_norm": 0.3388855457305908,
+      "learning_rate": 0.0005847937062937063,
+      "loss": 3.9794,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3104432407245616,
+      "grad_norm": 0.32652390003204346,
+      "learning_rate": 0.0005846188811188811,
+      "loss": 3.9708,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3250043683382842,
+      "grad_norm": 0.3443945050239563,
+      "learning_rate": 0.0005844440559440559,
+      "loss": 3.9656,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3395654959520065,
+      "grad_norm": 0.341253399848938,
+      "learning_rate": 0.0005842692307692308,
+      "loss": 3.9506,
+      "step": 4600
+    },
+    {
+      "epoch": 1.354126623565729,
+      "grad_norm": 0.3474685549736023,
+      "learning_rate": 0.0005840944055944056,
+      "loss": 3.9478,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3686877511794513,
+      "grad_norm": 0.34360435605049133,
+      "learning_rate": 0.0005839195804195804,
+      "loss": 3.9412,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3832488787931738,
+      "grad_norm": 0.35134294629096985,
+      "learning_rate": 0.0005837447552447552,
+      "loss": 3.9533,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3978100064068961,
+      "grad_norm": 0.3539522588253021,
+      "learning_rate": 0.0005835699300699301,
+      "loss": 3.9555,
+      "step": 4800
+    },
+    {
+      "epoch": 1.4123711340206184,
+      "grad_norm": 0.34184563159942627,
+      "learning_rate": 0.0005833951048951048,
+      "loss": 3.9523,
+      "step": 4850
+    },
+    {
+      "epoch": 1.426932261634341,
+      "grad_norm": 0.3456088602542877,
+      "learning_rate": 0.0005832202797202797,
+      "loss": 3.9523,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4414933892480635,
+      "grad_norm": 0.3169538080692291,
+      "learning_rate": 0.0005830454545454546,
+      "loss": 3.933,
+      "step": 4950
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "grad_norm": 0.3226233720779419,
+      "learning_rate": 0.0005828706293706293,
+      "loss": 3.9142,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "eval_accuracy": 0.331768614216658,
+      "eval_loss": 3.915773391723633,
+      "eval_runtime": 180.4519,
+      "eval_samples_per_second": 92.246,
+      "eval_steps_per_second": 5.769,
+      "step": 5000
+    },
+    {
+      "epoch": 1.470615644475508,
+      "grad_norm": 0.3312673568725586,
+      "learning_rate": 0.0005826958041958042,
+      "loss": 3.9325,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4851767720892306,
+      "grad_norm": 0.38458746671676636,
+      "learning_rate": 0.000582520979020979,
+      "loss": 3.9231,
+      "step": 5100
+    },
+    {
+      "epoch": 1.4997378997029531,
+      "grad_norm": 0.3237328827381134,
+      "learning_rate": 0.0005823461538461538,
+      "loss": 3.9247,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5142990273166754,
+      "grad_norm": 0.31413206458091736,
+      "learning_rate": 0.0005821713286713286,
+      "loss": 3.9265,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5288601549303977,
+      "grad_norm": 0.35820019245147705,
+      "learning_rate": 0.0005819965034965035,
+      "loss": 3.9107,
+      "step": 5250
+    },
+    {
+      "epoch": 1.5434212825441203,
+      "grad_norm": 0.3079889416694641,
+      "learning_rate": 0.0005818216783216783,
+      "loss": 3.9191,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5579824101578428,
+      "grad_norm": 0.3229127526283264,
+      "learning_rate": 0.0005816468531468531,
+      "loss": 3.9095,
+      "step": 5350
+    },
+    {
+      "epoch": 1.572543537771565,
+      "grad_norm": 0.3325076997280121,
+      "learning_rate": 0.0005814720279720279,
+      "loss": 3.9086,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5871046653852874,
+      "grad_norm": 0.3287741541862488,
+      "learning_rate": 0.0005812972027972028,
+      "loss": 3.908,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6016657929990097,
+      "grad_norm": 0.31764456629753113,
+      "learning_rate": 0.0005811223776223776,
+      "loss": 3.9184,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6162269206127322,
+      "grad_norm": 0.3241208791732788,
+      "learning_rate": 0.0005809475524475524,
+      "loss": 3.8827,
+      "step": 5550
+    },
+    {
+      "epoch": 1.6307880482264547,
+      "grad_norm": 0.3249201774597168,
+      "learning_rate": 0.0005807727272727272,
+      "loss": 3.9047,
+      "step": 5600
+    },
+    {
+      "epoch": 1.645349175840177,
+      "grad_norm": 0.3344324827194214,
+      "learning_rate": 0.0005805979020979021,
+      "loss": 3.8918,
+      "step": 5650
+    },
+    {
+      "epoch": 1.6599103034538993,
+      "grad_norm": 0.3216317296028137,
+      "learning_rate": 0.0005804230769230769,
+      "loss": 3.8934,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6744714310676219,
+      "grad_norm": 0.3269466161727905,
+      "learning_rate": 0.0005802482517482517,
+      "loss": 3.8934,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6890325586813444,
+      "grad_norm": 0.32984817028045654,
+      "learning_rate": 0.0005800734265734265,
+      "loss": 3.8732,
+      "step": 5800
+    },
+    {
+      "epoch": 1.7035936862950667,
+      "grad_norm": 0.3205852508544922,
+      "learning_rate": 0.0005798986013986013,
+      "loss": 3.8806,
+      "step": 5850
+    },
+    {
+      "epoch": 1.718154813908789,
+      "grad_norm": 0.3284587860107422,
+      "learning_rate": 0.0005797237762237762,
+      "loss": 3.8772,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7327159415225115,
+      "grad_norm": 0.3408912420272827,
+      "learning_rate": 0.000579548951048951,
+      "loss": 3.8816,
+      "step": 5950
+    },
+    {
+      "epoch": 1.747277069136234,
+      "grad_norm": 0.32032084465026855,
+      "learning_rate": 0.0005793741258741258,
+      "loss": 3.8871,
+      "step": 6000
+    },
+    {
+      "epoch": 1.747277069136234,
+      "eval_accuracy": 0.33673681000448386,
+      "eval_loss": 3.857321262359619,
+      "eval_runtime": 180.7738,
+      "eval_samples_per_second": 92.082,
+      "eval_steps_per_second": 5.759,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7618381967499563,
+      "grad_norm": 0.3361666798591614,
+      "learning_rate": 0.0005791993006993006,
+      "loss": 3.8866,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7763993243636786,
+      "grad_norm": 0.3247644305229187,
+      "learning_rate": 0.0005790244755244755,
+      "loss": 3.8842,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7909604519774012,
+      "grad_norm": 0.31813907623291016,
+      "learning_rate": 0.0005788496503496503,
+      "loss": 3.8721,
+      "step": 6150
+    },
+    {
+      "epoch": 1.8055215795911237,
+      "grad_norm": 0.3214803636074066,
+      "learning_rate": 0.0005786748251748251,
+      "loss": 3.8562,
+      "step": 6200
+    },
+    {
+      "epoch": 1.820082707204846,
+      "grad_norm": 0.3097255229949951,
+      "learning_rate": 0.0005784999999999999,
+      "loss": 3.8573,
+      "step": 6250
+    },
+    {
+      "epoch": 1.8346438348185683,
+      "grad_norm": 0.32802045345306396,
+      "learning_rate": 0.0005783251748251748,
+      "loss": 3.8657,
+      "step": 6300
+    },
+    {
+      "epoch": 1.8492049624322906,
+      "grad_norm": 0.3414191007614136,
+      "learning_rate": 0.0005781503496503496,
+      "loss": 3.8632,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8637660900460131,
+      "grad_norm": 0.3283698856830597,
+      "learning_rate": 0.0005779755244755244,
+      "loss": 3.8643,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8783272176597356,
+      "grad_norm": 0.31000998616218567,
+      "learning_rate": 0.0005778006993006993,
+      "loss": 3.861,
+      "step": 6450
+    },
+    {
+      "epoch": 1.892888345273458,
+      "grad_norm": 0.319161593914032,
+      "learning_rate": 0.000577625874125874,
+      "loss": 3.8588,
+      "step": 6500
+    },
+    {
+      "epoch": 1.9074494728871803,
+      "grad_norm": 0.3341123163700104,
+      "learning_rate": 0.0005774510489510489,
+      "loss": 3.8575,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9220106005009028,
+      "grad_norm": 0.3260324001312256,
+      "learning_rate": 0.0005772762237762237,
+      "loss": 3.8503,
+      "step": 6600
+    },
+    {
+      "epoch": 1.9365717281146253,
+      "grad_norm": 0.32724907994270325,
+      "learning_rate": 0.0005771013986013985,
+      "loss": 3.8533,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9511328557283476,
+      "grad_norm": 0.3281218111515045,
+      "learning_rate": 0.0005769265734265733,
+      "loss": 3.8423,
+      "step": 6700
+    },
+    {
+      "epoch": 1.96569398334207,
+      "grad_norm": 0.3258511424064636,
+      "learning_rate": 0.0005767517482517482,
+      "loss": 3.8595,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9802551109557924,
+      "grad_norm": 0.3264133930206299,
+      "learning_rate": 0.000576576923076923,
+      "loss": 3.8518,
+      "step": 6800
+    },
+    {
+      "epoch": 1.994816238569515,
+      "grad_norm": 0.3205277621746063,
+      "learning_rate": 0.0005764020979020978,
+      "loss": 3.8442,
+      "step": 6850
+    },
+    {
+      "epoch": 2.009319121672782,
+      "grad_norm": 0.3255445659160614,
+      "learning_rate": 0.0005762272727272726,
+      "loss": 3.7817,
+      "step": 6900
+    },
+    {
+      "epoch": 2.023880249286505,
+      "grad_norm": 0.3084678649902344,
+      "learning_rate": 0.0005760524475524475,
+      "loss": 3.7401,
+      "step": 6950
+    },
+    {
+      "epoch": 2.038441376900227,
+      "grad_norm": 0.3164021968841553,
+      "learning_rate": 0.0005758776223776223,
+      "loss": 3.7577,
+      "step": 7000
+    },
+    {
+      "epoch": 2.038441376900227,
+      "eval_accuracy": 0.34121018477785253,
+      "eval_loss": 3.815640687942505,
+      "eval_runtime": 180.7874,
+      "eval_samples_per_second": 92.075,
+      "eval_steps_per_second": 5.758,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0530025045139495,
+      "grad_norm": 0.32217466831207275,
+      "learning_rate": 0.0005757027972027971,
+      "loss": 3.7505,
+      "step": 7050
+    },
+    {
+      "epoch": 2.067563632127672,
+      "grad_norm": 0.3208473324775696,
+      "learning_rate": 0.000575527972027972,
+      "loss": 3.7286,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0821247597413945,
+      "grad_norm": 0.3347488045692444,
+      "learning_rate": 0.0005753531468531468,
+      "loss": 3.7528,
+      "step": 7150
+    },
+    {
+      "epoch": 2.096685887355117,
+      "grad_norm": 0.31259018182754517,
+      "learning_rate": 0.0005751783216783216,
+      "loss": 3.7381,
+      "step": 7200
+    },
+    {
+      "epoch": 2.111247014968839,
+      "grad_norm": 0.31018945574760437,
+      "learning_rate": 0.0005750034965034964,
+      "loss": 3.754,
+      "step": 7250
+    },
+    {
+      "epoch": 2.1258081425825615,
+      "grad_norm": 0.3304862380027771,
+      "learning_rate": 0.0005748286713286712,
+      "loss": 3.7545,
+      "step": 7300
+    },
+    {
+      "epoch": 2.140369270196284,
+      "grad_norm": 0.30643802881240845,
+      "learning_rate": 0.000574653846153846,
+      "loss": 3.7614,
+      "step": 7350
+    },
+    {
+      "epoch": 2.1549303978100065,
+      "grad_norm": 0.31393590569496155,
+      "learning_rate": 0.000574479020979021,
+      "loss": 3.7581,
+      "step": 7400
+    },
+    {
+      "epoch": 2.169491525423729,
+      "grad_norm": 0.3010294735431671,
+      "learning_rate": 0.0005743041958041958,
+      "loss": 3.7627,
+      "step": 7450
+    },
+    {
+      "epoch": 2.184052653037451,
+      "grad_norm": 0.31778109073638916,
+      "learning_rate": 0.0005741293706293706,
+      "loss": 3.763,
+      "step": 7500
+    },
+    {
+      "epoch": 2.198613780651174,
+      "grad_norm": 0.314079612493515,
+      "learning_rate": 0.0005739545454545454,
+      "loss": 3.7594,
+      "step": 7550
+    },
+    {
+      "epoch": 2.213174908264896,
+      "grad_norm": 0.3207477629184723,
+      "learning_rate": 0.0005737797202797203,
+      "loss": 3.7588,
+      "step": 7600
+    },
+    {
+      "epoch": 2.2277360358786185,
+      "grad_norm": 0.323979914188385,
+      "learning_rate": 0.0005736048951048951,
+      "loss": 3.7551,
+      "step": 7650
+    },
+    {
+      "epoch": 2.2422971634923408,
+      "grad_norm": 0.32413870096206665,
+      "learning_rate": 0.0005734300699300699,
+      "loss": 3.7636,
+      "step": 7700
+    },
+    {
+      "epoch": 2.256858291106063,
+      "grad_norm": 0.3244387209415436,
+      "learning_rate": 0.0005732552447552448,
+      "loss": 3.7487,
+      "step": 7750
+    },
+    {
+      "epoch": 2.271419418719786,
+      "grad_norm": 0.3057880699634552,
+      "learning_rate": 0.0005730804195804196,
+      "loss": 3.754,
+      "step": 7800
+    },
+    {
+      "epoch": 2.285980546333508,
+      "grad_norm": 0.30727651715278625,
+      "learning_rate": 0.0005729055944055944,
+      "loss": 3.7595,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3005416739472304,
+      "grad_norm": 0.3170273005962372,
+      "learning_rate": 0.0005727307692307692,
+      "loss": 3.7625,
+      "step": 7900
+    },
+    {
+      "epoch": 2.3151028015609527,
+      "grad_norm": 0.32343780994415283,
+      "learning_rate": 0.0005725559440559441,
+      "loss": 3.7416,
+      "step": 7950
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "grad_norm": 0.31494140625,
+      "learning_rate": 0.0005723811188811188,
+      "loss": 3.754,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "eval_accuracy": 0.3443036096658095,
+      "eval_loss": 3.784417152404785,
+      "eval_runtime": 181.0164,
+      "eval_samples_per_second": 91.959,
+      "eval_steps_per_second": 5.751,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3442250567883978,
+      "grad_norm": 0.3380199372768402,
+      "learning_rate": 0.0005722062937062937,
+      "loss": 3.7634,
+      "step": 8050
+    },
+    {
+      "epoch": 2.35878618440212,
+      "grad_norm": 0.31097733974456787,
+      "learning_rate": 0.0005720314685314685,
+      "loss": 3.7577,
+      "step": 8100
+    },
+    {
+      "epoch": 2.3733473120158424,
+      "grad_norm": 0.3224102556705475,
+      "learning_rate": 0.0005718566433566433,
+      "loss": 3.7658,
+      "step": 8150
+    },
+    {
+      "epoch": 2.387908439629565,
+      "grad_norm": 0.30005016922950745,
+      "learning_rate": 0.0005716818181818181,
+      "loss": 3.7533,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4024695672432874,
+      "grad_norm": 0.32545462250709534,
+      "learning_rate": 0.000571506993006993,
+      "loss": 3.7441,
+      "step": 8250
+    },
+    {
+      "epoch": 2.4170306948570097,
+      "grad_norm": 0.3352498710155487,
+      "learning_rate": 0.0005713321678321678,
+      "loss": 3.7494,
+      "step": 8300
+    },
+    {
+      "epoch": 2.431591822470732,
+      "grad_norm": 0.326423317193985,
+      "learning_rate": 0.0005711573426573426,
+      "loss": 3.7618,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4461529500844543,
+      "grad_norm": 0.3356836438179016,
+      "learning_rate": 0.0005709825174825175,
+      "loss": 3.7467,
+      "step": 8400
+    },
+    {
+      "epoch": 2.460714077698177,
+      "grad_norm": 0.3288600742816925,
+      "learning_rate": 0.0005708076923076923,
+      "loss": 3.7519,
+      "step": 8450
+    },
+    {
+      "epoch": 2.4752752053118994,
+      "grad_norm": 0.307220458984375,
+      "learning_rate": 0.0005706328671328671,
+      "loss": 3.7423,
+      "step": 8500
+    },
+    {
+      "epoch": 2.4898363329256217,
+      "grad_norm": 0.33141323924064636,
+      "learning_rate": 0.0005704580419580419,
+      "loss": 3.7437,
+      "step": 8550
+    },
+    {
+      "epoch": 2.5043974605393444,
+      "grad_norm": 0.34977665543556213,
+      "learning_rate": 0.0005702832167832168,
+      "loss": 3.7535,
+      "step": 8600
+    },
+    {
+      "epoch": 2.5189585881530667,
+      "grad_norm": 0.32064738869667053,
+      "learning_rate": 0.0005701083916083916,
+      "loss": 3.7556,
+      "step": 8650
+    },
+    {
+      "epoch": 2.533519715766789,
+      "grad_norm": 0.30487334728240967,
+      "learning_rate": 0.0005699335664335664,
+      "loss": 3.7369,
+      "step": 8700
+    },
+    {
+      "epoch": 2.5480808433805113,
+      "grad_norm": 0.3201334476470947,
+      "learning_rate": 0.0005697587412587412,
+      "loss": 3.7461,
+      "step": 8750
+    },
+    {
+      "epoch": 2.5626419709942336,
+      "grad_norm": 0.2892219126224518,
+      "learning_rate": 0.000569583916083916,
+      "loss": 3.7463,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5772030986079564,
+      "grad_norm": 0.3132726848125458,
+      "learning_rate": 0.0005694090909090908,
+      "loss": 3.7544,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5917642262216787,
+      "grad_norm": 0.30421772599220276,
+      "learning_rate": 0.0005692342657342657,
+      "loss": 3.7395,
+      "step": 8900
+    },
+    {
+      "epoch": 2.606325353835401,
+      "grad_norm": 0.3089979290962219,
+      "learning_rate": 0.0005690594405594405,
+      "loss": 3.7349,
+      "step": 8950
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "grad_norm": 0.3203098177909851,
+      "learning_rate": 0.0005688846153846153,
+      "loss": 3.7424,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "eval_accuracy": 0.3470149560797855,
+      "eval_loss": 3.753990650177002,
+      "eval_runtime": 180.7846,
+      "eval_samples_per_second": 92.076,
+      "eval_steps_per_second": 5.758,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6354476090628456,
+      "grad_norm": 0.31155484914779663,
+      "learning_rate": 0.0005687097902097901,
+      "loss": 3.7402,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6500087366765683,
+      "grad_norm": 0.3341865539550781,
+      "learning_rate": 0.000568534965034965,
+      "loss": 3.7273,
+      "step": 9100
+    },
+    {
+      "epoch": 2.6645698642902906,
+      "grad_norm": 0.31667912006378174,
+      "learning_rate": 0.0005683601398601398,
+      "loss": 3.7479,
+      "step": 9150
+    },
+    {
+      "epoch": 2.679130991904013,
+      "grad_norm": 0.3061210513114929,
+      "learning_rate": 0.0005681853146853146,
+      "loss": 3.7272,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6936921195177357,
+      "grad_norm": 0.317827969789505,
+      "learning_rate": 0.0005680104895104895,
+      "loss": 3.7336,
+      "step": 9250
+    },
+    {
+      "epoch": 2.708253247131458,
+      "grad_norm": 0.3071555197238922,
+      "learning_rate": 0.0005678356643356643,
+      "loss": 3.7312,
+      "step": 9300
+    },
+    {
+      "epoch": 2.7228143747451803,
+      "grad_norm": 0.31722643971443176,
+      "learning_rate": 0.0005676608391608391,
+      "loss": 3.7306,
+      "step": 9350
+    },
+    {
+      "epoch": 2.7373755023589026,
+      "grad_norm": 0.321736216545105,
+      "learning_rate": 0.0005674860139860139,
+      "loss": 3.7327,
+      "step": 9400
+    },
+    {
+      "epoch": 2.751936629972625,
+      "grad_norm": 0.3194282352924347,
+      "learning_rate": 0.0005673111888111888,
+      "loss": 3.7152,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7664977575863476,
+      "grad_norm": 0.3411991000175476,
+      "learning_rate": 0.0005671363636363635,
+      "loss": 3.7404,
+      "step": 9500
+    },
+    {
+      "epoch": 2.78105888520007,
+      "grad_norm": 0.3048510253429413,
+      "learning_rate": 0.0005669615384615384,
+      "loss": 3.7335,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7956200128137922,
+      "grad_norm": 0.30494141578674316,
+      "learning_rate": 0.0005667867132867132,
+      "loss": 3.7217,
+      "step": 9600
+    },
+    {
+      "epoch": 2.8101811404275145,
+      "grad_norm": 0.324116051197052,
+      "learning_rate": 0.000566611888111888,
+      "loss": 3.7275,
+      "step": 9650
+    },
+    {
+      "epoch": 2.824742268041237,
+      "grad_norm": 0.3104328215122223,
+      "learning_rate": 0.0005664370629370628,
+      "loss": 3.7239,
+      "step": 9700
+    },
+    {
+      "epoch": 2.8393033956549596,
+      "grad_norm": 0.3002651035785675,
+      "learning_rate": 0.0005662622377622377,
+      "loss": 3.7217,
+      "step": 9750
+    },
+    {
+      "epoch": 2.853864523268682,
+      "grad_norm": 0.2895483374595642,
+      "learning_rate": 0.0005660874125874125,
+      "loss": 3.7101,
+      "step": 9800
+    },
+    {
+      "epoch": 2.868425650882404,
+      "grad_norm": 0.3242766261100769,
+      "learning_rate": 0.0005659125874125873,
+      "loss": 3.7279,
+      "step": 9850
+    },
+    {
+      "epoch": 2.882986778496127,
+      "grad_norm": 0.32792049646377563,
+      "learning_rate": 0.0005657377622377622,
+      "loss": 3.7205,
+      "step": 9900
+    },
+    {
+      "epoch": 2.8975479061098492,
+      "grad_norm": 0.3201243281364441,
+      "learning_rate": 0.000565562937062937,
+      "loss": 3.7099,
+      "step": 9950
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "grad_norm": 0.3127686083316803,
+      "learning_rate": 0.0005653881118881118,
+      "loss": 3.7116,
+      "step": 10000
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "eval_accuracy": 0.3493557451553037,
+      "eval_loss": 3.727367639541626,
+      "eval_runtime": 181.2018,
+      "eval_samples_per_second": 91.864,
+      "eval_steps_per_second": 5.745,
+      "step": 10000
+    },
+    {
+      "epoch": 2.926670161337294,
+      "grad_norm": 0.29951873421669006,
+      "learning_rate": 0.0005652132867132866,
+      "loss": 3.7143,
+      "step": 10050
+    },
+    {
+      "epoch": 2.941231288951016,
+      "grad_norm": 0.3104478418827057,
+      "learning_rate": 0.0005650384615384615,
+      "loss": 3.722,
+      "step": 10100
+    },
+    {
+      "epoch": 2.955792416564739,
+      "grad_norm": 0.3141760230064392,
+      "learning_rate": 0.0005648636363636363,
+      "loss": 3.7163,
+      "step": 10150
+    },
+    {
+      "epoch": 2.970353544178461,
+      "grad_norm": 0.3073762357234955,
+      "learning_rate": 0.0005646888111888111,
+      "loss": 3.7277,
+      "step": 10200
+    },
+    {
+      "epoch": 2.9849146717921835,
+      "grad_norm": 0.3173540234565735,
+      "learning_rate": 0.000564513986013986,
+      "loss": 3.7239,
+      "step": 10250
+    },
+    {
+      "epoch": 2.9994757994059063,
+      "grad_norm": 0.3042048513889313,
+      "learning_rate": 0.0005643391608391607,
+      "loss": 3.7166,
+      "step": 10300
+    },
+    {
+      "epoch": 3.0139786825091734,
+      "grad_norm": 0.32184624671936035,
+      "learning_rate": 0.0005641643356643355,
+      "loss": 3.6176,
+      "step": 10350
+    },
+    {
+      "epoch": 3.0285398101228957,
+      "grad_norm": 0.30402591824531555,
+      "learning_rate": 0.0005639895104895105,
+      "loss": 3.6089,
+      "step": 10400
+    },
+    {
+      "epoch": 3.0431009377366185,
+      "grad_norm": 0.3132675886154175,
+      "learning_rate": 0.0005638146853146853,
+      "loss": 3.6175,
+      "step": 10450
+    },
+    {
+      "epoch": 3.057662065350341,
+      "grad_norm": 0.308847039937973,
+      "learning_rate": 0.0005636398601398601,
+      "loss": 3.6257,
+      "step": 10500
+    },
+    {
+      "epoch": 3.072223192964063,
+      "grad_norm": 0.3209833800792694,
+      "learning_rate": 0.000563465034965035,
+      "loss": 3.6303,
+      "step": 10550
+    },
+    {
+      "epoch": 3.0867843205777854,
+      "grad_norm": 0.3201095759868622,
+      "learning_rate": 0.0005632902097902098,
+      "loss": 3.6251,
+      "step": 10600
+    },
+    {
+      "epoch": 3.101345448191508,
+      "grad_norm": 0.3169490694999695,
+      "learning_rate": 0.0005631153846153846,
+      "loss": 3.6231,
+      "step": 10650
+    },
+    {
+      "epoch": 3.1159065758052304,
+      "grad_norm": 0.31459590792655945,
+      "learning_rate": 0.0005629405594405594,
+      "loss": 3.6346,
+      "step": 10700
+    },
+    {
+      "epoch": 3.1304677034189528,
+      "grad_norm": 0.3260461390018463,
+      "learning_rate": 0.0005627657342657343,
+      "loss": 3.6224,
+      "step": 10750
+    },
+    {
+      "epoch": 3.145028831032675,
+      "grad_norm": 0.3369269073009491,
+      "learning_rate": 0.0005625909090909091,
+      "loss": 3.6224,
+      "step": 10800
+    },
+    {
+      "epoch": 3.1595899586463974,
+      "grad_norm": 0.3142642676830292,
+      "learning_rate": 0.0005624160839160839,
+      "loss": 3.6332,
+      "step": 10850
+    },
+    {
+      "epoch": 3.17415108626012,
+      "grad_norm": 0.33844444155693054,
+      "learning_rate": 0.0005622412587412587,
+      "loss": 3.6388,
+      "step": 10900
+    },
+    {
+      "epoch": 3.1887122138738424,
+      "grad_norm": 0.3050360083580017,
+      "learning_rate": 0.0005620664335664336,
+      "loss": 3.6383,
+      "step": 10950
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "grad_norm": 0.32840847969055176,
+      "learning_rate": 0.0005618916083916083,
+      "loss": 3.6438,
+      "step": 11000
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "eval_accuracy": 0.35157309349307425,
+      "eval_loss": 3.71468186378479,
+      "eval_runtime": 181.9058,
+      "eval_samples_per_second": 91.509,
+      "eval_steps_per_second": 5.723,
+      "step": 11000
+    },
+    {
+      "epoch": 3.217834469101287,
+      "grad_norm": 0.31019338965415955,
+      "learning_rate": 0.0005617167832167832,
+      "loss": 3.6341,
+      "step": 11050
+    },
+    {
+      "epoch": 3.2323955967150098,
+      "grad_norm": 0.3126959204673767,
+      "learning_rate": 0.000561541958041958,
+      "loss": 3.6271,
+      "step": 11100
+    },
+    {
+      "epoch": 3.246956724328732,
+      "grad_norm": 0.311865895986557,
+      "learning_rate": 0.0005613671328671328,
+      "loss": 3.6429,
+      "step": 11150
+    },
+    {
+      "epoch": 3.2615178519424544,
+      "grad_norm": 0.35393866896629333,
+      "learning_rate": 0.0005611923076923077,
+      "loss": 3.6427,
+      "step": 11200
+    },
+    {
+      "epoch": 3.2760789795561767,
+      "grad_norm": 0.314378559589386,
+      "learning_rate": 0.0005610174825174825,
+      "loss": 3.6465,
+      "step": 11250
+    },
+    {
+      "epoch": 3.2906401071698994,
+      "grad_norm": 0.3106997013092041,
+      "learning_rate": 0.0005608426573426573,
+      "loss": 3.6439,
+      "step": 11300
+    },
+    {
+      "epoch": 3.3052012347836217,
+      "grad_norm": 0.31017565727233887,
+      "learning_rate": 0.0005606678321678321,
+      "loss": 3.6422,
+      "step": 11350
+    },
+    {
+      "epoch": 3.319762362397344,
+      "grad_norm": 0.297059565782547,
+      "learning_rate": 0.000560493006993007,
+      "loss": 3.6435,
+      "step": 11400
+    },
+    {
+      "epoch": 3.3343234900110663,
+      "grad_norm": 0.3107824921607971,
+      "learning_rate": 0.0005603181818181818,
+      "loss": 3.6471,
+      "step": 11450
+    },
+    {
+      "epoch": 3.3488846176247886,
+      "grad_norm": 0.3141351342201233,
+      "learning_rate": 0.0005601433566433566,
+      "loss": 3.6388,
+      "step": 11500
+    },
+    {
+      "epoch": 3.3634457452385114,
+      "grad_norm": 0.32414788007736206,
+      "learning_rate": 0.0005599685314685314,
+      "loss": 3.6296,
+      "step": 11550
+    },
+    {
+      "epoch": 3.3780068728522337,
+      "grad_norm": 0.32148855924606323,
+      "learning_rate": 0.0005597937062937063,
+      "loss": 3.6379,
+      "step": 11600
+    },
+    {
+      "epoch": 3.392568000465956,
+      "grad_norm": 0.3060270845890045,
+      "learning_rate": 0.0005596188811188811,
+      "loss": 3.653,
+      "step": 11650
+    },
+    {
+      "epoch": 3.4071291280796787,
+      "grad_norm": 0.3094375431537628,
+      "learning_rate": 0.0005594440559440559,
+      "loss": 3.6363,
+      "step": 11700
+    },
+    {
+      "epoch": 3.421690255693401,
+      "grad_norm": 0.3291188180446625,
+      "learning_rate": 0.0005592692307692307,
+      "loss": 3.6504,
+      "step": 11750
+    },
+    {
+      "epoch": 3.4362513833071233,
+      "grad_norm": 0.31658145785331726,
+      "learning_rate": 0.0005590944055944055,
+      "loss": 3.6477,
+      "step": 11800
+    },
+    {
+      "epoch": 3.4508125109208456,
+      "grad_norm": 0.31207045912742615,
+      "learning_rate": 0.0005589195804195803,
+      "loss": 3.6462,
+      "step": 11850
+    },
+    {
+      "epoch": 3.465373638534568,
+      "grad_norm": 0.3267686367034912,
+      "learning_rate": 0.0005587447552447552,
+      "loss": 3.6405,
+      "step": 11900
+    },
+    {
+      "epoch": 3.4799347661482907,
+      "grad_norm": 0.3283142149448395,
+      "learning_rate": 0.00055856993006993,
+      "loss": 3.6475,
+      "step": 11950
+    },
+    {
+      "epoch": 3.494495893762013,
+      "grad_norm": 0.32450154423713684,
+      "learning_rate": 0.0005583951048951048,
+      "loss": 3.6323,
+      "step": 12000
+    },
+    {
+      "epoch": 3.494495893762013,
+      "eval_accuracy": 0.35301782037515167,
+      "eval_loss": 3.6957285404205322,
+      "eval_runtime": 180.5967,
+      "eval_samples_per_second": 92.172,
+      "eval_steps_per_second": 5.764,
+      "step": 12000
+    },
+    {
+      "epoch": 3.5090570213757353,
+      "grad_norm": 0.3060498833656311,
+      "learning_rate": 0.0005582202797202797,
+      "loss": 3.6332,
+      "step": 12050
+    },
+    {
+      "epoch": 3.523618148989458,
+      "grad_norm": 0.3046874403953552,
+      "learning_rate": 0.0005580454545454545,
+      "loss": 3.6427,
+      "step": 12100
+    },
+    {
+      "epoch": 3.53817927660318,
+      "grad_norm": 0.3221018314361572,
+      "learning_rate": 0.0005578706293706293,
+      "loss": 3.6475,
+      "step": 12150
+    },
+    {
+      "epoch": 3.5527404042169026,
+      "grad_norm": 0.33598005771636963,
+      "learning_rate": 0.0005576958041958041,
+      "loss": 3.6484,
+      "step": 12200
+    },
+    {
+      "epoch": 3.567301531830625,
+      "grad_norm": 0.3117700219154358,
+      "learning_rate": 0.000557520979020979,
+      "loss": 3.6374,
+      "step": 12250
+    },
+    {
+      "epoch": 3.5818626594443472,
+      "grad_norm": 0.3346507251262665,
+      "learning_rate": 0.0005573461538461538,
+      "loss": 3.6269,
+      "step": 12300
+    },
+    {
+      "epoch": 3.59642378705807,
+      "grad_norm": 0.3148074448108673,
+      "learning_rate": 0.0005571713286713286,
+      "loss": 3.6348,
+      "step": 12350
+    },
+    {
+      "epoch": 3.6109849146717923,
+      "grad_norm": 0.30282357335090637,
+      "learning_rate": 0.0005569965034965034,
+      "loss": 3.6468,
+      "step": 12400
+    },
+    {
+      "epoch": 3.6255460422855146,
+      "grad_norm": 0.306007981300354,
+      "learning_rate": 0.0005568216783216783,
+      "loss": 3.6379,
+      "step": 12450
+    },
+    {
+      "epoch": 3.640107169899237,
+      "grad_norm": 0.3162916600704193,
+      "learning_rate": 0.000556646853146853,
+      "loss": 3.6354,
+      "step": 12500
+    },
+    {
+      "epoch": 3.654668297512959,
+      "grad_norm": 0.31878116726875305,
+      "learning_rate": 0.0005564720279720279,
+      "loss": 3.6363,
+      "step": 12550
+    },
+    {
+      "epoch": 3.669229425126682,
+      "grad_norm": 0.3057233691215515,
+      "learning_rate": 0.0005562972027972027,
+      "loss": 3.6419,
+      "step": 12600
+    },
+    {
+      "epoch": 3.6837905527404042,
+      "grad_norm": 0.33499786257743835,
+      "learning_rate": 0.0005561223776223775,
+      "loss": 3.6456,
+      "step": 12650
+    },
+    {
+      "epoch": 3.6983516803541265,
+      "grad_norm": 0.3105752170085907,
+      "learning_rate": 0.0005559475524475524,
+      "loss": 3.6356,
+      "step": 12700
+    },
+    {
+      "epoch": 3.7129128079678493,
+      "grad_norm": 0.3092792332172394,
+      "learning_rate": 0.0005557727272727272,
+      "loss": 3.6494,
+      "step": 12750
+    },
+    {
+      "epoch": 3.7274739355815716,
+      "grad_norm": 0.3000400960445404,
+      "learning_rate": 0.000555597902097902,
+      "loss": 3.6326,
+      "step": 12800
+    },
+    {
+      "epoch": 3.742035063195294,
+      "grad_norm": 0.3095425069332123,
+      "learning_rate": 0.0005554230769230768,
+      "loss": 3.6408,
+      "step": 12850
+    },
+    {
+      "epoch": 3.756596190809016,
+      "grad_norm": 0.3014494478702545,
+      "learning_rate": 0.0005552482517482517,
+      "loss": 3.6444,
+      "step": 12900
+    },
+    {
+      "epoch": 3.7711573184227385,
+      "grad_norm": 0.3192651569843292,
+      "learning_rate": 0.0005550734265734265,
+      "loss": 3.6436,
+      "step": 12950
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "grad_norm": 0.32174089550971985,
+      "learning_rate": 0.0005548986013986013,
+      "loss": 3.6429,
+      "step": 13000
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "eval_accuracy": 0.35489482496456076,
+      "eval_loss": 3.6800663471221924,
+      "eval_runtime": 180.4037,
+      "eval_samples_per_second": 92.271,
+      "eval_steps_per_second": 5.77,
+      "step": 13000
+    },
+    {
+      "epoch": 3.8002795736501835,
+      "grad_norm": 0.31617510318756104,
+      "learning_rate": 0.0005547237762237761,
+      "loss": 3.6465,
+      "step": 13050
+    },
+    {
+      "epoch": 3.814840701263906,
+      "grad_norm": 0.34271636605262756,
+      "learning_rate": 0.000554548951048951,
+      "loss": 3.6347,
+      "step": 13100
+    },
+    {
+      "epoch": 3.829401828877628,
+      "grad_norm": 0.3209848999977112,
+      "learning_rate": 0.0005543741258741258,
+      "loss": 3.6414,
+      "step": 13150
+    },
+    {
+      "epoch": 3.8439629564913504,
+      "grad_norm": 0.33069539070129395,
+      "learning_rate": 0.0005541993006993006,
+      "loss": 3.6252,
+      "step": 13200
+    },
+    {
+      "epoch": 3.858524084105073,
+      "grad_norm": 0.3266202211380005,
+      "learning_rate": 0.0005540244755244756,
+      "loss": 3.6354,
+      "step": 13250
+    },
+    {
+      "epoch": 3.8730852117187955,
+      "grad_norm": 0.3188350796699524,
+      "learning_rate": 0.0005538496503496502,
+      "loss": 3.6536,
+      "step": 13300
+    },
+    {
+      "epoch": 3.887646339332518,
+      "grad_norm": 0.3059855103492737,
+      "learning_rate": 0.0005536748251748252,
+      "loss": 3.6327,
+      "step": 13350
+    },
+    {
+      "epoch": 3.9022074669462405,
+      "grad_norm": 0.31143465638160706,
+      "learning_rate": 0.0005535,
+      "loss": 3.6204,
+      "step": 13400
+    },
+    {
+      "epoch": 3.916768594559963,
+      "grad_norm": 0.3238065838813782,
+      "learning_rate": 0.0005533251748251748,
+      "loss": 3.6292,
+      "step": 13450
+    },
+    {
+      "epoch": 3.931329722173685,
+      "grad_norm": 0.2959144115447998,
+      "learning_rate": 0.0005531503496503496,
+      "loss": 3.6453,
+      "step": 13500
+    },
+    {
+      "epoch": 3.9458908497874075,
+      "grad_norm": 0.32706987857818604,
+      "learning_rate": 0.0005529755244755245,
+      "loss": 3.6269,
+      "step": 13550
+    },
+    {
+      "epoch": 3.9604519774011298,
+      "grad_norm": 0.3075656294822693,
+      "learning_rate": 0.0005528006993006993,
+      "loss": 3.6376,
+      "step": 13600
+    },
+    {
+      "epoch": 3.9750131050148525,
+      "grad_norm": 0.3198637366294861,
+      "learning_rate": 0.0005526258741258741,
+      "loss": 3.6372,
+      "step": 13650
+    },
+    {
+      "epoch": 3.989574232628575,
+      "grad_norm": 0.3145425319671631,
+      "learning_rate": 0.0005524510489510489,
+      "loss": 3.6378,
+      "step": 13700
+    },
+    {
+      "epoch": 4.004077115731842,
+      "grad_norm": 0.31634321808815,
+      "learning_rate": 0.0005522762237762238,
+      "loss": 3.6062,
+      "step": 13750
+    },
+    {
+      "epoch": 4.018638243345564,
+      "grad_norm": 0.3056029677391052,
+      "learning_rate": 0.0005521013986013986,
+      "loss": 3.5304,
+      "step": 13800
+    },
+    {
+      "epoch": 4.033199370959287,
+      "grad_norm": 0.29673531651496887,
+      "learning_rate": 0.0005519265734265734,
+      "loss": 3.5444,
+      "step": 13850
+    },
+    {
+      "epoch": 4.04776049857301,
+      "grad_norm": 0.32092323899269104,
+      "learning_rate": 0.0005517517482517482,
+      "loss": 3.5251,
+      "step": 13900
+    },
+    {
+      "epoch": 4.062321626186732,
+      "grad_norm": 0.32128646969795227,
+      "learning_rate": 0.0005515769230769231,
+      "loss": 3.5352,
+      "step": 13950
+    },
+    {
+      "epoch": 4.076882753800454,
+      "grad_norm": 0.3217810094356537,
+      "learning_rate": 0.0005514020979020979,
+      "loss": 3.5448,
+      "step": 14000
+    },
+    {
+      "epoch": 4.076882753800454,
+      "eval_accuracy": 0.3563979804625054,
+      "eval_loss": 3.6696760654449463,
+      "eval_runtime": 180.6933,
+      "eval_samples_per_second": 92.123,
+      "eval_steps_per_second": 5.761,
+      "step": 14000
+    },
+    {
+      "epoch": 4.091443881414177,
+      "grad_norm": 0.320950984954834,
+      "learning_rate": 0.0005512272727272727,
+      "loss": 3.5395,
+      "step": 14050
+    },
+    {
+      "epoch": 4.106005009027899,
+      "grad_norm": 0.3130992650985718,
+      "learning_rate": 0.0005510524475524475,
+      "loss": 3.5323,
+      "step": 14100
+    },
+    {
+      "epoch": 4.120566136641622,
+      "grad_norm": 0.32641854882240295,
+      "learning_rate": 0.0005508776223776223,
+      "loss": 3.5339,
+      "step": 14150
+    },
+    {
+      "epoch": 4.135127264255344,
+      "grad_norm": 0.3011496961116791,
+      "learning_rate": 0.0005507027972027972,
+      "loss": 3.5519,
+      "step": 14200
+    },
+    {
+      "epoch": 4.149688391869066,
+      "grad_norm": 0.31833386421203613,
+      "learning_rate": 0.000550527972027972,
+      "loss": 3.5554,
+      "step": 14250
+    },
+    {
+      "epoch": 4.164249519482789,
+      "grad_norm": 0.3174136281013489,
+      "learning_rate": 0.0005503531468531468,
+      "loss": 3.5548,
+      "step": 14300
+    },
+    {
+      "epoch": 4.178810647096511,
+      "grad_norm": 0.30939123034477234,
+      "learning_rate": 0.0005501783216783216,
+      "loss": 3.5442,
+      "step": 14350
+    },
+    {
+      "epoch": 4.193371774710234,
+      "grad_norm": 0.31899625062942505,
+      "learning_rate": 0.0005500034965034965,
+      "loss": 3.5547,
+      "step": 14400
+    },
+    {
+      "epoch": 4.207932902323956,
+      "grad_norm": 0.3191762864589691,
+      "learning_rate": 0.0005498286713286713,
+      "loss": 3.5635,
+      "step": 14450
+    },
+    {
+      "epoch": 4.222494029937678,
+      "grad_norm": 0.32480305433273315,
+      "learning_rate": 0.0005496538461538461,
+      "loss": 3.5434,
+      "step": 14500
+    },
+    {
+      "epoch": 4.237055157551401,
+      "grad_norm": 0.32042625546455383,
+      "learning_rate": 0.0005494790209790209,
+      "loss": 3.5771,
+      "step": 14550
+    },
+    {
+      "epoch": 4.251616285165123,
+      "grad_norm": 0.34208381175994873,
+      "learning_rate": 0.0005493041958041958,
+      "loss": 3.5499,
+      "step": 14600
+    },
+    {
+      "epoch": 4.266177412778846,
+      "grad_norm": 0.29826879501342773,
+      "learning_rate": 0.0005491293706293706,
+      "loss": 3.554,
+      "step": 14650
+    },
+    {
+      "epoch": 4.280738540392568,
+      "grad_norm": 0.2985898554325104,
+      "learning_rate": 0.0005489545454545454,
+      "loss": 3.5696,
+      "step": 14700
+    },
+    {
+      "epoch": 4.29529966800629,
+      "grad_norm": 0.3151024878025055,
+      "learning_rate": 0.0005487797202797203,
+      "loss": 3.5591,
+      "step": 14750
+    },
+    {
+      "epoch": 4.309860795620013,
+      "grad_norm": 0.31103193759918213,
+      "learning_rate": 0.000548604895104895,
+      "loss": 3.5667,
+      "step": 14800
+    },
+    {
+      "epoch": 4.324421923233735,
+      "grad_norm": 0.30722710490226746,
+      "learning_rate": 0.0005484300699300699,
+      "loss": 3.5591,
+      "step": 14850
+    },
+    {
+      "epoch": 4.338983050847458,
+      "grad_norm": 0.32191506028175354,
+      "learning_rate": 0.0005482552447552447,
+      "loss": 3.5589,
+      "step": 14900
+    },
+    {
+      "epoch": 4.35354417846118,
+      "grad_norm": 0.30484798550605774,
+      "learning_rate": 0.0005480804195804195,
+      "loss": 3.5703,
+      "step": 14950
+    },
+    {
+      "epoch": 4.368105306074902,
+      "grad_norm": 0.3188960552215576,
+      "learning_rate": 0.0005479055944055943,
+      "loss": 3.5583,
+      "step": 15000
+    },
+    {
+      "epoch": 4.368105306074902,
+      "eval_accuracy": 0.35716825066605096,
+      "eval_loss": 3.6608121395111084,
+      "eval_runtime": 180.9462,
+      "eval_samples_per_second": 91.994,
+      "eval_steps_per_second": 5.753,
+      "step": 15000
+    },
+    {
+      "epoch": 4.382666433688625,
+      "grad_norm": 0.30769604444503784,
+      "learning_rate": 0.0005477307692307692,
+      "loss": 3.5692,
+      "step": 15050
+    },
+    {
+      "epoch": 4.397227561302348,
+      "grad_norm": 0.3089370131492615,
+      "learning_rate": 0.000547555944055944,
+      "loss": 3.5697,
+      "step": 15100
+    },
+    {
+      "epoch": 4.41178868891607,
+      "grad_norm": 0.2947642207145691,
+      "learning_rate": 0.0005473811188811188,
+      "loss": 3.5753,
+      "step": 15150
+    },
+    {
+      "epoch": 4.426349816529792,
+      "grad_norm": 0.30472230911254883,
+      "learning_rate": 0.0005472062937062936,
+      "loss": 3.5844,
+      "step": 15200
+    },
+    {
+      "epoch": 4.440910944143514,
+      "grad_norm": 0.31481656432151794,
+      "learning_rate": 0.0005470314685314685,
+      "loss": 3.5755,
+      "step": 15250
+    },
+    {
+      "epoch": 4.455472071757237,
+      "grad_norm": 0.29861533641815186,
+      "learning_rate": 0.0005468566433566433,
+      "loss": 3.5762,
+      "step": 15300
+    },
+    {
+      "epoch": 4.47003319937096,
+      "grad_norm": 0.3105551600456238,
+      "learning_rate": 0.0005466818181818181,
+      "loss": 3.5571,
+      "step": 15350
+    },
+    {
+      "epoch": 4.4845943269846815,
+      "grad_norm": 0.318051278591156,
+      "learning_rate": 0.000546506993006993,
+      "loss": 3.579,
+      "step": 15400
+    },
+    {
+      "epoch": 4.499155454598404,
+      "grad_norm": 0.3215668797492981,
+      "learning_rate": 0.0005463321678321678,
+      "loss": 3.5776,
+      "step": 15450
+    },
+    {
+      "epoch": 4.513716582212126,
+      "grad_norm": 0.3257148861885071,
+      "learning_rate": 0.0005461573426573426,
+      "loss": 3.5733,
+      "step": 15500
+    },
+    {
+      "epoch": 4.528277709825849,
+      "grad_norm": 0.3296310603618622,
+      "learning_rate": 0.0005459825174825174,
+      "loss": 3.57,
+      "step": 15550
+    },
+    {
+      "epoch": 4.542838837439572,
+      "grad_norm": 0.349429726600647,
+      "learning_rate": 0.0005458076923076922,
+      "loss": 3.5644,
+      "step": 15600
+    },
+    {
+      "epoch": 4.5573999650532935,
+      "grad_norm": 0.3020808696746826,
+      "learning_rate": 0.000545632867132867,
+      "loss": 3.5642,
+      "step": 15650
+    },
+    {
+      "epoch": 4.571961092667016,
+      "grad_norm": 0.3027339577674866,
+      "learning_rate": 0.0005454580419580419,
+      "loss": 3.5719,
+      "step": 15700
+    },
+    {
+      "epoch": 4.586522220280738,
+      "grad_norm": 0.32000523805618286,
+      "learning_rate": 0.0005452832167832167,
+      "loss": 3.5646,
+      "step": 15750
+    },
+    {
+      "epoch": 4.601083347894461,
+      "grad_norm": 0.32438185811042786,
+      "learning_rate": 0.0005451083916083915,
+      "loss": 3.5665,
+      "step": 15800
+    },
+    {
+      "epoch": 4.615644475508184,
+      "grad_norm": 0.3186527192592621,
+      "learning_rate": 0.0005449335664335663,
+      "loss": 3.5831,
+      "step": 15850
+    },
+    {
+      "epoch": 4.630205603121905,
+      "grad_norm": 0.31081655621528625,
+      "learning_rate": 0.0005447587412587412,
+      "loss": 3.5725,
+      "step": 15900
+    },
+    {
+      "epoch": 4.644766730735628,
+      "grad_norm": 0.3225807547569275,
+      "learning_rate": 0.000544583916083916,
+      "loss": 3.5615,
+      "step": 15950
+    },
+    {
+      "epoch": 4.659327858349351,
+      "grad_norm": 0.30369243025779724,
+      "learning_rate": 0.0005444090909090908,
+      "loss": 3.5833,
+      "step": 16000
+    },
+    {
+      "epoch": 4.659327858349351,
+      "eval_accuracy": 0.358389373468894,
+      "eval_loss": 3.6469626426696777,
+      "eval_runtime": 180.7892,
+      "eval_samples_per_second": 92.074,
+      "eval_steps_per_second": 5.758,
+      "step": 16000
+    },
+    {
+      "epoch": 4.673888985963073,
+      "grad_norm": 0.31460005044937134,
+      "learning_rate": 0.0005442342657342657,
+      "loss": 3.5811,
+      "step": 16050
+    },
+    {
+      "epoch": 4.6884501135767955,
+      "grad_norm": 0.3051797151565552,
+      "learning_rate": 0.0005440594405594405,
+      "loss": 3.5772,
+      "step": 16100
+    },
+    {
+      "epoch": 4.703011241190518,
+      "grad_norm": 0.3245795667171478,
+      "learning_rate": 0.0005438846153846153,
+      "loss": 3.5612,
+      "step": 16150
+    },
+    {
+      "epoch": 4.71757236880424,
+      "grad_norm": 0.30969762802124023,
+      "learning_rate": 0.0005437097902097901,
+      "loss": 3.5658,
+      "step": 16200
+    },
+    {
+      "epoch": 4.732133496417963,
+      "grad_norm": 0.31821149587631226,
+      "learning_rate": 0.0005435349650349651,
+      "loss": 3.5825,
+      "step": 16250
+    },
+    {
+      "epoch": 4.746694624031685,
+      "grad_norm": 0.3063112199306488,
+      "learning_rate": 0.0005433601398601397,
+      "loss": 3.5774,
+      "step": 16300
+    },
+    {
+      "epoch": 4.7612557516454075,
+      "grad_norm": 0.31970885396003723,
+      "learning_rate": 0.0005431853146853147,
+      "loss": 3.5815,
+      "step": 16350
+    },
+    {
+      "epoch": 4.77581687925913,
+      "grad_norm": 0.3243388235569,
+      "learning_rate": 0.0005430104895104895,
+      "loss": 3.588,
+      "step": 16400
+    },
+    {
+      "epoch": 4.790378006872852,
+      "grad_norm": 0.30511829257011414,
+      "learning_rate": 0.0005428356643356643,
+      "loss": 3.5775,
+      "step": 16450
+    },
+    {
+      "epoch": 4.804939134486575,
+      "grad_norm": 0.3213348984718323,
+      "learning_rate": 0.0005426608391608391,
+      "loss": 3.5811,
+      "step": 16500
+    },
+    {
+      "epoch": 4.819500262100297,
+      "grad_norm": 0.3377111554145813,
+      "learning_rate": 0.000542486013986014,
+      "loss": 3.5669,
+      "step": 16550
+    },
+    {
+      "epoch": 4.834061389714019,
+      "grad_norm": 0.3088950216770172,
+      "learning_rate": 0.0005423111888111888,
+      "loss": 3.5772,
+      "step": 16600
+    },
+    {
+      "epoch": 4.848622517327742,
+      "grad_norm": 0.30557340383529663,
+      "learning_rate": 0.0005421363636363636,
+      "loss": 3.5807,
+      "step": 16650
+    },
+    {
+      "epoch": 4.863183644941464,
+      "grad_norm": 0.3154135048389435,
+      "learning_rate": 0.0005419615384615385,
+      "loss": 3.5719,
+      "step": 16700
+    },
+    {
+      "epoch": 4.877744772555187,
+      "grad_norm": 0.31455299258232117,
+      "learning_rate": 0.0005417867132867133,
+      "loss": 3.5717,
+      "step": 16750
+    },
+    {
+      "epoch": 4.892305900168909,
+      "grad_norm": 0.31194445490837097,
+      "learning_rate": 0.0005416118881118881,
+      "loss": 3.5834,
+      "step": 16800
+    },
+    {
+      "epoch": 4.906867027782631,
+      "grad_norm": 0.31651851534843445,
+      "learning_rate": 0.0005414370629370629,
+      "loss": 3.5798,
+      "step": 16850
+    },
+    {
+      "epoch": 4.921428155396354,
+      "grad_norm": 0.3114611506462097,
+      "learning_rate": 0.0005412622377622378,
+      "loss": 3.5846,
+      "step": 16900
+    },
+    {
+      "epoch": 4.935989283010076,
+      "grad_norm": 0.3295888602733612,
+      "learning_rate": 0.0005410874125874126,
+      "loss": 3.5719,
+      "step": 16950
+    },
+    {
+      "epoch": 4.950550410623799,
+      "grad_norm": 0.3076690137386322,
+      "learning_rate": 0.0005409125874125874,
+      "loss": 3.5773,
+      "step": 17000
+    },
+    {
+      "epoch": 4.950550410623799,
+      "eval_accuracy": 0.3596430611139809,
+      "eval_loss": 3.63307523727417,
+      "eval_runtime": 180.6842,
+      "eval_samples_per_second": 92.128,
+      "eval_steps_per_second": 5.761,
+      "step": 17000
+    },
+    {
+      "epoch": 4.9651115382375215,
+      "grad_norm": 0.3147895038127899,
+      "learning_rate": 0.0005407377622377622,
+      "loss": 3.5766,
+      "step": 17050
+    },
+    {
+      "epoch": 4.979672665851243,
+      "grad_norm": 0.3255283832550049,
+      "learning_rate": 0.000540562937062937,
+      "loss": 3.5795,
+      "step": 17100
+    },
+    {
+      "epoch": 4.994233793464966,
+      "grad_norm": 0.31261691451072693,
+      "learning_rate": 0.0005403881118881118,
+      "loss": 3.5825,
+      "step": 17150
+    },
+    {
+      "epoch": 5.008736676568233,
+      "grad_norm": 0.31291091442108154,
+      "learning_rate": 0.0005402132867132867,
+      "loss": 3.4917,
+      "step": 17200
+    },
+    {
+      "epoch": 5.023297804181956,
+      "grad_norm": 0.32165074348449707,
+      "learning_rate": 0.0005400384615384615,
+      "loss": 3.4626,
+      "step": 17250
+    },
+    {
+      "epoch": 5.037858931795678,
+      "grad_norm": 0.33376818895339966,
+      "learning_rate": 0.0005398636363636363,
+      "loss": 3.4814,
+      "step": 17300
+    },
+    {
+      "epoch": 5.052420059409401,
+      "grad_norm": 0.3262355625629425,
+      "learning_rate": 0.0005396888111888111,
+      "loss": 3.4666,
+      "step": 17350
+    },
+    {
+      "epoch": 5.066981187023123,
+      "grad_norm": 0.33096253871917725,
+      "learning_rate": 0.000539513986013986,
+      "loss": 3.4784,
+      "step": 17400
+    },
+    {
+      "epoch": 5.081542314636845,
+      "grad_norm": 0.3339770436286926,
+      "learning_rate": 0.0005393391608391608,
+      "loss": 3.4858,
+      "step": 17450
+    },
+    {
+      "epoch": 5.096103442250568,
+      "grad_norm": 0.31180301308631897,
+      "learning_rate": 0.0005391643356643356,
+      "loss": 3.4845,
+      "step": 17500
+    },
+    {
+      "epoch": 5.110664569864291,
+      "grad_norm": 0.3266642987728119,
+      "learning_rate": 0.0005389895104895105,
+      "loss": 3.4799,
+      "step": 17550
+    },
+    {
+      "epoch": 5.125225697478013,
+      "grad_norm": 0.3353632092475891,
+      "learning_rate": 0.0005388146853146853,
+      "loss": 3.4778,
+      "step": 17600
+    },
+    {
+      "epoch": 5.139786825091735,
+      "grad_norm": 0.3286963999271393,
+      "learning_rate": 0.0005386398601398601,
+      "loss": 3.4885,
+      "step": 17650
+    },
+    {
+      "epoch": 5.154347952705457,
+      "grad_norm": 0.3285607695579529,
+      "learning_rate": 0.0005384650349650349,
+      "loss": 3.4878,
+      "step": 17700
+    },
+    {
+      "epoch": 5.16890908031918,
+      "grad_norm": 0.32417571544647217,
+      "learning_rate": 0.0005382902097902098,
+      "loss": 3.486,
+      "step": 17750
+    },
+    {
+      "epoch": 5.183470207932903,
+      "grad_norm": 0.3437209129333496,
+      "learning_rate": 0.0005381153846153845,
+      "loss": 3.5001,
+      "step": 17800
+    },
+    {
+      "epoch": 5.1980313355466246,
+      "grad_norm": 0.3151189088821411,
+      "learning_rate": 0.0005379405594405594,
+      "loss": 3.4985,
+      "step": 17850
+    },
+    {
+      "epoch": 5.212592463160347,
+      "grad_norm": 0.30223751068115234,
+      "learning_rate": 0.0005377657342657342,
+      "loss": 3.5054,
+      "step": 17900
+    },
+    {
+      "epoch": 5.227153590774069,
+      "grad_norm": 0.334331214427948,
+      "learning_rate": 0.000537590909090909,
+      "loss": 3.4948,
+      "step": 17950
+    },
+    {
+      "epoch": 5.241714718387792,
+      "grad_norm": 0.33861103653907776,
+      "learning_rate": 0.0005374160839160838,
+      "loss": 3.4997,
+      "step": 18000
+    },
+    {
+      "epoch": 5.241714718387792,
+      "eval_accuracy": 0.35962095934379373,
+      "eval_loss": 3.6378488540649414,
+      "eval_runtime": 180.5829,
+      "eval_samples_per_second": 92.179,
+      "eval_steps_per_second": 5.765,
+      "step": 18000
+    },
+    {
+      "epoch": 5.256275846001515,
+      "grad_norm": 0.3312193751335144,
+      "learning_rate": 0.0005372412587412587,
+      "loss": 3.51,
+      "step": 18050
+    },
+    {
+      "epoch": 5.2708369736152365,
+      "grad_norm": 0.3368973731994629,
+      "learning_rate": 0.0005370664335664335,
+      "loss": 3.5077,
+      "step": 18100
+    },
+    {
+      "epoch": 5.285398101228959,
+      "grad_norm": 0.3362269997596741,
+      "learning_rate": 0.0005368916083916083,
+      "loss": 3.505,
+      "step": 18150
+    },
+    {
+      "epoch": 5.299959228842681,
+      "grad_norm": 0.3283269703388214,
+      "learning_rate": 0.0005367167832167832,
+      "loss": 3.5147,
+      "step": 18200
+    },
+    {
+      "epoch": 5.314520356456404,
+      "grad_norm": 0.33599990606307983,
+      "learning_rate": 0.000536541958041958,
+      "loss": 3.5038,
+      "step": 18250
+    },
+    {
+      "epoch": 5.329081484070127,
+      "grad_norm": 0.325891375541687,
+      "learning_rate": 0.0005363671328671328,
+      "loss": 3.5007,
+      "step": 18300
+    },
+    {
+      "epoch": 5.3436426116838485,
+      "grad_norm": 0.31639760732650757,
+      "learning_rate": 0.0005361923076923076,
+      "loss": 3.5059,
+      "step": 18350
+    },
+    {
+      "epoch": 5.358203739297571,
+      "grad_norm": 0.3361245095729828,
+      "learning_rate": 0.0005360174825174825,
+      "loss": 3.5089,
+      "step": 18400
+    },
+    {
+      "epoch": 5.372764866911294,
+      "grad_norm": 0.308998167514801,
+      "learning_rate": 0.0005358426573426573,
+      "loss": 3.517,
+      "step": 18450
+    },
+    {
+      "epoch": 5.387325994525016,
+      "grad_norm": 0.30970314145088196,
+      "learning_rate": 0.0005356678321678321,
+      "loss": 3.5048,
+      "step": 18500
+    },
+    {
+      "epoch": 5.401887122138739,
+      "grad_norm": 0.3119153380393982,
+      "learning_rate": 0.0005354930069930069,
+      "loss": 3.5108,
+      "step": 18550
+    },
+    {
+      "epoch": 5.41644824975246,
+      "grad_norm": 0.32366570830345154,
+      "learning_rate": 0.0005353181818181817,
+      "loss": 3.5209,
+      "step": 18600
+    },
+    {
+      "epoch": 5.431009377366183,
+      "grad_norm": 0.34180885553359985,
+      "learning_rate": 0.0005351433566433565,
+      "loss": 3.5116,
+      "step": 18650
+    },
+    {
+      "epoch": 5.445570504979906,
+      "grad_norm": 0.30780646204948425,
+      "learning_rate": 0.0005349685314685314,
+      "loss": 3.5122,
+      "step": 18700
+    },
+    {
+      "epoch": 5.460131632593628,
+      "grad_norm": 0.32243868708610535,
+      "learning_rate": 0.0005347937062937062,
+      "loss": 3.5189,
+      "step": 18750
+    },
+    {
+      "epoch": 5.4746927602073505,
+      "grad_norm": 0.30452656745910645,
+      "learning_rate": 0.000534618881118881,
+      "loss": 3.5301,
+      "step": 18800
+    },
+    {
+      "epoch": 5.489253887821073,
+      "grad_norm": 0.31950345635414124,
+      "learning_rate": 0.0005344440559440559,
+      "loss": 3.5165,
+      "step": 18850
+    },
+    {
+      "epoch": 5.503815015434795,
+      "grad_norm": 0.32063764333724976,
+      "learning_rate": 0.0005342692307692307,
+      "loss": 3.5179,
+      "step": 18900
+    },
+    {
+      "epoch": 5.518376143048518,
+      "grad_norm": 0.31415075063705444,
+      "learning_rate": 0.0005340944055944055,
+      "loss": 3.5253,
+      "step": 18950
+    },
+    {
+      "epoch": 5.53293727066224,
+      "grad_norm": 0.2983076572418213,
+      "learning_rate": 0.0005339195804195803,
+      "loss": 3.5261,
+      "step": 19000
+    },
+    {
+      "epoch": 5.53293727066224,
+      "eval_accuracy": 0.3608592462873141,
+      "eval_loss": 3.6253974437713623,
+      "eval_runtime": 180.632,
+      "eval_samples_per_second": 92.154,
+      "eval_steps_per_second": 5.763,
+      "step": 19000
+    },
+    {
+      "epoch": 5.5474983982759625,
+      "grad_norm": 0.3204125463962555,
+      "learning_rate": 0.0005337447552447552,
+      "loss": 3.5249,
+      "step": 19050
+    },
+    {
+      "epoch": 5.562059525889685,
+      "grad_norm": 0.3400876224040985,
+      "learning_rate": 0.00053356993006993,
+      "loss": 3.5125,
+      "step": 19100
+    },
+    {
+      "epoch": 5.576620653503407,
+      "grad_norm": 0.31529760360717773,
+      "learning_rate": 0.0005333951048951048,
+      "loss": 3.522,
+      "step": 19150
+    },
+    {
+      "epoch": 5.59118178111713,
+      "grad_norm": 0.30710023641586304,
+      "learning_rate": 0.0005332202797202796,
+      "loss": 3.5183,
+      "step": 19200
+    },
+    {
+      "epoch": 5.605742908730852,
+      "grad_norm": 0.30690956115722656,
+      "learning_rate": 0.0005330454545454546,
+      "loss": 3.5081,
+      "step": 19250
+    },
+    {
+      "epoch": 5.620304036344574,
+      "grad_norm": 0.3141026794910431,
+      "learning_rate": 0.0005328706293706292,
+      "loss": 3.5201,
+      "step": 19300
+    },
+    {
+      "epoch": 5.634865163958297,
+      "grad_norm": 0.3454573452472687,
+      "learning_rate": 0.0005326958041958042,
+      "loss": 3.5377,
+      "step": 19350
+    },
+    {
+      "epoch": 5.649426291572019,
+      "grad_norm": 0.32653287053108215,
+      "learning_rate": 0.000532520979020979,
+      "loss": 3.5207,
+      "step": 19400
+    },
+    {
+      "epoch": 5.663987419185742,
+      "grad_norm": 0.30469658970832825,
+      "learning_rate": 0.0005323461538461538,
+      "loss": 3.5373,
+      "step": 19450
+    },
+    {
+      "epoch": 5.6785485467994645,
+      "grad_norm": 0.29301902651786804,
+      "learning_rate": 0.0005321713286713287,
+      "loss": 3.5197,
+      "step": 19500
+    },
+    {
+      "epoch": 5.693109674413186,
+      "grad_norm": 0.34159839153289795,
+      "learning_rate": 0.0005319965034965035,
+      "loss": 3.5231,
+      "step": 19550
+    },
+    {
+      "epoch": 5.707670802026909,
+      "grad_norm": 0.30907705426216125,
+      "learning_rate": 0.0005318216783216783,
+      "loss": 3.527,
+      "step": 19600
+    },
+    {
+      "epoch": 5.722231929640631,
+      "grad_norm": 0.3162166178226471,
+      "learning_rate": 0.0005316468531468531,
+      "loss": 3.5296,
+      "step": 19650
+    },
+    {
+      "epoch": 5.736793057254354,
+      "grad_norm": 0.3204457461833954,
+      "learning_rate": 0.000531472027972028,
+      "loss": 3.5307,
+      "step": 19700
+    },
+    {
+      "epoch": 5.7513541848680765,
+      "grad_norm": 0.3057008683681488,
+      "learning_rate": 0.0005312972027972028,
+      "loss": 3.53,
+      "step": 19750
+    },
+    {
+      "epoch": 5.765915312481798,
+      "grad_norm": 0.32276445627212524,
+      "learning_rate": 0.0005311223776223776,
+      "loss": 3.5437,
+      "step": 19800
+    },
+    {
+      "epoch": 5.780476440095521,
+      "grad_norm": 0.30552002787590027,
+      "learning_rate": 0.0005309475524475524,
+      "loss": 3.5316,
+      "step": 19850
+    },
+    {
+      "epoch": 5.795037567709244,
+      "grad_norm": 0.2986491918563843,
+      "learning_rate": 0.0005307727272727273,
+      "loss": 3.5291,
+      "step": 19900
+    },
+    {
+      "epoch": 5.809598695322966,
+      "grad_norm": 0.3022374212741852,
+      "learning_rate": 0.0005305979020979021,
+      "loss": 3.5171,
+      "step": 19950
+    },
+    {
+      "epoch": 5.824159822936688,
+      "grad_norm": 0.3142791986465454,
+      "learning_rate": 0.0005304230769230769,
+      "loss": 3.5262,
+      "step": 20000
+    },
+    {
+      "epoch": 5.824159822936688,
+      "eval_accuracy": 0.36232043193442454,
+      "eval_loss": 3.614790439605713,
+      "eval_runtime": 180.8701,
+      "eval_samples_per_second": 92.033,
+      "eval_steps_per_second": 5.756,
+      "step": 20000
+    },
+    {
+      "epoch": 5.83872095055041,
+      "grad_norm": 0.300855427980423,
+      "learning_rate": 0.0005302482517482517,
+      "loss": 3.538,
+      "step": 20050
+    },
+    {
+      "epoch": 5.853282078164133,
+      "grad_norm": 0.286253958940506,
+      "learning_rate": 0.0005300734265734265,
+      "loss": 3.5309,
+      "step": 20100
+    },
+    {
+      "epoch": 5.867843205777856,
+      "grad_norm": 0.2925945222377777,
+      "learning_rate": 0.0005298986013986013,
+      "loss": 3.5247,
+      "step": 20150
+    },
+    {
+      "epoch": 5.882404333391578,
+      "grad_norm": 0.31259238719940186,
+      "learning_rate": 0.0005297237762237762,
+      "loss": 3.5267,
+      "step": 20200
+    },
+    {
+      "epoch": 5.8969654610053,
+      "grad_norm": 0.30897215008735657,
+      "learning_rate": 0.000529548951048951,
+      "loss": 3.5384,
+      "step": 20250
+    },
+    {
+      "epoch": 5.911526588619022,
+      "grad_norm": 0.3039945363998413,
+      "learning_rate": 0.0005293741258741258,
+      "loss": 3.5347,
+      "step": 20300
+    },
+    {
+      "epoch": 5.926087716232745,
+      "grad_norm": 0.30515971779823303,
+      "learning_rate": 0.0005291993006993007,
+      "loss": 3.5327,
+      "step": 20350
+    },
+    {
+      "epoch": 5.940648843846468,
+      "grad_norm": 0.3250819742679596,
+      "learning_rate": 0.0005290244755244755,
+      "loss": 3.5112,
+      "step": 20400
+    },
+    {
+      "epoch": 5.95520997146019,
+      "grad_norm": 0.31303057074546814,
+      "learning_rate": 0.0005288496503496503,
+      "loss": 3.5379,
+      "step": 20450
+    },
+    {
+      "epoch": 5.969771099073912,
+      "grad_norm": 0.30073851346969604,
+      "learning_rate": 0.0005286748251748251,
+      "loss": 3.5411,
+      "step": 20500
+    },
+    {
+      "epoch": 5.984332226687634,
+      "grad_norm": 0.3091214597225189,
+      "learning_rate": 0.0005285,
+      "loss": 3.5171,
+      "step": 20550
+    },
+    {
+      "epoch": 5.998893354301357,
+      "grad_norm": 0.32308459281921387,
+      "learning_rate": 0.0005283251748251748,
+      "loss": 3.5282,
+      "step": 20600
+    },
+    {
+      "epoch": 6.013396237404625,
+      "grad_norm": 0.33346566557884216,
+      "learning_rate": 0.0005281503496503496,
+      "loss": 3.4215,
+      "step": 20650
+    },
+    {
+      "epoch": 6.027957365018347,
+      "grad_norm": 0.2992725372314453,
+      "learning_rate": 0.0005279755244755244,
+      "loss": 3.42,
+      "step": 20700
+    },
+    {
+      "epoch": 6.04251849263207,
+      "grad_norm": 0.3181827664375305,
+      "learning_rate": 0.0005278006993006993,
+      "loss": 3.4206,
+      "step": 20750
+    },
+    {
+      "epoch": 6.0570796202457915,
+      "grad_norm": 0.3308713734149933,
+      "learning_rate": 0.000527625874125874,
+      "loss": 3.436,
+      "step": 20800
+    },
+    {
+      "epoch": 6.071640747859514,
+      "grad_norm": 0.33250924944877625,
+      "learning_rate": 0.0005274510489510489,
+      "loss": 3.43,
+      "step": 20850
+    },
+    {
+      "epoch": 6.086201875473237,
+      "grad_norm": 0.31477755308151245,
+      "learning_rate": 0.0005272762237762238,
+      "loss": 3.4421,
+      "step": 20900
+    },
+    {
+      "epoch": 6.100763003086959,
+      "grad_norm": 0.30721914768218994,
+      "learning_rate": 0.0005271013986013985,
+      "loss": 3.4443,
+      "step": 20950
+    },
+    {
+      "epoch": 6.115324130700682,
+      "grad_norm": 0.3389357924461365,
+      "learning_rate": 0.0005269265734265734,
+      "loss": 3.4388,
+      "step": 21000
+    },
+    {
+      "epoch": 6.115324130700682,
+      "eval_accuracy": 0.3626586595558532,
+      "eval_loss": 3.616656541824341,
+      "eval_runtime": 180.8276,
+      "eval_samples_per_second": 92.055,
+      "eval_steps_per_second": 5.757,
+      "step": 21000
+    },
+    {
+      "epoch": 6.1298852583144035,
+      "grad_norm": 0.3342684507369995,
+      "learning_rate": 0.0005267517482517482,
+      "loss": 3.433,
+      "step": 21050
+    },
+    {
+      "epoch": 6.144446385928126,
+      "grad_norm": 0.3244004547595978,
+      "learning_rate": 0.000526576923076923,
+      "loss": 3.4507,
+      "step": 21100
+    },
+    {
+      "epoch": 6.159007513541849,
+      "grad_norm": 0.30746179819107056,
+      "learning_rate": 0.0005264020979020978,
+      "loss": 3.4594,
+      "step": 21150
+    },
+    {
+      "epoch": 6.173568641155571,
+      "grad_norm": 0.30511513352394104,
+      "learning_rate": 0.0005262272727272727,
+      "loss": 3.4499,
+      "step": 21200
+    },
+    {
+      "epoch": 6.1881297687692935,
+      "grad_norm": 0.30683988332748413,
+      "learning_rate": 0.0005260524475524475,
+      "loss": 3.4515,
+      "step": 21250
+    },
+    {
+      "epoch": 6.202690896383016,
+      "grad_norm": 0.3043258488178253,
+      "learning_rate": 0.0005258776223776223,
+      "loss": 3.4542,
+      "step": 21300
+    },
+    {
+      "epoch": 6.217252023996738,
+      "grad_norm": 0.32519418001174927,
+      "learning_rate": 0.0005257027972027971,
+      "loss": 3.4434,
+      "step": 21350
+    },
+    {
+      "epoch": 6.231813151610461,
+      "grad_norm": 0.3197823762893677,
+      "learning_rate": 0.000525527972027972,
+      "loss": 3.4388,
+      "step": 21400
+    },
+    {
+      "epoch": 6.246374279224183,
+      "grad_norm": 0.3240765333175659,
+      "learning_rate": 0.0005253531468531468,
+      "loss": 3.4588,
+      "step": 21450
+    },
+    {
+      "epoch": 6.2609354068379055,
+      "grad_norm": 0.34707140922546387,
+      "learning_rate": 0.0005251783216783216,
+      "loss": 3.4539,
+      "step": 21500
+    },
+    {
+      "epoch": 6.275496534451628,
+      "grad_norm": 0.337259441614151,
+      "learning_rate": 0.0005250034965034965,
+      "loss": 3.4567,
+      "step": 21550
+    },
+    {
+      "epoch": 6.29005766206535,
+      "grad_norm": 0.31958332657814026,
+      "learning_rate": 0.0005248286713286712,
+      "loss": 3.4585,
+      "step": 21600
+    },
+    {
+      "epoch": 6.304618789679073,
+      "grad_norm": 0.32298600673675537,
+      "learning_rate": 0.0005246538461538461,
+      "loss": 3.463,
+      "step": 21650
+    },
+    {
+      "epoch": 6.319179917292795,
+      "grad_norm": 0.3255809247493744,
+      "learning_rate": 0.0005244790209790209,
+      "loss": 3.4608,
+      "step": 21700
+    },
+    {
+      "epoch": 6.3337410449065175,
+      "grad_norm": 0.32428938150405884,
+      "learning_rate": 0.0005243041958041957,
+      "loss": 3.4699,
+      "step": 21750
+    },
+    {
+      "epoch": 6.34830217252024,
+      "grad_norm": 0.3299717307090759,
+      "learning_rate": 0.0005241293706293705,
+      "loss": 3.4656,
+      "step": 21800
+    },
+    {
+      "epoch": 6.362863300133962,
+      "grad_norm": 0.32020100951194763,
+      "learning_rate": 0.0005239545454545454,
+      "loss": 3.4592,
+      "step": 21850
+    },
+    {
+      "epoch": 6.377424427747685,
+      "grad_norm": 0.3127721846103668,
+      "learning_rate": 0.0005237797202797202,
+      "loss": 3.4843,
+      "step": 21900
+    },
+    {
+      "epoch": 6.391985555361408,
+      "grad_norm": 0.3199755847454071,
+      "learning_rate": 0.000523604895104895,
+      "loss": 3.4847,
+      "step": 21950
+    },
+    {
+      "epoch": 6.406546682975129,
+      "grad_norm": 0.3283131718635559,
+      "learning_rate": 0.0005234300699300698,
+      "loss": 3.4634,
+      "step": 22000
+    },
+    {
+      "epoch": 6.406546682975129,
+      "eval_accuracy": 0.36310868921689904,
+      "eval_loss": 3.6098170280456543,
+      "eval_runtime": 181.4687,
+      "eval_samples_per_second": 91.729,
+      "eval_steps_per_second": 5.737,
+      "step": 22000
+    },
+    {
+      "epoch": 6.421107810588852,
+      "grad_norm": 0.3273344933986664,
+      "learning_rate": 0.0005232552447552447,
+      "loss": 3.4702,
+      "step": 22050
+    },
+    {
+      "epoch": 6.435668938202574,
+      "grad_norm": 0.3400958478450775,
+      "learning_rate": 0.0005230804195804195,
+      "loss": 3.4588,
+      "step": 22100
+    },
+    {
+      "epoch": 6.450230065816297,
+      "grad_norm": 0.3155398666858673,
+      "learning_rate": 0.0005229055944055943,
+      "loss": 3.486,
+      "step": 22150
+    },
+    {
+      "epoch": 6.4647911934300195,
+      "grad_norm": 0.32073867321014404,
+      "learning_rate": 0.0005227307692307691,
+      "loss": 3.482,
+      "step": 22200
+    },
+    {
+      "epoch": 6.479352321043741,
+      "grad_norm": 0.30934587121009827,
+      "learning_rate": 0.0005225559440559441,
+      "loss": 3.4709,
+      "step": 22250
+    },
+    {
+      "epoch": 6.493913448657464,
+      "grad_norm": 0.3552294671535492,
+      "learning_rate": 0.0005223811188811189,
+      "loss": 3.4915,
+      "step": 22300
+    },
+    {
+      "epoch": 6.508474576271187,
+      "grad_norm": 0.32506152987480164,
+      "learning_rate": 0.0005222062937062937,
+      "loss": 3.4893,
+      "step": 22350
+    },
+    {
+      "epoch": 6.523035703884909,
+      "grad_norm": 0.31292617321014404,
+      "learning_rate": 0.0005220314685314686,
+      "loss": 3.4771,
+      "step": 22400
+    },
+    {
+      "epoch": 6.5375968314986315,
+      "grad_norm": 0.3158280849456787,
+      "learning_rate": 0.0005218566433566433,
+      "loss": 3.4818,
+      "step": 22450
+    },
+    {
+      "epoch": 6.552157959112353,
+      "grad_norm": 0.3074539303779602,
+      "learning_rate": 0.0005216818181818182,
+      "loss": 3.4723,
+      "step": 22500
+    },
+    {
+      "epoch": 6.566719086726076,
+      "grad_norm": 0.30536600947380066,
+      "learning_rate": 0.000521506993006993,
+      "loss": 3.4755,
+      "step": 22550
+    },
+    {
+      "epoch": 6.581280214339799,
+      "grad_norm": 0.3104475438594818,
+      "learning_rate": 0.0005213321678321678,
+      "loss": 3.4779,
+      "step": 22600
+    },
+    {
+      "epoch": 6.595841341953521,
+      "grad_norm": 0.3355538547039032,
+      "learning_rate": 0.0005211573426573426,
+      "loss": 3.4812,
+      "step": 22650
+    },
+    {
+      "epoch": 6.610402469567243,
+      "grad_norm": 0.3326936364173889,
+      "learning_rate": 0.0005209825174825175,
+      "loss": 3.4838,
+      "step": 22700
+    },
+    {
+      "epoch": 6.624963597180965,
+      "grad_norm": 0.311422735452652,
+      "learning_rate": 0.0005208076923076923,
+      "loss": 3.4823,
+      "step": 22750
+    },
+    {
+      "epoch": 6.639524724794688,
+      "grad_norm": 0.3325813412666321,
+      "learning_rate": 0.0005206328671328671,
+      "loss": 3.4808,
+      "step": 22800
+    },
+    {
+      "epoch": 6.654085852408411,
+      "grad_norm": 0.32317134737968445,
+      "learning_rate": 0.0005204580419580419,
+      "loss": 3.4865,
+      "step": 22850
+    },
+    {
+      "epoch": 6.668646980022133,
+      "grad_norm": 0.307810515165329,
+      "learning_rate": 0.0005202832167832168,
+      "loss": 3.4835,
+      "step": 22900
+    },
+    {
+      "epoch": 6.683208107635855,
+      "grad_norm": 0.3085881471633911,
+      "learning_rate": 0.0005201083916083916,
+      "loss": 3.4681,
+      "step": 22950
+    },
+    {
+      "epoch": 6.697769235249577,
+      "grad_norm": 0.33852365612983704,
+      "learning_rate": 0.0005199335664335664,
+      "loss": 3.5019,
+      "step": 23000
+    },
+    {
+      "epoch": 6.697769235249577,
+      "eval_accuracy": 0.3637609265626363,
+      "eval_loss": 3.6008996963500977,
+      "eval_runtime": 180.919,
+      "eval_samples_per_second": 92.008,
+      "eval_steps_per_second": 5.754,
+      "step": 23000
+    },
+    {
+      "epoch": 6.7123303628633,
+      "grad_norm": 0.3305760622024536,
+      "learning_rate": 0.0005197587412587413,
+      "loss": 3.4833,
+      "step": 23050
+    },
+    {
+      "epoch": 6.726891490477023,
+      "grad_norm": 0.30366402864456177,
+      "learning_rate": 0.0005195839160839161,
+      "loss": 3.4811,
+      "step": 23100
+    },
+    {
+      "epoch": 6.741452618090745,
+      "grad_norm": 0.2957024872303009,
+      "learning_rate": 0.0005194090909090909,
+      "loss": 3.4887,
+      "step": 23150
+    },
+    {
+      "epoch": 6.756013745704467,
+      "grad_norm": 0.33022427558898926,
+      "learning_rate": 0.0005192342657342657,
+      "loss": 3.4863,
+      "step": 23200
+    },
+    {
+      "epoch": 6.77057487331819,
+      "grad_norm": 0.3377295732498169,
+      "learning_rate": 0.0005190594405594405,
+      "loss": 3.4843,
+      "step": 23250
+    },
+    {
+      "epoch": 6.785136000931912,
+      "grad_norm": 0.288236141204834,
+      "learning_rate": 0.0005188846153846153,
+      "loss": 3.5076,
+      "step": 23300
+    },
+    {
+      "epoch": 6.799697128545635,
+      "grad_norm": 0.3370327353477478,
+      "learning_rate": 0.0005187097902097902,
+      "loss": 3.4846,
+      "step": 23350
+    },
+    {
+      "epoch": 6.814258256159357,
+      "grad_norm": 0.32287853956222534,
+      "learning_rate": 0.000518534965034965,
+      "loss": 3.4971,
+      "step": 23400
+    },
+    {
+      "epoch": 6.828819383773079,
+      "grad_norm": 0.3225182294845581,
+      "learning_rate": 0.0005183601398601398,
+      "loss": 3.4699,
+      "step": 23450
+    },
+    {
+      "epoch": 6.843380511386802,
+      "grad_norm": 0.32776007056236267,
+      "learning_rate": 0.0005181853146853146,
+      "loss": 3.4867,
+      "step": 23500
+    },
+    {
+      "epoch": 6.857941639000524,
+      "grad_norm": 0.3112199306488037,
+      "learning_rate": 0.0005180104895104895,
+      "loss": 3.4929,
+      "step": 23550
+    },
+    {
+      "epoch": 6.872502766614247,
+      "grad_norm": 0.3101305663585663,
+      "learning_rate": 0.0005178356643356643,
+      "loss": 3.4852,
+      "step": 23600
+    },
+    {
+      "epoch": 6.887063894227969,
+      "grad_norm": 0.34980154037475586,
+      "learning_rate": 0.0005176608391608391,
+      "loss": 3.4916,
+      "step": 23650
+    },
+    {
+      "epoch": 6.901625021841691,
+      "grad_norm": 0.3215571939945221,
+      "learning_rate": 0.000517486013986014,
+      "loss": 3.4847,
+      "step": 23700
+    },
+    {
+      "epoch": 6.916186149455414,
+      "grad_norm": 0.33672600984573364,
+      "learning_rate": 0.0005173111888111888,
+      "loss": 3.4901,
+      "step": 23750
+    },
+    {
+      "epoch": 6.930747277069136,
+      "grad_norm": 0.3331395387649536,
+      "learning_rate": 0.0005171363636363636,
+      "loss": 3.4927,
+      "step": 23800
+    },
+    {
+      "epoch": 6.945308404682859,
+      "grad_norm": 0.3124479055404663,
+      "learning_rate": 0.0005169615384615384,
+      "loss": 3.506,
+      "step": 23850
+    },
+    {
+      "epoch": 6.959869532296581,
+      "grad_norm": 0.33297938108444214,
+      "learning_rate": 0.0005167867132867133,
+      "loss": 3.4942,
+      "step": 23900
+    },
+    {
+      "epoch": 6.974430659910303,
+      "grad_norm": 0.31658828258514404,
+      "learning_rate": 0.000516611888111888,
+      "loss": 3.5013,
+      "step": 23950
+    },
+    {
+      "epoch": 6.988991787524026,
+      "grad_norm": 0.32839858531951904,
+      "learning_rate": 0.0005164370629370629,
+      "loss": 3.4874,
+      "step": 24000
+    },
+    {
+      "epoch": 6.988991787524026,
+      "eval_accuracy": 0.36462360097558155,
+      "eval_loss": 3.5910537242889404,
+      "eval_runtime": 180.6847,
+      "eval_samples_per_second": 92.127,
+      "eval_steps_per_second": 5.761,
+      "step": 24000
+    },
+    {
+      "epoch": 7.003494670627293,
+      "grad_norm": 0.32422536611557007,
+      "learning_rate": 0.0005162622377622377,
+      "loss": 3.4613,
+      "step": 24050
+    },
+    {
+      "epoch": 7.018055798241016,
+      "grad_norm": 0.31693896651268005,
+      "learning_rate": 0.0005160874125874125,
+      "loss": 3.3777,
+      "step": 24100
+    },
+    {
+      "epoch": 7.032616925854738,
+      "grad_norm": 0.3402048349380493,
+      "learning_rate": 0.0005159125874125873,
+      "loss": 3.3793,
+      "step": 24150
+    },
+    {
+      "epoch": 7.0471780534684605,
+      "grad_norm": 0.3285346031188965,
+      "learning_rate": 0.0005157377622377622,
+      "loss": 3.387,
+      "step": 24200
+    },
+    {
+      "epoch": 7.061739181082183,
+      "grad_norm": 0.330010324716568,
+      "learning_rate": 0.000515562937062937,
+      "loss": 3.392,
+      "step": 24250
+    },
+    {
+      "epoch": 7.076300308695905,
+      "grad_norm": 0.32283395528793335,
+      "learning_rate": 0.0005153881118881118,
+      "loss": 3.4001,
+      "step": 24300
+    },
+    {
+      "epoch": 7.090861436309628,
+      "grad_norm": 0.3089471459388733,
+      "learning_rate": 0.0005152132867132867,
+      "loss": 3.4023,
+      "step": 24350
+    },
+    {
+      "epoch": 7.105422563923351,
+      "grad_norm": 0.32891589403152466,
+      "learning_rate": 0.0005150384615384615,
+      "loss": 3.3962,
+      "step": 24400
+    },
+    {
+      "epoch": 7.1199836915370724,
+      "grad_norm": 0.34227320551872253,
+      "learning_rate": 0.0005148636363636363,
+      "loss": 3.4008,
+      "step": 24450
+    },
+    {
+      "epoch": 7.134544819150795,
+      "grad_norm": 0.31279921531677246,
+      "learning_rate": 0.0005146888111888111,
+      "loss": 3.4203,
+      "step": 24500
+    },
+    {
+      "epoch": 7.149105946764517,
+      "grad_norm": 0.3075491189956665,
+      "learning_rate": 0.000514513986013986,
+      "loss": 3.4162,
+      "step": 24550
+    },
+    {
+      "epoch": 7.16366707437824,
+      "grad_norm": 0.32692641019821167,
+      "learning_rate": 0.0005143391608391608,
+      "loss": 3.4139,
+      "step": 24600
+    },
+    {
+      "epoch": 7.1782282019919625,
+      "grad_norm": 0.34598174691200256,
+      "learning_rate": 0.0005141643356643356,
+      "loss": 3.4051,
+      "step": 24650
+    },
+    {
+      "epoch": 7.192789329605684,
+      "grad_norm": 0.33088117837905884,
+      "learning_rate": 0.0005139895104895104,
+      "loss": 3.4026,
+      "step": 24700
+    },
+    {
+      "epoch": 7.207350457219407,
+      "grad_norm": 0.33946093916893005,
+      "learning_rate": 0.0005138146853146852,
+      "loss": 3.412,
+      "step": 24750
+    },
+    {
+      "epoch": 7.22191158483313,
+      "grad_norm": 0.32189738750457764,
+      "learning_rate": 0.00051363986013986,
+      "loss": 3.4248,
+      "step": 24800
+    },
+    {
+      "epoch": 7.236472712446852,
+      "grad_norm": 0.3421887457370758,
+      "learning_rate": 0.0005134650349650349,
+      "loss": 3.4235,
+      "step": 24850
+    },
+    {
+      "epoch": 7.2510338400605745,
+      "grad_norm": 0.32535749673843384,
+      "learning_rate": 0.0005132902097902097,
+      "loss": 3.4198,
+      "step": 24900
+    },
+    {
+      "epoch": 7.265594967674296,
+      "grad_norm": 0.30061206221580505,
+      "learning_rate": 0.0005131153846153845,
+      "loss": 3.418,
+      "step": 24950
+    },
+    {
+      "epoch": 7.280156095288019,
+      "grad_norm": 0.31006714701652527,
+      "learning_rate": 0.0005129405594405594,
+      "loss": 3.4222,
+      "step": 25000
+    },
+    {
+      "epoch": 7.280156095288019,
+      "eval_accuracy": 0.3645690519257578,
+      "eval_loss": 3.5986101627349854,
+      "eval_runtime": 180.7195,
+      "eval_samples_per_second": 92.11,
+      "eval_steps_per_second": 5.76,
+      "step": 25000
+    },
+    {
+      "epoch": 7.294717222901742,
+      "grad_norm": 0.32596707344055176,
+      "learning_rate": 0.0005127657342657342,
+      "loss": 3.4334,
+      "step": 25050
+    },
+    {
+      "epoch": 7.309278350515464,
+      "grad_norm": 0.3368932604789734,
+      "learning_rate": 0.000512590909090909,
+      "loss": 3.4284,
+      "step": 25100
+    },
+    {
+      "epoch": 7.3238394781291865,
+      "grad_norm": 0.3360914885997772,
+      "learning_rate": 0.0005124160839160838,
+      "loss": 3.4356,
+      "step": 25150
+    },
+    {
+      "epoch": 7.338400605742908,
+      "grad_norm": 0.3240918517112732,
+      "learning_rate": 0.0005122412587412588,
+      "loss": 3.4355,
+      "step": 25200
+    },
+    {
+      "epoch": 7.352961733356631,
+      "grad_norm": 0.3215958774089813,
+      "learning_rate": 0.0005120664335664336,
+      "loss": 3.4247,
+      "step": 25250
+    },
+    {
+      "epoch": 7.367522860970354,
+      "grad_norm": 0.3438575267791748,
+      "learning_rate": 0.0005118916083916084,
+      "loss": 3.4302,
+      "step": 25300
+    },
+    {
+      "epoch": 7.382083988584076,
+      "grad_norm": 0.3164159655570984,
+      "learning_rate": 0.0005117167832167832,
+      "loss": 3.4298,
+      "step": 25350
+    },
+    {
+      "epoch": 7.396645116197798,
+      "grad_norm": 0.32072940468788147,
+      "learning_rate": 0.0005115419580419581,
+      "loss": 3.4205,
+      "step": 25400
+    },
+    {
+      "epoch": 7.411206243811521,
+      "grad_norm": 0.32689517736434937,
+      "learning_rate": 0.0005113671328671328,
+      "loss": 3.4321,
+      "step": 25450
+    },
+    {
+      "epoch": 7.425767371425243,
+      "grad_norm": 0.31307026743888855,
+      "learning_rate": 0.0005111923076923077,
+      "loss": 3.4414,
+      "step": 25500
+    },
+    {
+      "epoch": 7.440328499038966,
+      "grad_norm": 0.3382061719894409,
+      "learning_rate": 0.0005110174825174825,
+      "loss": 3.4383,
+      "step": 25550
+    },
+    {
+      "epoch": 7.454889626652688,
+      "grad_norm": 0.32721343636512756,
+      "learning_rate": 0.0005108426573426573,
+      "loss": 3.4344,
+      "step": 25600
+    },
+    {
+      "epoch": 7.46945075426641,
+      "grad_norm": 0.3294874131679535,
+      "learning_rate": 0.0005106678321678321,
+      "loss": 3.4324,
+      "step": 25650
+    },
+    {
+      "epoch": 7.484011881880133,
+      "grad_norm": 0.3590655028820038,
+      "learning_rate": 0.000510493006993007,
+      "loss": 3.4372,
+      "step": 25700
+    },
+    {
+      "epoch": 7.498573009493855,
+      "grad_norm": 0.3113061189651489,
+      "learning_rate": 0.0005103181818181818,
+      "loss": 3.4376,
+      "step": 25750
+    },
+    {
+      "epoch": 7.513134137107578,
+      "grad_norm": 0.3551453948020935,
+      "learning_rate": 0.0005101433566433566,
+      "loss": 3.4468,
+      "step": 25800
+    },
+    {
+      "epoch": 7.5276952647213005,
+      "grad_norm": 0.3093542456626892,
+      "learning_rate": 0.0005099685314685315,
+      "loss": 3.453,
+      "step": 25850
+    },
+    {
+      "epoch": 7.542256392335022,
+      "grad_norm": 0.32737603783607483,
+      "learning_rate": 0.0005097937062937063,
+      "loss": 3.4581,
+      "step": 25900
+    },
+    {
+      "epoch": 7.556817519948745,
+      "grad_norm": 0.3264789581298828,
+      "learning_rate": 0.0005096188811188811,
+      "loss": 3.4601,
+      "step": 25950
+    },
+    {
+      "epoch": 7.571378647562467,
+      "grad_norm": 0.312288761138916,
+      "learning_rate": 0.0005094440559440559,
+      "loss": 3.4584,
+      "step": 26000
+    },
+    {
+      "epoch": 7.571378647562467,
+      "eval_accuracy": 0.36503354178751124,
+      "eval_loss": 3.5893683433532715,
+      "eval_runtime": 180.8574,
+      "eval_samples_per_second": 92.039,
+      "eval_steps_per_second": 5.756,
+      "step": 26000
+    },
+    {
+      "epoch": 7.58593977517619,
+      "grad_norm": 0.30291152000427246,
+      "learning_rate": 0.0005092692307692308,
+      "loss": 3.4588,
+      "step": 26050
+    },
+    {
+      "epoch": 7.600500902789912,
+      "grad_norm": 0.3086663484573364,
+      "learning_rate": 0.0005090944055944056,
+      "loss": 3.4499,
+      "step": 26100
+    },
+    {
+      "epoch": 7.615062030403634,
+      "grad_norm": 0.31779009103775024,
+      "learning_rate": 0.0005089195804195804,
+      "loss": 3.4592,
+      "step": 26150
+    },
+    {
+      "epoch": 7.629623158017357,
+      "grad_norm": 0.3134089708328247,
+      "learning_rate": 0.0005087447552447552,
+      "loss": 3.4564,
+      "step": 26200
+    },
+    {
+      "epoch": 7.644184285631079,
+      "grad_norm": 0.316002756357193,
+      "learning_rate": 0.00050856993006993,
+      "loss": 3.4563,
+      "step": 26250
+    },
+    {
+      "epoch": 7.658745413244802,
+      "grad_norm": 0.3355805277824402,
+      "learning_rate": 0.0005083951048951048,
+      "loss": 3.447,
+      "step": 26300
+    },
+    {
+      "epoch": 7.673306540858524,
+      "grad_norm": 0.3496020436286926,
+      "learning_rate": 0.0005082202797202797,
+      "loss": 3.4441,
+      "step": 26350
+    },
+    {
+      "epoch": 7.687867668472246,
+      "grad_norm": 0.31140685081481934,
+      "learning_rate": 0.0005080454545454545,
+      "loss": 3.4427,
+      "step": 26400
+    },
+    {
+      "epoch": 7.702428796085969,
+      "grad_norm": 0.35752904415130615,
+      "learning_rate": 0.0005078706293706293,
+      "loss": 3.4618,
+      "step": 26450
+    },
+    {
+      "epoch": 7.716989923699691,
+      "grad_norm": 0.34923386573791504,
+      "learning_rate": 0.0005076958041958042,
+      "loss": 3.4498,
+      "step": 26500
+    },
+    {
+      "epoch": 7.731551051313414,
+      "grad_norm": 0.3136448562145233,
+      "learning_rate": 0.000507520979020979,
+      "loss": 3.4591,
+      "step": 26550
+    },
+    {
+      "epoch": 7.746112178927136,
+      "grad_norm": 0.31951072812080383,
+      "learning_rate": 0.0005073461538461538,
+      "loss": 3.4538,
+      "step": 26600
+    },
+    {
+      "epoch": 7.760673306540858,
+      "grad_norm": 0.3343602120876312,
+      "learning_rate": 0.0005071713286713286,
+      "loss": 3.4467,
+      "step": 26650
+    },
+    {
+      "epoch": 7.775234434154581,
+      "grad_norm": 0.33834147453308105,
+      "learning_rate": 0.0005069965034965035,
+      "loss": 3.462,
+      "step": 26700
+    },
+    {
+      "epoch": 7.789795561768304,
+      "grad_norm": 0.3536285161972046,
+      "learning_rate": 0.0005068216783216783,
+      "loss": 3.4647,
+      "step": 26750
+    },
+    {
+      "epoch": 7.8043566893820255,
+      "grad_norm": 0.31148087978363037,
+      "learning_rate": 0.0005066468531468531,
+      "loss": 3.4641,
+      "step": 26800
+    },
+    {
+      "epoch": 7.818917816995748,
+      "grad_norm": 0.32618945837020874,
+      "learning_rate": 0.0005064720279720279,
+      "loss": 3.4615,
+      "step": 26850
+    },
+    {
+      "epoch": 7.833478944609471,
+      "grad_norm": 0.31293508410453796,
+      "learning_rate": 0.0005062972027972028,
+      "loss": 3.4553,
+      "step": 26900
+    },
+    {
+      "epoch": 7.848040072223193,
+      "grad_norm": 0.3196380138397217,
+      "learning_rate": 0.0005061223776223775,
+      "loss": 3.4592,
+      "step": 26950
+    },
+    {
+      "epoch": 7.862601199836916,
+      "grad_norm": 0.3242320418357849,
+      "learning_rate": 0.0005059475524475524,
+      "loss": 3.4485,
+      "step": 27000
+    },
+    {
+      "epoch": 7.862601199836916,
+      "eval_accuracy": 0.3659237258505831,
+      "eval_loss": 3.580866575241089,
+      "eval_runtime": 180.9494,
+      "eval_samples_per_second": 91.993,
+      "eval_steps_per_second": 5.753,
+      "step": 27000
+    },
+    {
+      "epoch": 7.8771623274506375,
+      "grad_norm": 0.3238949477672577,
+      "learning_rate": 0.0005057727272727272,
+      "loss": 3.4538,
+      "step": 27050
+    },
+    {
+      "epoch": 7.89172345506436,
+      "grad_norm": 0.31904834508895874,
+      "learning_rate": 0.000505597902097902,
+      "loss": 3.4514,
+      "step": 27100
+    },
+    {
+      "epoch": 7.906284582678083,
+      "grad_norm": 0.32475146651268005,
+      "learning_rate": 0.0005054230769230769,
+      "loss": 3.4581,
+      "step": 27150
+    },
+    {
+      "epoch": 7.920845710291805,
+      "grad_norm": 0.32322096824645996,
+      "learning_rate": 0.0005052482517482517,
+      "loss": 3.449,
+      "step": 27200
+    },
+    {
+      "epoch": 7.935406837905528,
+      "grad_norm": 0.3155551552772522,
+      "learning_rate": 0.0005050734265734265,
+      "loss": 3.454,
+      "step": 27250
+    },
+    {
+      "epoch": 7.9499679655192494,
+      "grad_norm": 0.36903294920921326,
+      "learning_rate": 0.0005048986013986013,
+      "loss": 3.4722,
+      "step": 27300
+    },
+    {
+      "epoch": 7.964529093132972,
+      "grad_norm": 0.32118940353393555,
+      "learning_rate": 0.0005047237762237762,
+      "loss": 3.4534,
+      "step": 27350
+    },
+    {
+      "epoch": 7.979090220746695,
+      "grad_norm": 0.3173547387123108,
+      "learning_rate": 0.000504548951048951,
+      "loss": 3.4603,
+      "step": 27400
+    },
+    {
+      "epoch": 7.993651348360417,
+      "grad_norm": 0.30384644865989685,
+      "learning_rate": 0.0005043741258741258,
+      "loss": 3.4583,
+      "step": 27450
+    },
+    {
+      "epoch": 8.008154231463685,
+      "grad_norm": 0.3187147378921509,
+      "learning_rate": 0.0005041993006993006,
+      "loss": 3.4017,
+      "step": 27500
+    },
+    {
+      "epoch": 8.022715359077408,
+      "grad_norm": 0.32749438285827637,
+      "learning_rate": 0.0005040244755244755,
+      "loss": 3.3413,
+      "step": 27550
+    },
+    {
+      "epoch": 8.037276486691129,
+      "grad_norm": 0.3408881425857544,
+      "learning_rate": 0.0005038496503496503,
+      "loss": 3.3453,
+      "step": 27600
+    },
+    {
+      "epoch": 8.051837614304851,
+      "grad_norm": 0.3273647129535675,
+      "learning_rate": 0.0005036748251748251,
+      "loss": 3.3509,
+      "step": 27650
+    },
+    {
+      "epoch": 8.066398741918574,
+      "grad_norm": 0.3485855162143707,
+      "learning_rate": 0.0005034999999999999,
+      "loss": 3.3585,
+      "step": 27700
+    },
+    {
+      "epoch": 8.080959869532297,
+      "grad_norm": 0.31149420142173767,
+      "learning_rate": 0.0005033251748251747,
+      "loss": 3.3763,
+      "step": 27750
+    },
+    {
+      "epoch": 8.09552099714602,
+      "grad_norm": 0.35035839676856995,
+      "learning_rate": 0.0005031503496503496,
+      "loss": 3.3758,
+      "step": 27800
+    },
+    {
+      "epoch": 8.11008212475974,
+      "grad_norm": 0.3395816683769226,
+      "learning_rate": 0.0005029755244755244,
+      "loss": 3.3757,
+      "step": 27850
+    },
+    {
+      "epoch": 8.124643252373463,
+      "grad_norm": 0.3167729079723358,
+      "learning_rate": 0.0005028006993006992,
+      "loss": 3.3735,
+      "step": 27900
+    },
+    {
+      "epoch": 8.139204379987186,
+      "grad_norm": 0.31024643778800964,
+      "learning_rate": 0.000502625874125874,
+      "loss": 3.3797,
+      "step": 27950
+    },
+    {
+      "epoch": 8.153765507600909,
+      "grad_norm": 0.3287062346935272,
+      "learning_rate": 0.000502451048951049,
+      "loss": 3.3673,
+      "step": 28000
+    },
+    {
+      "epoch": 8.153765507600909,
+      "eval_accuracy": 0.36575196688120276,
+      "eval_loss": 3.587451457977295,
+      "eval_runtime": 180.7644,
+      "eval_samples_per_second": 92.087,
+      "eval_steps_per_second": 5.759,
+      "step": 28000
+    },
+    {
+      "epoch": 8.168326635214632,
+      "grad_norm": 0.3316042125225067,
+      "learning_rate": 0.0005022762237762237,
+      "loss": 3.3777,
+      "step": 28050
+    },
+    {
+      "epoch": 8.182887762828354,
+      "grad_norm": 0.3350354731082916,
+      "learning_rate": 0.0005021013986013985,
+      "loss": 3.3789,
+      "step": 28100
+    },
+    {
+      "epoch": 8.197448890442075,
+      "grad_norm": 0.29669225215911865,
+      "learning_rate": 0.0005019265734265733,
+      "loss": 3.3892,
+      "step": 28150
+    },
+    {
+      "epoch": 8.212010018055798,
+      "grad_norm": 0.33734533190727234,
+      "learning_rate": 0.0005017517482517483,
+      "loss": 3.389,
+      "step": 28200
+    },
+    {
+      "epoch": 8.22657114566952,
+      "grad_norm": 0.3267245292663574,
+      "learning_rate": 0.0005015769230769231,
+      "loss": 3.3855,
+      "step": 28250
+    },
+    {
+      "epoch": 8.241132273283243,
+      "grad_norm": 0.32976293563842773,
+      "learning_rate": 0.0005014020979020979,
+      "loss": 3.3962,
+      "step": 28300
+    },
+    {
+      "epoch": 8.255693400896966,
+      "grad_norm": 0.36868104338645935,
+      "learning_rate": 0.0005012272727272727,
+      "loss": 3.3907,
+      "step": 28350
+    },
+    {
+      "epoch": 8.270254528510687,
+      "grad_norm": 0.3128938376903534,
+      "learning_rate": 0.0005010524475524476,
+      "loss": 3.392,
+      "step": 28400
+    },
+    {
+      "epoch": 8.28481565612441,
+      "grad_norm": 0.32096031308174133,
+      "learning_rate": 0.0005008776223776223,
+      "loss": 3.3917,
+      "step": 28450
+    },
+    {
+      "epoch": 8.299376783738133,
+      "grad_norm": 0.3255106806755066,
+      "learning_rate": 0.0005007027972027972,
+      "loss": 3.3873,
+      "step": 28500
+    },
+    {
+      "epoch": 8.313937911351855,
+      "grad_norm": 0.33877506852149963,
+      "learning_rate": 0.000500527972027972,
+      "loss": 3.4003,
+      "step": 28550
+    },
+    {
+      "epoch": 8.328499038965578,
+      "grad_norm": 0.323604553937912,
+      "learning_rate": 0.0005003531468531468,
+      "loss": 3.4088,
+      "step": 28600
+    },
+    {
+      "epoch": 8.3430601665793,
+      "grad_norm": 0.34601426124572754,
+      "learning_rate": 0.0005001783216783217,
+      "loss": 3.4023,
+      "step": 28650
+    },
+    {
+      "epoch": 8.357621294193022,
+      "grad_norm": 0.3355138599872589,
+      "learning_rate": 0.0005000034965034965,
+      "loss": 3.4155,
+      "step": 28700
+    },
+    {
+      "epoch": 8.372182421806745,
+      "grad_norm": 0.33060452342033386,
+      "learning_rate": 0.0004998286713286713,
+      "loss": 3.4031,
+      "step": 28750
+    },
+    {
+      "epoch": 8.386743549420467,
+      "grad_norm": 0.33665457367897034,
+      "learning_rate": 0.0004996538461538461,
+      "loss": 3.3994,
+      "step": 28800
+    },
+    {
+      "epoch": 8.40130467703419,
+      "grad_norm": 0.3132437765598297,
+      "learning_rate": 0.000499479020979021,
+      "loss": 3.408,
+      "step": 28850
+    },
+    {
+      "epoch": 8.415865804647911,
+      "grad_norm": 0.35093778371810913,
+      "learning_rate": 0.0004993041958041958,
+      "loss": 3.4177,
+      "step": 28900
+    },
+    {
+      "epoch": 8.430426932261634,
+      "grad_norm": 0.3508143126964569,
+      "learning_rate": 0.0004991293706293706,
+      "loss": 3.4109,
+      "step": 28950
+    },
+    {
+      "epoch": 8.444988059875357,
+      "grad_norm": 0.3553667962551117,
+      "learning_rate": 0.0004989545454545454,
+      "loss": 3.4158,
+      "step": 29000
+    },
+    {
+      "epoch": 8.444988059875357,
+      "eval_accuracy": 0.366446409202989,
+      "eval_loss": 3.5817067623138428,
+      "eval_runtime": 180.8846,
+      "eval_samples_per_second": 92.026,
+      "eval_steps_per_second": 5.755,
+      "step": 29000
+    },
+    {
+      "epoch": 8.45954918748908,
+      "grad_norm": 0.34186291694641113,
+      "learning_rate": 0.0004987797202797203,
+      "loss": 3.4113,
+      "step": 29050
+    },
+    {
+      "epoch": 8.474110315102802,
+      "grad_norm": 0.33059263229370117,
+      "learning_rate": 0.0004986048951048951,
+      "loss": 3.4114,
+      "step": 29100
+    },
+    {
+      "epoch": 8.488671442716523,
+      "grad_norm": 0.3274041712284088,
+      "learning_rate": 0.0004984300699300699,
+      "loss": 3.3992,
+      "step": 29150
+    },
+    {
+      "epoch": 8.503232570330246,
+      "grad_norm": 0.3322159945964813,
+      "learning_rate": 0.0004982552447552448,
+      "loss": 3.4132,
+      "step": 29200
+    },
+    {
+      "epoch": 8.517793697943969,
+      "grad_norm": 0.3492433428764343,
+      "learning_rate": 0.0004980804195804195,
+      "loss": 3.4041,
+      "step": 29250
+    },
+    {
+      "epoch": 8.532354825557691,
+      "grad_norm": 0.3179911971092224,
+      "learning_rate": 0.0004979055944055944,
+      "loss": 3.421,
+      "step": 29300
+    },
+    {
+      "epoch": 8.546915953171414,
+      "grad_norm": 0.3256663978099823,
+      "learning_rate": 0.0004977307692307692,
+      "loss": 3.4057,
+      "step": 29350
+    },
+    {
+      "epoch": 8.561477080785137,
+      "grad_norm": 0.3634396195411682,
+      "learning_rate": 0.000497555944055944,
+      "loss": 3.4194,
+      "step": 29400
+    },
+    {
+      "epoch": 8.576038208398858,
+      "grad_norm": 0.3208930492401123,
+      "learning_rate": 0.0004973811188811188,
+      "loss": 3.4217,
+      "step": 29450
+    },
+    {
+      "epoch": 8.59059933601258,
+      "grad_norm": 0.3529151678085327,
+      "learning_rate": 0.0004972062937062937,
+      "loss": 3.4254,
+      "step": 29500
+    },
+    {
+      "epoch": 8.605160463626303,
+      "grad_norm": 0.33356186747550964,
+      "learning_rate": 0.0004970314685314685,
+      "loss": 3.4137,
+      "step": 29550
+    },
+    {
+      "epoch": 8.619721591240026,
+      "grad_norm": 0.3451177179813385,
+      "learning_rate": 0.0004968566433566433,
+      "loss": 3.4248,
+      "step": 29600
+    },
+    {
+      "epoch": 8.634282718853749,
+      "grad_norm": 0.32635167241096497,
+      "learning_rate": 0.0004966818181818181,
+      "loss": 3.4141,
+      "step": 29650
+    },
+    {
+      "epoch": 8.64884384646747,
+      "grad_norm": 0.3231564462184906,
+      "learning_rate": 0.000496506993006993,
+      "loss": 3.4239,
+      "step": 29700
+    },
+    {
+      "epoch": 8.663404974081192,
+      "grad_norm": 0.3018019497394562,
+      "learning_rate": 0.0004963321678321678,
+      "loss": 3.4006,
+      "step": 29750
+    },
+    {
+      "epoch": 8.677966101694915,
+      "grad_norm": 0.3364706337451935,
+      "learning_rate": 0.0004961573426573426,
+      "loss": 3.4213,
+      "step": 29800
+    },
+    {
+      "epoch": 8.692527229308638,
+      "grad_norm": 0.32945576310157776,
+      "learning_rate": 0.0004959825174825175,
+      "loss": 3.4371,
+      "step": 29850
+    },
+    {
+      "epoch": 8.70708835692236,
+      "grad_norm": 0.3232404589653015,
+      "learning_rate": 0.0004958076923076923,
+      "loss": 3.4317,
+      "step": 29900
+    },
+    {
+      "epoch": 8.721649484536082,
+      "grad_norm": 0.33957934379577637,
+      "learning_rate": 0.0004956328671328671,
+      "loss": 3.4244,
+      "step": 29950
+    },
+    {
+      "epoch": 8.736210612149804,
+      "grad_norm": 0.3381136655807495,
+      "learning_rate": 0.0004954580419580419,
+      "loss": 3.4307,
+      "step": 30000
+    },
+    {
+      "epoch": 8.736210612149804,
+      "eval_accuracy": 0.3671908156329112,
+      "eval_loss": 3.5734803676605225,
+      "eval_runtime": 180.8975,
+      "eval_samples_per_second": 92.019,
+      "eval_steps_per_second": 5.755,
+      "step": 30000
+    },
+    {
+      "epoch": 8.750771739763527,
+      "grad_norm": 0.33459433913230896,
+      "learning_rate": 0.0004952832167832167,
+      "loss": 3.4191,
+      "step": 30050
+    },
+    {
+      "epoch": 8.76533286737725,
+      "grad_norm": 0.32491710782051086,
+      "learning_rate": 0.0004951083916083915,
+      "loss": 3.436,
+      "step": 30100
+    },
+    {
+      "epoch": 8.779893994990973,
+      "grad_norm": 0.3690783977508545,
+      "learning_rate": 0.0004949335664335664,
+      "loss": 3.4343,
+      "step": 30150
+    },
+    {
+      "epoch": 8.794455122604695,
+      "grad_norm": 0.3273313343524933,
+      "learning_rate": 0.0004947587412587412,
+      "loss": 3.411,
+      "step": 30200
+    },
+    {
+      "epoch": 8.809016250218416,
+      "grad_norm": 0.3403497040271759,
+      "learning_rate": 0.000494583916083916,
+      "loss": 3.4318,
+      "step": 30250
+    },
+    {
+      "epoch": 8.82357737783214,
+      "grad_norm": 0.29834118485450745,
+      "learning_rate": 0.0004944090909090908,
+      "loss": 3.4454,
+      "step": 30300
+    },
+    {
+      "epoch": 8.838138505445862,
+      "grad_norm": 0.34332728385925293,
+      "learning_rate": 0.0004942342657342657,
+      "loss": 3.4399,
+      "step": 30350
+    },
+    {
+      "epoch": 8.852699633059585,
+      "grad_norm": 0.32075831294059753,
+      "learning_rate": 0.0004940594405594405,
+      "loss": 3.4269,
+      "step": 30400
+    },
+    {
+      "epoch": 8.867260760673307,
+      "grad_norm": 0.3218145966529846,
+      "learning_rate": 0.0004938846153846153,
+      "loss": 3.4325,
+      "step": 30450
+    },
+    {
+      "epoch": 8.881821888287028,
+      "grad_norm": 0.322322815656662,
+      "learning_rate": 0.0004937097902097901,
+      "loss": 3.4382,
+      "step": 30500
+    },
+    {
+      "epoch": 8.896383015900751,
+      "grad_norm": 0.3216952681541443,
+      "learning_rate": 0.000493534965034965,
+      "loss": 3.4371,
+      "step": 30550
+    },
+    {
+      "epoch": 8.910944143514474,
+      "grad_norm": 0.31134024262428284,
+      "learning_rate": 0.0004933601398601398,
+      "loss": 3.4368,
+      "step": 30600
+    },
+    {
+      "epoch": 8.925505271128197,
+      "grad_norm": 0.33784356713294983,
+      "learning_rate": 0.0004931853146853146,
+      "loss": 3.4262,
+      "step": 30650
+    },
+    {
+      "epoch": 8.94006639874192,
+      "grad_norm": 0.3245190680027008,
+      "learning_rate": 0.0004930104895104895,
+      "loss": 3.4206,
+      "step": 30700
+    },
+    {
+      "epoch": 8.95462752635564,
+      "grad_norm": 0.32766658067703247,
+      "learning_rate": 0.0004928356643356642,
+      "loss": 3.4303,
+      "step": 30750
+    },
+    {
+      "epoch": 8.969188653969363,
+      "grad_norm": 0.327048122882843,
+      "learning_rate": 0.0004926608391608391,
+      "loss": 3.4377,
+      "step": 30800
+    },
+    {
+      "epoch": 8.983749781583086,
+      "grad_norm": 0.3345113694667816,
+      "learning_rate": 0.0004924860139860139,
+      "loss": 3.4446,
+      "step": 30850
+    },
+    {
+      "epoch": 8.998310909196809,
+      "grad_norm": 0.31201016902923584,
+      "learning_rate": 0.0004923111888111887,
+      "loss": 3.4229,
+      "step": 30900
+    },
+    {
+      "epoch": 9.012813792300076,
+      "grad_norm": 0.31976842880249023,
+      "learning_rate": 0.0004921363636363635,
+      "loss": 3.3365,
+      "step": 30950
+    },
+    {
+      "epoch": 9.027374919913798,
+      "grad_norm": 0.31214338541030884,
+      "learning_rate": 0.0004919615384615384,
+      "loss": 3.3191,
+      "step": 31000
+    },
+    {
+      "epoch": 9.027374919913798,
+      "eval_accuracy": 0.36729027359875366,
+      "eval_loss": 3.579190492630005,
+      "eval_runtime": 180.6255,
+      "eval_samples_per_second": 92.158,
+      "eval_steps_per_second": 5.763,
+      "step": 31000
+    },
+    {
+      "epoch": 9.041936047527521,
+      "grad_norm": 0.3289821445941925,
+      "learning_rate": 0.0004917867132867132,
+      "loss": 3.3282,
+      "step": 31050
+    },
+    {
+      "epoch": 9.056497175141242,
+      "grad_norm": 0.3382094204425812,
+      "learning_rate": 0.000491611888111888,
+      "loss": 3.3278,
+      "step": 31100
+    },
+    {
+      "epoch": 9.071058302754965,
+      "grad_norm": 0.3234216570854187,
+      "learning_rate": 0.0004914370629370628,
+      "loss": 3.3318,
+      "step": 31150
+    },
+    {
+      "epoch": 9.085619430368688,
+      "grad_norm": 0.3528585135936737,
+      "learning_rate": 0.0004912622377622378,
+      "loss": 3.3381,
+      "step": 31200
+    },
+    {
+      "epoch": 9.10018055798241,
+      "grad_norm": 0.32475900650024414,
+      "learning_rate": 0.0004910874125874126,
+      "loss": 3.3417,
+      "step": 31250
+    },
+    {
+      "epoch": 9.114741685596133,
+      "grad_norm": 0.38442665338516235,
+      "learning_rate": 0.0004909125874125874,
+      "loss": 3.3351,
+      "step": 31300
+    },
+    {
+      "epoch": 9.129302813209854,
+      "grad_norm": 0.3471496105194092,
+      "learning_rate": 0.0004907377622377623,
+      "loss": 3.3419,
+      "step": 31350
+    },
+    {
+      "epoch": 9.143863940823577,
+      "grad_norm": 0.34800395369529724,
+      "learning_rate": 0.0004905629370629371,
+      "loss": 3.3577,
+      "step": 31400
+    },
+    {
+      "epoch": 9.1584250684373,
+      "grad_norm": 0.3251023292541504,
+      "learning_rate": 0.0004903881118881119,
+      "loss": 3.3453,
+      "step": 31450
+    },
+    {
+      "epoch": 9.172986196051022,
+      "grad_norm": 0.3403095602989197,
+      "learning_rate": 0.0004902132867132867,
+      "loss": 3.3494,
+      "step": 31500
+    },
+    {
+      "epoch": 9.187547323664745,
+      "grad_norm": 0.34055787324905396,
+      "learning_rate": 0.0004900384615384615,
+      "loss": 3.3536,
+      "step": 31550
+    },
+    {
+      "epoch": 9.202108451278466,
+      "grad_norm": 0.3373563587665558,
+      "learning_rate": 0.0004898636363636363,
+      "loss": 3.3497,
+      "step": 31600
+    },
+    {
+      "epoch": 9.216669578892189,
+      "grad_norm": 0.3197869062423706,
+      "learning_rate": 0.0004896888111888112,
+      "loss": 3.3548,
+      "step": 31650
+    },
+    {
+      "epoch": 9.231230706505912,
+      "grad_norm": 0.3400711715221405,
+      "learning_rate": 0.000489513986013986,
+      "loss": 3.365,
+      "step": 31700
+    },
+    {
+      "epoch": 9.245791834119634,
+      "grad_norm": 0.338986337184906,
+      "learning_rate": 0.0004893391608391608,
+      "loss": 3.3731,
+      "step": 31750
+    },
+    {
+      "epoch": 9.260352961733357,
+      "grad_norm": 0.3736262321472168,
+      "learning_rate": 0.0004891643356643356,
+      "loss": 3.3684,
+      "step": 31800
+    },
+    {
+      "epoch": 9.27491408934708,
+      "grad_norm": 0.32736754417419434,
+      "learning_rate": 0.0004889895104895105,
+      "loss": 3.3747,
+      "step": 31850
+    },
+    {
+      "epoch": 9.2894752169608,
+      "grad_norm": 0.3265363872051239,
+      "learning_rate": 0.0004888146853146853,
+      "loss": 3.3615,
+      "step": 31900
+    },
+    {
+      "epoch": 9.304036344574524,
+      "grad_norm": 0.35288384556770325,
+      "learning_rate": 0.0004886398601398601,
+      "loss": 3.3709,
+      "step": 31950
+    },
+    {
+      "epoch": 9.318597472188246,
+      "grad_norm": 0.3371695280075073,
+      "learning_rate": 0.000488465034965035,
+      "loss": 3.3673,
+      "step": 32000
+    },
+    {
+      "epoch": 9.318597472188246,
+      "eval_accuracy": 0.3674448684274567,
+      "eval_loss": 3.577155113220215,
+      "eval_runtime": 180.6862,
+      "eval_samples_per_second": 92.127,
+      "eval_steps_per_second": 5.761,
+      "step": 32000
+    },
+    {
+      "epoch": 9.333158599801969,
+      "grad_norm": 0.32006871700286865,
+      "learning_rate": 0.0004882902097902098,
+      "loss": 3.3764,
+      "step": 32050
+    },
+    {
+      "epoch": 9.347719727415692,
+      "grad_norm": 0.3183046877384186,
+      "learning_rate": 0.0004881153846153846,
+      "loss": 3.3797,
+      "step": 32100
+    },
+    {
+      "epoch": 9.362280855029413,
+      "grad_norm": 0.34943392872810364,
+      "learning_rate": 0.0004879405594405594,
+      "loss": 3.3896,
+      "step": 32150
+    },
+    {
+      "epoch": 9.376841982643136,
+      "grad_norm": 0.32475075125694275,
+      "learning_rate": 0.00048776573426573424,
+      "loss": 3.3837,
+      "step": 32200
+    },
+    {
+      "epoch": 9.391403110256858,
+      "grad_norm": 0.3197793662548065,
+      "learning_rate": 0.00048759090909090904,
+      "loss": 3.3888,
+      "step": 32250
+    },
+    {
+      "epoch": 9.405964237870581,
+      "grad_norm": 0.34314748644828796,
+      "learning_rate": 0.0004874160839160839,
+      "loss": 3.3873,
+      "step": 32300
+    },
+    {
+      "epoch": 9.420525365484304,
+      "grad_norm": 0.35972434282302856,
+      "learning_rate": 0.0004872412587412587,
+      "loss": 3.3898,
+      "step": 32350
+    },
+    {
+      "epoch": 9.435086493098025,
+      "grad_norm": 0.34308797121047974,
+      "learning_rate": 0.00048706643356643354,
+      "loss": 3.3825,
+      "step": 32400
+    },
+    {
+      "epoch": 9.449647620711747,
+      "grad_norm": 0.3321487307548523,
+      "learning_rate": 0.00048689160839160834,
+      "loss": 3.3777,
+      "step": 32450
+    },
+    {
+      "epoch": 9.46420874832547,
+      "grad_norm": 0.3517487049102783,
+      "learning_rate": 0.0004867167832167832,
+      "loss": 3.3981,
+      "step": 32500
+    },
+    {
+      "epoch": 9.478769875939193,
+      "grad_norm": 0.3309260606765747,
+      "learning_rate": 0.00048654195804195794,
+      "loss": 3.3946,
+      "step": 32550
+    },
+    {
+      "epoch": 9.493331003552916,
+      "grad_norm": 0.32394567131996155,
+      "learning_rate": 0.00048636713286713285,
+      "loss": 3.3976,
+      "step": 32600
+    },
+    {
+      "epoch": 9.507892131166638,
+      "grad_norm": 0.3531663119792938,
+      "learning_rate": 0.0004861923076923077,
+      "loss": 3.3897,
+      "step": 32650
+    },
+    {
+      "epoch": 9.52245325878036,
+      "grad_norm": 0.327006995677948,
+      "learning_rate": 0.00048601748251748245,
+      "loss": 3.3908,
+      "step": 32700
+    },
+    {
+      "epoch": 9.537014386394082,
+      "grad_norm": 0.3265383839607239,
+      "learning_rate": 0.0004858426573426573,
+      "loss": 3.3903,
+      "step": 32750
+    },
+    {
+      "epoch": 9.551575514007805,
+      "grad_norm": 0.34400323033332825,
+      "learning_rate": 0.0004856678321678321,
+      "loss": 3.3941,
+      "step": 32800
+    },
+    {
+      "epoch": 9.566136641621528,
+      "grad_norm": 0.3338717520236969,
+      "learning_rate": 0.00048549300699300696,
+      "loss": 3.3895,
+      "step": 32850
+    },
+    {
+      "epoch": 9.58069776923525,
+      "grad_norm": 0.3492511510848999,
+      "learning_rate": 0.00048531818181818176,
+      "loss": 3.4021,
+      "step": 32900
+    },
+    {
+      "epoch": 9.595258896848971,
+      "grad_norm": 0.346670538187027,
+      "learning_rate": 0.0004851433566433566,
+      "loss": 3.4054,
+      "step": 32950
+    },
+    {
+      "epoch": 9.609820024462694,
+      "grad_norm": 0.32477083802223206,
+      "learning_rate": 0.0004849685314685314,
+      "loss": 3.4042,
+      "step": 33000
+    },
+    {
+      "epoch": 9.609820024462694,
+      "eval_accuracy": 0.36810098533923746,
+      "eval_loss": 3.5701539516448975,
+      "eval_runtime": 180.6914,
+      "eval_samples_per_second": 92.124,
+      "eval_steps_per_second": 5.761,
+      "step": 33000
+    },
+    {
+      "epoch": 9.624381152076417,
+      "grad_norm": 0.335056871175766,
+      "learning_rate": 0.00048479370629370627,
+      "loss": 3.3859,
+      "step": 33050
+    },
+    {
+      "epoch": 9.63894227969014,
+      "grad_norm": 0.3331140875816345,
+      "learning_rate": 0.00048461888111888106,
+      "loss": 3.3992,
+      "step": 33100
+    },
+    {
+      "epoch": 9.653503407303862,
+      "grad_norm": 0.3186609148979187,
+      "learning_rate": 0.0004844440559440559,
+      "loss": 3.3985,
+      "step": 33150
+    },
+    {
+      "epoch": 9.668064534917583,
+      "grad_norm": 0.3806893527507782,
+      "learning_rate": 0.0004842692307692307,
+      "loss": 3.3957,
+      "step": 33200
+    },
+    {
+      "epoch": 9.682625662531306,
+      "grad_norm": 0.3387463688850403,
+      "learning_rate": 0.00048409440559440557,
+      "loss": 3.407,
+      "step": 33250
+    },
+    {
+      "epoch": 9.697186790145029,
+      "grad_norm": 0.3401879370212555,
+      "learning_rate": 0.0004839195804195803,
+      "loss": 3.4044,
+      "step": 33300
+    },
+    {
+      "epoch": 9.711747917758752,
+      "grad_norm": 0.34021636843681335,
+      "learning_rate": 0.0004837447552447552,
+      "loss": 3.3997,
+      "step": 33350
+    },
+    {
+      "epoch": 9.726309045372474,
+      "grad_norm": 0.33192554116249084,
+      "learning_rate": 0.0004835699300699301,
+      "loss": 3.4043,
+      "step": 33400
+    },
+    {
+      "epoch": 9.740870172986195,
+      "grad_norm": 0.34780585765838623,
+      "learning_rate": 0.0004833951048951048,
+      "loss": 3.3918,
+      "step": 33450
+    },
+    {
+      "epoch": 9.755431300599918,
+      "grad_norm": 0.3295852243900299,
+      "learning_rate": 0.0004832202797202797,
+      "loss": 3.399,
+      "step": 33500
+    },
+    {
+      "epoch": 9.76999242821364,
+      "grad_norm": 0.3357504904270172,
+      "learning_rate": 0.0004830454545454545,
+      "loss": 3.403,
+      "step": 33550
+    },
+    {
+      "epoch": 9.784553555827364,
+      "grad_norm": 0.34999728202819824,
+      "learning_rate": 0.00048287062937062933,
+      "loss": 3.4003,
+      "step": 33600
+    },
+    {
+      "epoch": 9.799114683441086,
+      "grad_norm": 0.33418694138526917,
+      "learning_rate": 0.00048269580419580413,
+      "loss": 3.408,
+      "step": 33650
+    },
+    {
+      "epoch": 9.813675811054807,
+      "grad_norm": 0.34634068608283997,
+      "learning_rate": 0.000482520979020979,
+      "loss": 3.4006,
+      "step": 33700
+    },
+    {
+      "epoch": 9.82823693866853,
+      "grad_norm": 0.3158990144729614,
+      "learning_rate": 0.0004823461538461538,
+      "loss": 3.3964,
+      "step": 33750
+    },
+    {
+      "epoch": 9.842798066282253,
+      "grad_norm": 0.3415282666683197,
+      "learning_rate": 0.00048217132867132864,
+      "loss": 3.4045,
+      "step": 33800
+    },
+    {
+      "epoch": 9.857359193895975,
+      "grad_norm": 0.32026195526123047,
+      "learning_rate": 0.00048199650349650344,
+      "loss": 3.4094,
+      "step": 33850
+    },
+    {
+      "epoch": 9.871920321509698,
+      "grad_norm": 0.3347398638725281,
+      "learning_rate": 0.0004818216783216783,
+      "loss": 3.4243,
+      "step": 33900
+    },
+    {
+      "epoch": 9.88648144912342,
+      "grad_norm": 0.33100831508636475,
+      "learning_rate": 0.0004816468531468531,
+      "loss": 3.4099,
+      "step": 33950
+    },
+    {
+      "epoch": 9.901042576737142,
+      "grad_norm": 0.34802931547164917,
+      "learning_rate": 0.00048147202797202795,
+      "loss": 3.4059,
+      "step": 34000
+    },
+    {
+      "epoch": 9.901042576737142,
+      "eval_accuracy": 0.3684669577360075,
+      "eval_loss": 3.563434362411499,
+      "eval_runtime": 180.8937,
+      "eval_samples_per_second": 92.021,
+      "eval_steps_per_second": 5.755,
+      "step": 34000
+    },
+    {
+      "epoch": 9.915603704350865,
+      "grad_norm": 0.31292399764060974,
+      "learning_rate": 0.0004812972027972028,
+      "loss": 3.3956,
+      "step": 34050
+    },
+    {
+      "epoch": 9.930164831964587,
+      "grad_norm": 0.3332271873950958,
+      "learning_rate": 0.0004811223776223776,
+      "loss": 3.4103,
+      "step": 34100
+    },
+    {
+      "epoch": 9.94472595957831,
+      "grad_norm": 0.3372020721435547,
+      "learning_rate": 0.00048094755244755245,
+      "loss": 3.4071,
+      "step": 34150
+    },
+    {
+      "epoch": 9.959287087192033,
+      "grad_norm": 0.35507315397262573,
+      "learning_rate": 0.0004807727272727272,
+      "loss": 3.3987,
+      "step": 34200
+    },
+    {
+      "epoch": 9.973848214805754,
+      "grad_norm": 0.31724709272384644,
+      "learning_rate": 0.00048059790209790205,
+      "loss": 3.4112,
+      "step": 34250
+    },
+    {
+      "epoch": 9.988409342419477,
+      "grad_norm": 0.33160942792892456,
+      "learning_rate": 0.00048042307692307685,
+      "loss": 3.4104,
+      "step": 34300
+    },
+    {
+      "epoch": 10.002912225522744,
+      "grad_norm": 0.3155555725097656,
+      "learning_rate": 0.0004802482517482517,
+      "loss": 3.39,
+      "step": 34350
+    },
+    {
+      "epoch": 10.017473353136467,
+      "grad_norm": 0.364051878452301,
+      "learning_rate": 0.0004800734265734265,
+      "loss": 3.3051,
+      "step": 34400
+    },
+    {
+      "epoch": 10.03203448075019,
+      "grad_norm": 0.3165052533149719,
+      "learning_rate": 0.00047989860139860136,
+      "loss": 3.3003,
+      "step": 34450
+    },
+    {
+      "epoch": 10.046595608363912,
+      "grad_norm": 0.3623712956905365,
+      "learning_rate": 0.00047972377622377616,
+      "loss": 3.312,
+      "step": 34500
+    },
+    {
+      "epoch": 10.061156735977635,
+      "grad_norm": 0.33263257145881653,
+      "learning_rate": 0.000479548951048951,
+      "loss": 3.3194,
+      "step": 34550
+    },
+    {
+      "epoch": 10.075717863591356,
+      "grad_norm": 0.3498550355434418,
+      "learning_rate": 0.0004793741258741258,
+      "loss": 3.3114,
+      "step": 34600
+    },
+    {
+      "epoch": 10.090278991205079,
+      "grad_norm": 0.32290056347846985,
+      "learning_rate": 0.00047919930069930067,
+      "loss": 3.3143,
+      "step": 34650
+    },
+    {
+      "epoch": 10.104840118818801,
+      "grad_norm": 0.3660222589969635,
+      "learning_rate": 0.0004790244755244755,
+      "loss": 3.3239,
+      "step": 34700
+    },
+    {
+      "epoch": 10.119401246432524,
+      "grad_norm": 0.3473450839519501,
+      "learning_rate": 0.0004788496503496503,
+      "loss": 3.3249,
+      "step": 34750
+    },
+    {
+      "epoch": 10.133962374046247,
+      "grad_norm": 0.3577515482902527,
+      "learning_rate": 0.0004786748251748252,
+      "loss": 3.3254,
+      "step": 34800
+    },
+    {
+      "epoch": 10.148523501659968,
+      "grad_norm": 0.33307579159736633,
+      "learning_rate": 0.0004785,
+      "loss": 3.3295,
+      "step": 34850
+    },
+    {
+      "epoch": 10.16308462927369,
+      "grad_norm": 0.3364368677139282,
+      "learning_rate": 0.00047832517482517483,
+      "loss": 3.3365,
+      "step": 34900
+    },
+    {
+      "epoch": 10.177645756887413,
+      "grad_norm": 0.3440118432044983,
+      "learning_rate": 0.0004781503496503496,
+      "loss": 3.3399,
+      "step": 34950
+    },
+    {
+      "epoch": 10.192206884501136,
+      "grad_norm": 0.34188905358314514,
+      "learning_rate": 0.00047797552447552443,
+      "loss": 3.3375,
+      "step": 35000
+    },
+    {
+      "epoch": 10.192206884501136,
+      "eval_accuracy": 0.36812226417117305,
+      "eval_loss": 3.572165012359619,
+      "eval_runtime": 180.6318,
+      "eval_samples_per_second": 92.154,
+      "eval_steps_per_second": 5.763,
+      "step": 35000
+    },
+    {
+      "epoch": 10.206768012114859,
+      "grad_norm": 0.3428874909877777,
+      "learning_rate": 0.00047780069930069923,
+      "loss": 3.3332,
+      "step": 35050
+    },
+    {
+      "epoch": 10.221329139728581,
+      "grad_norm": 0.3249208927154541,
+      "learning_rate": 0.0004776258741258741,
+      "loss": 3.3323,
+      "step": 35100
+    },
+    {
+      "epoch": 10.235890267342302,
+      "grad_norm": 0.3354872167110443,
+      "learning_rate": 0.0004774510489510489,
+      "loss": 3.3385,
+      "step": 35150
+    },
+    {
+      "epoch": 10.250451394956025,
+      "grad_norm": 0.33717766404151917,
+      "learning_rate": 0.00047727622377622374,
+      "loss": 3.3466,
+      "step": 35200
+    },
+    {
+      "epoch": 10.265012522569748,
+      "grad_norm": 0.37344837188720703,
+      "learning_rate": 0.00047710139860139854,
+      "loss": 3.3522,
+      "step": 35250
+    },
+    {
+      "epoch": 10.27957365018347,
+      "grad_norm": 0.3368532061576843,
+      "learning_rate": 0.0004769265734265734,
+      "loss": 3.3487,
+      "step": 35300
+    },
+    {
+      "epoch": 10.294134777797193,
+      "grad_norm": 0.3481936752796173,
+      "learning_rate": 0.0004767517482517482,
+      "loss": 3.3462,
+      "step": 35350
+    },
+    {
+      "epoch": 10.308695905410914,
+      "grad_norm": 0.3315824866294861,
+      "learning_rate": 0.00047657692307692304,
+      "loss": 3.3595,
+      "step": 35400
+    },
+    {
+      "epoch": 10.323257033024637,
+      "grad_norm": 0.329924076795578,
+      "learning_rate": 0.0004764020979020979,
+      "loss": 3.3562,
+      "step": 35450
+    },
+    {
+      "epoch": 10.33781816063836,
+      "grad_norm": 0.356904000043869,
+      "learning_rate": 0.0004762272727272727,
+      "loss": 3.3541,
+      "step": 35500
+    },
+    {
+      "epoch": 10.352379288252083,
+      "grad_norm": 0.33224600553512573,
+      "learning_rate": 0.00047605244755244755,
+      "loss": 3.3563,
+      "step": 35550
+    },
+    {
+      "epoch": 10.366940415865805,
+      "grad_norm": 0.33060455322265625,
+      "learning_rate": 0.00047587762237762235,
+      "loss": 3.3526,
+      "step": 35600
+    },
+    {
+      "epoch": 10.381501543479526,
+      "grad_norm": 0.3506428599357605,
+      "learning_rate": 0.0004757027972027972,
+      "loss": 3.3612,
+      "step": 35650
+    },
+    {
+      "epoch": 10.396062671093249,
+      "grad_norm": 0.3286277651786804,
+      "learning_rate": 0.00047552797202797195,
+      "loss": 3.3555,
+      "step": 35700
+    },
+    {
+      "epoch": 10.410623798706972,
+      "grad_norm": 0.3427405059337616,
+      "learning_rate": 0.0004753531468531468,
+      "loss": 3.3529,
+      "step": 35750
+    },
+    {
+      "epoch": 10.425184926320695,
+      "grad_norm": 0.36271488666534424,
+      "learning_rate": 0.0004751783216783216,
+      "loss": 3.3544,
+      "step": 35800
+    },
+    {
+      "epoch": 10.439746053934417,
+      "grad_norm": 0.3364131450653076,
+      "learning_rate": 0.00047500349650349646,
+      "loss": 3.3693,
+      "step": 35850
+    },
+    {
+      "epoch": 10.454307181548138,
+      "grad_norm": 0.34665197134017944,
+      "learning_rate": 0.00047482867132867126,
+      "loss": 3.3763,
+      "step": 35900
+    },
+    {
+      "epoch": 10.468868309161861,
+      "grad_norm": 0.4717924892902374,
+      "learning_rate": 0.0004746538461538461,
+      "loss": 3.3666,
+      "step": 35950
+    },
+    {
+      "epoch": 10.483429436775584,
+      "grad_norm": 0.32863062620162964,
+      "learning_rate": 0.0004744790209790209,
+      "loss": 3.3636,
+      "step": 36000
+    },
+    {
+      "epoch": 10.483429436775584,
+      "eval_accuracy": 0.36863377907587797,
+      "eval_loss": 3.5674543380737305,
+      "eval_runtime": 180.888,
+      "eval_samples_per_second": 92.024,
+      "eval_steps_per_second": 5.755,
+      "step": 36000
+    },
+    {
+      "epoch": 10.497990564389307,
+      "grad_norm": 0.3689253628253937,
+      "learning_rate": 0.00047430419580419576,
+      "loss": 3.3647,
+      "step": 36050
+    },
+    {
+      "epoch": 10.51255169200303,
+      "grad_norm": 0.33551663160324097,
+      "learning_rate": 0.0004741293706293706,
+      "loss": 3.3471,
+      "step": 36100
+    },
+    {
+      "epoch": 10.52711281961675,
+      "grad_norm": 0.32269003987312317,
+      "learning_rate": 0.0004739545454545454,
+      "loss": 3.3725,
+      "step": 36150
+    },
+    {
+      "epoch": 10.541673947230473,
+      "grad_norm": 0.3469913899898529,
+      "learning_rate": 0.00047377972027972027,
+      "loss": 3.3801,
+      "step": 36200
+    },
+    {
+      "epoch": 10.556235074844196,
+      "grad_norm": 0.32835647463798523,
+      "learning_rate": 0.00047360489510489507,
+      "loss": 3.3726,
+      "step": 36250
+    },
+    {
+      "epoch": 10.570796202457919,
+      "grad_norm": 0.34614914655685425,
+      "learning_rate": 0.0004734300699300699,
+      "loss": 3.3698,
+      "step": 36300
+    },
+    {
+      "epoch": 10.585357330071641,
+      "grad_norm": 0.3445773720741272,
+      "learning_rate": 0.0004732552447552447,
+      "loss": 3.3589,
+      "step": 36350
+    },
+    {
+      "epoch": 10.599918457685362,
+      "grad_norm": 0.33179157972335815,
+      "learning_rate": 0.0004730804195804196,
+      "loss": 3.3738,
+      "step": 36400
+    },
+    {
+      "epoch": 10.614479585299085,
+      "grad_norm": 0.34509310126304626,
+      "learning_rate": 0.0004729055944055943,
+      "loss": 3.37,
+      "step": 36450
+    },
+    {
+      "epoch": 10.629040712912808,
+      "grad_norm": 0.34073859453201294,
+      "learning_rate": 0.0004727307692307692,
+      "loss": 3.3782,
+      "step": 36500
+    },
+    {
+      "epoch": 10.64360184052653,
+      "grad_norm": 0.33959853649139404,
+      "learning_rate": 0.000472555944055944,
+      "loss": 3.3636,
+      "step": 36550
+    },
+    {
+      "epoch": 10.658162968140253,
+      "grad_norm": 0.3635091483592987,
+      "learning_rate": 0.00047238111888111883,
+      "loss": 3.3765,
+      "step": 36600
+    },
+    {
+      "epoch": 10.672724095753976,
+      "grad_norm": 0.3816806375980377,
+      "learning_rate": 0.00047220629370629363,
+      "loss": 3.3727,
+      "step": 36650
+    },
+    {
+      "epoch": 10.687285223367697,
+      "grad_norm": 0.44146403670310974,
+      "learning_rate": 0.0004720314685314685,
+      "loss": 3.3843,
+      "step": 36700
+    },
+    {
+      "epoch": 10.70184635098142,
+      "grad_norm": 0.3365342319011688,
+      "learning_rate": 0.0004718566433566433,
+      "loss": 3.3746,
+      "step": 36750
+    },
+    {
+      "epoch": 10.716407478595142,
+      "grad_norm": 0.3512522578239441,
+      "learning_rate": 0.00047168181818181814,
+      "loss": 3.3824,
+      "step": 36800
+    },
+    {
+      "epoch": 10.730968606208865,
+      "grad_norm": 0.34449753165245056,
+      "learning_rate": 0.000471506993006993,
+      "loss": 3.3708,
+      "step": 36850
+    },
+    {
+      "epoch": 10.745529733822588,
+      "grad_norm": 0.3496144115924835,
+      "learning_rate": 0.0004713321678321678,
+      "loss": 3.3733,
+      "step": 36900
+    },
+    {
+      "epoch": 10.760090861436309,
+      "grad_norm": 0.34519726037979126,
+      "learning_rate": 0.00047115734265734265,
+      "loss": 3.3864,
+      "step": 36950
+    },
+    {
+      "epoch": 10.774651989050032,
+      "grad_norm": 0.362666517496109,
+      "learning_rate": 0.00047098251748251745,
+      "loss": 3.3783,
+      "step": 37000
+    },
+    {
+      "epoch": 10.774651989050032,
+      "eval_accuracy": 0.36897800239028294,
+      "eval_loss": 3.558612108230591,
+      "eval_runtime": 181.139,
+      "eval_samples_per_second": 91.896,
+      "eval_steps_per_second": 5.747,
+      "step": 37000
+    },
+    {
+      "epoch": 10.789213116663754,
+      "grad_norm": 0.31553319096565247,
+      "learning_rate": 0.0004708076923076923,
+      "loss": 3.3889,
+      "step": 37050
+    },
+    {
+      "epoch": 10.803774244277477,
+      "grad_norm": 0.33135709166526794,
+      "learning_rate": 0.0004706328671328671,
+      "loss": 3.3838,
+      "step": 37100
+    },
+    {
+      "epoch": 10.8183353718912,
+      "grad_norm": 0.3278547525405884,
+      "learning_rate": 0.00047045804195804195,
+      "loss": 3.3758,
+      "step": 37150
+    },
+    {
+      "epoch": 10.83289649950492,
+      "grad_norm": 0.3490781784057617,
+      "learning_rate": 0.0004702832167832167,
+      "loss": 3.396,
+      "step": 37200
+    },
+    {
+      "epoch": 10.847457627118644,
+      "grad_norm": 0.35344281792640686,
+      "learning_rate": 0.00047010839160839155,
+      "loss": 3.3873,
+      "step": 37250
+    },
+    {
+      "epoch": 10.862018754732366,
+      "grad_norm": 0.34710001945495605,
+      "learning_rate": 0.00046993356643356635,
+      "loss": 3.3902,
+      "step": 37300
+    },
+    {
+      "epoch": 10.876579882346089,
+      "grad_norm": 0.36521968245506287,
+      "learning_rate": 0.0004697587412587412,
+      "loss": 3.3968,
+      "step": 37350
+    },
+    {
+      "epoch": 10.891141009959812,
+      "grad_norm": 0.33776748180389404,
+      "learning_rate": 0.000469583916083916,
+      "loss": 3.3824,
+      "step": 37400
+    },
+    {
+      "epoch": 10.905702137573535,
+      "grad_norm": 0.3304058015346527,
+      "learning_rate": 0.00046940909090909086,
+      "loss": 3.3784,
+      "step": 37450
+    },
+    {
+      "epoch": 10.920263265187256,
+      "grad_norm": 0.3333076536655426,
+      "learning_rate": 0.0004692342657342657,
+      "loss": 3.3872,
+      "step": 37500
+    },
+    {
+      "epoch": 10.934824392800978,
+      "grad_norm": 0.3497754633426666,
+      "learning_rate": 0.0004690594405594405,
+      "loss": 3.3924,
+      "step": 37550
+    },
+    {
+      "epoch": 10.949385520414701,
+      "grad_norm": 0.3230236768722534,
+      "learning_rate": 0.00046888461538461537,
+      "loss": 3.3794,
+      "step": 37600
+    },
+    {
+      "epoch": 10.963946648028424,
+      "grad_norm": 0.34919098019599915,
+      "learning_rate": 0.00046870979020979017,
+      "loss": 3.3816,
+      "step": 37650
+    },
+    {
+      "epoch": 10.978507775642147,
+      "grad_norm": 0.32327717542648315,
+      "learning_rate": 0.000468534965034965,
+      "loss": 3.3837,
+      "step": 37700
+    },
+    {
+      "epoch": 10.993068903255867,
+      "grad_norm": 0.3353258967399597,
+      "learning_rate": 0.0004683601398601398,
+      "loss": 3.3962,
+      "step": 37750
+    },
+    {
+      "epoch": 11.007571786359136,
+      "grad_norm": 0.3746544122695923,
+      "learning_rate": 0.0004681853146853147,
+      "loss": 3.3268,
+      "step": 37800
+    },
+    {
+      "epoch": 11.022132913972857,
+      "grad_norm": 0.37094318866729736,
+      "learning_rate": 0.0004680104895104895,
+      "loss": 3.2658,
+      "step": 37850
+    },
+    {
+      "epoch": 11.03669404158658,
+      "grad_norm": 0.3441932797431946,
+      "learning_rate": 0.00046783566433566433,
+      "loss": 3.2796,
+      "step": 37900
+    },
+    {
+      "epoch": 11.051255169200303,
+      "grad_norm": 0.31547367572784424,
+      "learning_rate": 0.0004676608391608391,
+      "loss": 3.2876,
+      "step": 37950
+    },
+    {
+      "epoch": 11.065816296814026,
+      "grad_norm": 0.36404022574424744,
+      "learning_rate": 0.00046748601398601393,
+      "loss": 3.2792,
+      "step": 38000
+    },
+    {
+      "epoch": 11.065816296814026,
+      "eval_accuracy": 0.3686197891255999,
+      "eval_loss": 3.572824001312256,
+      "eval_runtime": 180.8426,
+      "eval_samples_per_second": 92.047,
+      "eval_steps_per_second": 5.756,
+      "step": 38000
+    },
+    {
+      "epoch": 11.080377424427748,
+      "grad_norm": 0.35687056183815,
+      "learning_rate": 0.00046731118881118873,
+      "loss": 3.2936,
+      "step": 38050
+    },
+    {
+      "epoch": 11.09493855204147,
+      "grad_norm": 0.33898624777793884,
+      "learning_rate": 0.0004671363636363636,
+      "loss": 3.2977,
+      "step": 38100
+    },
+    {
+      "epoch": 11.109499679655192,
+      "grad_norm": 0.33681026101112366,
+      "learning_rate": 0.00046696153846153844,
+      "loss": 3.3047,
+      "step": 38150
+    },
+    {
+      "epoch": 11.124060807268915,
+      "grad_norm": 0.3504103124141693,
+      "learning_rate": 0.00046678671328671324,
+      "loss": 3.3054,
+      "step": 38200
+    },
+    {
+      "epoch": 11.138621934882638,
+      "grad_norm": 0.35253238677978516,
+      "learning_rate": 0.0004666118881118881,
+      "loss": 3.3048,
+      "step": 38250
+    },
+    {
+      "epoch": 11.15318306249636,
+      "grad_norm": 0.37646666169166565,
+      "learning_rate": 0.0004664370629370629,
+      "loss": 3.2982,
+      "step": 38300
+    },
+    {
+      "epoch": 11.167744190110081,
+      "grad_norm": 0.32107439637184143,
+      "learning_rate": 0.00046626223776223774,
+      "loss": 3.2995,
+      "step": 38350
+    },
+    {
+      "epoch": 11.182305317723804,
+      "grad_norm": 0.37649980187416077,
+      "learning_rate": 0.00046608741258741254,
+      "loss": 3.304,
+      "step": 38400
+    },
+    {
+      "epoch": 11.196866445337527,
+      "grad_norm": 0.365675687789917,
+      "learning_rate": 0.0004659125874125874,
+      "loss": 3.3121,
+      "step": 38450
+    },
+    {
+      "epoch": 11.21142757295125,
+      "grad_norm": 0.3648703396320343,
+      "learning_rate": 0.0004657377622377622,
+      "loss": 3.3139,
+      "step": 38500
+    },
+    {
+      "epoch": 11.225988700564972,
+      "grad_norm": 0.34117260575294495,
+      "learning_rate": 0.00046556293706293705,
+      "loss": 3.3322,
+      "step": 38550
+    },
+    {
+      "epoch": 11.240549828178693,
+      "grad_norm": 0.3431043028831482,
+      "learning_rate": 0.00046538811188811185,
+      "loss": 3.312,
+      "step": 38600
+    },
+    {
+      "epoch": 11.255110955792416,
+      "grad_norm": 0.34056955575942993,
+      "learning_rate": 0.0004652132867132867,
+      "loss": 3.3255,
+      "step": 38650
+    },
+    {
+      "epoch": 11.269672083406139,
+      "grad_norm": 0.3726236820220947,
+      "learning_rate": 0.00046503846153846145,
+      "loss": 3.3273,
+      "step": 38700
+    },
+    {
+      "epoch": 11.284233211019862,
+      "grad_norm": 0.3636229336261749,
+      "learning_rate": 0.0004648636363636363,
+      "loss": 3.3231,
+      "step": 38750
+    },
+    {
+      "epoch": 11.298794338633584,
+      "grad_norm": 0.3742581903934479,
+      "learning_rate": 0.0004646888111888111,
+      "loss": 3.3272,
+      "step": 38800
+    },
+    {
+      "epoch": 11.313355466247307,
+      "grad_norm": 0.3781585693359375,
+      "learning_rate": 0.00046451398601398596,
+      "loss": 3.3396,
+      "step": 38850
+    },
+    {
+      "epoch": 11.327916593861028,
+      "grad_norm": 0.349326491355896,
+      "learning_rate": 0.0004643391608391608,
+      "loss": 3.3466,
+      "step": 38900
+    },
+    {
+      "epoch": 11.34247772147475,
+      "grad_norm": 0.38100987672805786,
+      "learning_rate": 0.0004641643356643356,
+      "loss": 3.3351,
+      "step": 38950
+    },
+    {
+      "epoch": 11.357038849088473,
+      "grad_norm": 0.34632235765457153,
+      "learning_rate": 0.00046398951048951046,
+      "loss": 3.3318,
+      "step": 39000
+    },
+    {
+      "epoch": 11.357038849088473,
+      "eval_accuracy": 0.36974063102434884,
+      "eval_loss": 3.5641088485717773,
+      "eval_runtime": 180.754,
+      "eval_samples_per_second": 92.092,
+      "eval_steps_per_second": 5.759,
+      "step": 39000
+    },
+    {
+      "epoch": 11.371599976702196,
+      "grad_norm": 0.36289188265800476,
+      "learning_rate": 0.00046381468531468526,
+      "loss": 3.3394,
+      "step": 39050
+    },
+    {
+      "epoch": 11.386161104315919,
+      "grad_norm": 0.32132551074028015,
+      "learning_rate": 0.0004636398601398601,
+      "loss": 3.3275,
+      "step": 39100
+    },
+    {
+      "epoch": 11.40072223192964,
+      "grad_norm": 0.3607240617275238,
+      "learning_rate": 0.0004634650349650349,
+      "loss": 3.3365,
+      "step": 39150
+    },
+    {
+      "epoch": 11.415283359543363,
+      "grad_norm": 0.3579674959182739,
+      "learning_rate": 0.00046329020979020977,
+      "loss": 3.3283,
+      "step": 39200
+    },
+    {
+      "epoch": 11.429844487157085,
+      "grad_norm": 0.35449928045272827,
+      "learning_rate": 0.00046311538461538457,
+      "loss": 3.3373,
+      "step": 39250
+    },
+    {
+      "epoch": 11.444405614770808,
+      "grad_norm": 0.3389507234096527,
+      "learning_rate": 0.0004629405594405594,
+      "loss": 3.3359,
+      "step": 39300
+    },
+    {
+      "epoch": 11.458966742384531,
+      "grad_norm": 0.35343387722969055,
+      "learning_rate": 0.0004627657342657342,
+      "loss": 3.3483,
+      "step": 39350
+    },
+    {
+      "epoch": 11.473527869998252,
+      "grad_norm": 0.3741236627101898,
+      "learning_rate": 0.0004625909090909091,
+      "loss": 3.3399,
+      "step": 39400
+    },
+    {
+      "epoch": 11.488088997611975,
+      "grad_norm": 0.3604460656642914,
+      "learning_rate": 0.0004624160839160838,
+      "loss": 3.3536,
+      "step": 39450
+    },
+    {
+      "epoch": 11.502650125225697,
+      "grad_norm": 0.374704509973526,
+      "learning_rate": 0.0004622412587412587,
+      "loss": 3.3552,
+      "step": 39500
+    },
+    {
+      "epoch": 11.51721125283942,
+      "grad_norm": 0.34912702441215515,
+      "learning_rate": 0.00046206643356643353,
+      "loss": 3.3618,
+      "step": 39550
+    },
+    {
+      "epoch": 11.531772380453143,
+      "grad_norm": 0.35853371024131775,
+      "learning_rate": 0.00046189160839160833,
+      "loss": 3.3609,
+      "step": 39600
+    },
+    {
+      "epoch": 11.546333508066864,
+      "grad_norm": 0.34523260593414307,
+      "learning_rate": 0.0004617167832167832,
+      "loss": 3.3525,
+      "step": 39650
+    },
+    {
+      "epoch": 11.560894635680587,
+      "grad_norm": 0.32813966274261475,
+      "learning_rate": 0.000461541958041958,
+      "loss": 3.3592,
+      "step": 39700
+    },
+    {
+      "epoch": 11.57545576329431,
+      "grad_norm": 0.3415012061595917,
+      "learning_rate": 0.00046136713286713284,
+      "loss": 3.342,
+      "step": 39750
+    },
+    {
+      "epoch": 11.590016890908032,
+      "grad_norm": 0.34844455122947693,
+      "learning_rate": 0.00046119230769230764,
+      "loss": 3.3556,
+      "step": 39800
+    },
+    {
+      "epoch": 11.604578018521755,
+      "grad_norm": 0.37255018949508667,
+      "learning_rate": 0.0004610174825174825,
+      "loss": 3.3545,
+      "step": 39850
+    },
+    {
+      "epoch": 11.619139146135478,
+      "grad_norm": 0.3670998215675354,
+      "learning_rate": 0.0004608426573426573,
+      "loss": 3.359,
+      "step": 39900
+    },
+    {
+      "epoch": 11.633700273749199,
+      "grad_norm": 0.32997700572013855,
+      "learning_rate": 0.00046066783216783215,
+      "loss": 3.3547,
+      "step": 39950
+    },
+    {
+      "epoch": 11.648261401362921,
+      "grad_norm": 0.3327796757221222,
+      "learning_rate": 0.00046049300699300695,
+      "loss": 3.3586,
+      "step": 40000
+    },
+    {
+      "epoch": 11.648261401362921,
+      "eval_accuracy": 0.3697354582696242,
+      "eval_loss": 3.558627128601074,
+      "eval_runtime": 180.8619,
+      "eval_samples_per_second": 92.037,
+      "eval_steps_per_second": 5.756,
+      "step": 40000
+    },
+    {
+      "epoch": 11.662822528976644,
+      "grad_norm": 0.3510940670967102,
+      "learning_rate": 0.0004603181818181818,
+      "loss": 3.361,
+      "step": 40050
+    },
+    {
+      "epoch": 11.677383656590367,
+      "grad_norm": 0.3497365117073059,
+      "learning_rate": 0.0004601433566433566,
+      "loss": 3.3581,
+      "step": 40100
+    },
+    {
+      "epoch": 11.69194478420409,
+      "grad_norm": 0.35127291083335876,
+      "learning_rate": 0.00045996853146853145,
+      "loss": 3.3681,
+      "step": 40150
+    },
+    {
+      "epoch": 11.70650591181781,
+      "grad_norm": 0.3371525704860687,
+      "learning_rate": 0.0004597937062937062,
+      "loss": 3.3521,
+      "step": 40200
+    },
+    {
+      "epoch": 11.721067039431533,
+      "grad_norm": 0.3377763330936432,
+      "learning_rate": 0.00045961888111888105,
+      "loss": 3.3748,
+      "step": 40250
+    },
+    {
+      "epoch": 11.735628167045256,
+      "grad_norm": 0.3272308111190796,
+      "learning_rate": 0.0004594440559440559,
+      "loss": 3.3714,
+      "step": 40300
+    },
+    {
+      "epoch": 11.750189294658979,
+      "grad_norm": 0.3513683080673218,
+      "learning_rate": 0.0004592692307692307,
+      "loss": 3.3533,
+      "step": 40350
+    },
+    {
+      "epoch": 11.764750422272702,
+      "grad_norm": 0.34825000166893005,
+      "learning_rate": 0.00045909440559440556,
+      "loss": 3.372,
+      "step": 40400
+    },
+    {
+      "epoch": 11.779311549886422,
+      "grad_norm": 0.3494202494621277,
+      "learning_rate": 0.00045891958041958036,
+      "loss": 3.3673,
+      "step": 40450
+    },
+    {
+      "epoch": 11.793872677500145,
+      "grad_norm": 0.357833594083786,
+      "learning_rate": 0.0004587447552447552,
+      "loss": 3.3581,
+      "step": 40500
+    },
+    {
+      "epoch": 11.808433805113868,
+      "grad_norm": 0.32943588495254517,
+      "learning_rate": 0.00045856993006993,
+      "loss": 3.3642,
+      "step": 40550
+    },
+    {
+      "epoch": 11.82299493272759,
+      "grad_norm": 0.3362137973308563,
+      "learning_rate": 0.00045839510489510487,
+      "loss": 3.3649,
+      "step": 40600
+    },
+    {
+      "epoch": 11.837556060341313,
+      "grad_norm": 0.35426944494247437,
+      "learning_rate": 0.00045822027972027967,
+      "loss": 3.3644,
+      "step": 40650
+    },
+    {
+      "epoch": 11.852117187955034,
+      "grad_norm": 0.3524153232574463,
+      "learning_rate": 0.0004580454545454545,
+      "loss": 3.3766,
+      "step": 40700
+    },
+    {
+      "epoch": 11.866678315568757,
+      "grad_norm": 0.34207257628440857,
+      "learning_rate": 0.0004578706293706293,
+      "loss": 3.356,
+      "step": 40750
+    },
+    {
+      "epoch": 11.88123944318248,
+      "grad_norm": 0.33475518226623535,
+      "learning_rate": 0.0004576958041958042,
+      "loss": 3.3619,
+      "step": 40800
+    },
+    {
+      "epoch": 11.895800570796203,
+      "grad_norm": 0.3188439607620239,
+      "learning_rate": 0.000457520979020979,
+      "loss": 3.3655,
+      "step": 40850
+    },
+    {
+      "epoch": 11.910361698409925,
+      "grad_norm": 0.3339869976043701,
+      "learning_rate": 0.00045734615384615383,
+      "loss": 3.3671,
+      "step": 40900
+    },
+    {
+      "epoch": 11.924922826023646,
+      "grad_norm": 0.3358663022518158,
+      "learning_rate": 0.0004571713286713287,
+      "loss": 3.3729,
+      "step": 40950
+    },
+    {
+      "epoch": 11.93948395363737,
+      "grad_norm": 0.3430688679218292,
+      "learning_rate": 0.00045699650349650343,
+      "loss": 3.355,
+      "step": 41000
+    },
+    {
+      "epoch": 11.93948395363737,
+      "eval_accuracy": 0.37045693999110757,
+      "eval_loss": 3.551067590713501,
+      "eval_runtime": 180.9029,
+      "eval_samples_per_second": 92.016,
+      "eval_steps_per_second": 5.754,
+      "step": 41000
+    },
+    {
+      "epoch": 11.954045081251092,
+      "grad_norm": 0.3405440151691437,
+      "learning_rate": 0.0004568216783216783,
+      "loss": 3.3736,
+      "step": 41050
+    },
+    {
+      "epoch": 11.968606208864815,
+      "grad_norm": 0.3274223208427429,
+      "learning_rate": 0.0004566468531468531,
+      "loss": 3.3775,
+      "step": 41100
+    },
+    {
+      "epoch": 11.983167336478537,
+      "grad_norm": 0.34064793586730957,
+      "learning_rate": 0.00045647202797202794,
+      "loss": 3.3726,
+      "step": 41150
+    },
+    {
+      "epoch": 11.99772846409226,
+      "grad_norm": 0.4026555120944977,
+      "learning_rate": 0.00045629720279720274,
+      "loss": 3.3781,
+      "step": 41200
+    },
+    {
+      "epoch": 12.012231347195527,
+      "grad_norm": 0.341229110956192,
+      "learning_rate": 0.0004561223776223776,
+      "loss": 3.2739,
+      "step": 41250
+    },
+    {
+      "epoch": 12.02679247480925,
+      "grad_norm": 0.3594837486743927,
+      "learning_rate": 0.0004559475524475524,
+      "loss": 3.262,
+      "step": 41300
+    },
+    {
+      "epoch": 12.041353602422971,
+      "grad_norm": 0.3388396203517914,
+      "learning_rate": 0.00045577272727272724,
+      "loss": 3.257,
+      "step": 41350
+    },
+    {
+      "epoch": 12.055914730036694,
+      "grad_norm": 0.36203816533088684,
+      "learning_rate": 0.00045559790209790204,
+      "loss": 3.2833,
+      "step": 41400
+    },
+    {
+      "epoch": 12.070475857650417,
+      "grad_norm": 0.37606170773506165,
+      "learning_rate": 0.0004554230769230769,
+      "loss": 3.2727,
+      "step": 41450
+    },
+    {
+      "epoch": 12.08503698526414,
+      "grad_norm": 0.33072522282600403,
+      "learning_rate": 0.0004552482517482517,
+      "loss": 3.2844,
+      "step": 41500
+    },
+    {
+      "epoch": 12.099598112877862,
+      "grad_norm": 0.3422619104385376,
+      "learning_rate": 0.00045507342657342655,
+      "loss": 3.2729,
+      "step": 41550
+    },
+    {
+      "epoch": 12.114159240491583,
+      "grad_norm": 0.33112087845802307,
+      "learning_rate": 0.00045489860139860135,
+      "loss": 3.2855,
+      "step": 41600
+    },
+    {
+      "epoch": 12.128720368105306,
+      "grad_norm": 0.3561849296092987,
+      "learning_rate": 0.0004547237762237762,
+      "loss": 3.2809,
+      "step": 41650
+    },
+    {
+      "epoch": 12.143281495719028,
+      "grad_norm": 0.356389582157135,
+      "learning_rate": 0.00045454895104895106,
+      "loss": 3.2899,
+      "step": 41700
+    },
+    {
+      "epoch": 12.157842623332751,
+      "grad_norm": 0.37144315242767334,
+      "learning_rate": 0.0004543741258741258,
+      "loss": 3.2904,
+      "step": 41750
+    },
+    {
+      "epoch": 12.172403750946474,
+      "grad_norm": 0.3500952124595642,
+      "learning_rate": 0.00045419930069930066,
+      "loss": 3.2895,
+      "step": 41800
+    },
+    {
+      "epoch": 12.186964878560195,
+      "grad_norm": 0.3198910355567932,
+      "learning_rate": 0.00045402447552447546,
+      "loss": 3.2839,
+      "step": 41850
+    },
+    {
+      "epoch": 12.201526006173918,
+      "grad_norm": 0.37378618121147156,
+      "learning_rate": 0.0004538496503496503,
+      "loss": 3.3,
+      "step": 41900
+    },
+    {
+      "epoch": 12.21608713378764,
+      "grad_norm": 0.36039999127388,
+      "learning_rate": 0.0004536748251748251,
+      "loss": 3.2992,
+      "step": 41950
+    },
+    {
+      "epoch": 12.230648261401363,
+      "grad_norm": 0.3335278630256653,
+      "learning_rate": 0.00045349999999999996,
+      "loss": 3.3012,
+      "step": 42000
+    },
+    {
+      "epoch": 12.230648261401363,
+      "eval_accuracy": 0.3701739667951469,
+      "eval_loss": 3.5618510246276855,
+      "eval_runtime": 180.658,
+      "eval_samples_per_second": 92.141,
+      "eval_steps_per_second": 5.762,
+      "step": 42000
+    },
+    {
+      "epoch": 12.245209389015086,
+      "grad_norm": 0.38318052887916565,
+      "learning_rate": 0.00045332517482517476,
+      "loss": 3.3091,
+      "step": 42050
+    },
+    {
+      "epoch": 12.259770516628807,
+      "grad_norm": 0.3642372786998749,
+      "learning_rate": 0.0004531503496503496,
+      "loss": 3.3112,
+      "step": 42100
+    },
+    {
+      "epoch": 12.27433164424253,
+      "grad_norm": 0.3355906307697296,
+      "learning_rate": 0.0004529755244755244,
+      "loss": 3.3095,
+      "step": 42150
+    },
+    {
+      "epoch": 12.288892771856252,
+      "grad_norm": 0.3374500274658203,
+      "learning_rate": 0.00045280069930069927,
+      "loss": 3.3135,
+      "step": 42200
+    },
+    {
+      "epoch": 12.303453899469975,
+      "grad_norm": 0.3533301055431366,
+      "learning_rate": 0.00045262587412587407,
+      "loss": 3.3096,
+      "step": 42250
+    },
+    {
+      "epoch": 12.318015027083698,
+      "grad_norm": 0.34568679332733154,
+      "learning_rate": 0.0004524510489510489,
+      "loss": 3.3137,
+      "step": 42300
+    },
+    {
+      "epoch": 12.33257615469742,
+      "grad_norm": 0.3552064597606659,
+      "learning_rate": 0.0004522762237762238,
+      "loss": 3.3188,
+      "step": 42350
+    },
+    {
+      "epoch": 12.347137282311142,
+      "grad_norm": 0.365732878446579,
+      "learning_rate": 0.0004521013986013986,
+      "loss": 3.3069,
+      "step": 42400
+    },
+    {
+      "epoch": 12.361698409924864,
+      "grad_norm": 0.32813093066215515,
+      "learning_rate": 0.00045192657342657343,
+      "loss": 3.3176,
+      "step": 42450
+    },
+    {
+      "epoch": 12.376259537538587,
+      "grad_norm": 0.35801804065704346,
+      "learning_rate": 0.0004517517482517482,
+      "loss": 3.3266,
+      "step": 42500
+    },
+    {
+      "epoch": 12.39082066515231,
+      "grad_norm": 0.3690122067928314,
+      "learning_rate": 0.00045157692307692303,
+      "loss": 3.3077,
+      "step": 42550
+    },
+    {
+      "epoch": 12.405381792766033,
+      "grad_norm": 0.3640967905521393,
+      "learning_rate": 0.00045140209790209783,
+      "loss": 3.3327,
+      "step": 42600
+    },
+    {
+      "epoch": 12.419942920379754,
+      "grad_norm": 0.3548601567745209,
+      "learning_rate": 0.0004512272727272727,
+      "loss": 3.3232,
+      "step": 42650
+    },
+    {
+      "epoch": 12.434504047993476,
+      "grad_norm": 0.34999150037765503,
+      "learning_rate": 0.0004510524475524475,
+      "loss": 3.3247,
+      "step": 42700
+    },
+    {
+      "epoch": 12.449065175607199,
+      "grad_norm": 0.36435720324516296,
+      "learning_rate": 0.00045087762237762234,
+      "loss": 3.3183,
+      "step": 42750
+    },
+    {
+      "epoch": 12.463626303220922,
+      "grad_norm": 0.36453813314437866,
+      "learning_rate": 0.00045070279720279714,
+      "loss": 3.3269,
+      "step": 42800
+    },
+    {
+      "epoch": 12.478187430834645,
+      "grad_norm": 0.35878297686576843,
+      "learning_rate": 0.000450527972027972,
+      "loss": 3.3212,
+      "step": 42850
+    },
+    {
+      "epoch": 12.492748558448366,
+      "grad_norm": 0.36037617921829224,
+      "learning_rate": 0.0004503531468531468,
+      "loss": 3.3282,
+      "step": 42900
+    },
+    {
+      "epoch": 12.507309686062088,
+      "grad_norm": 0.35079270601272583,
+      "learning_rate": 0.00045017832167832165,
+      "loss": 3.3298,
+      "step": 42950
+    },
+    {
+      "epoch": 12.521870813675811,
+      "grad_norm": 0.3421091139316559,
+      "learning_rate": 0.0004500034965034965,
+      "loss": 3.3375,
+      "step": 43000
+    },
+    {
+      "epoch": 12.521870813675811,
+      "eval_accuracy": 0.370077447894489,
+      "eval_loss": 3.5572597980499268,
+      "eval_runtime": 180.7328,
+      "eval_samples_per_second": 92.103,
+      "eval_steps_per_second": 5.76,
+      "step": 43000
+    },
+    {
+      "epoch": 12.536431941289534,
+      "grad_norm": 0.33686643838882446,
+      "learning_rate": 0.0004498286713286713,
+      "loss": 3.3295,
+      "step": 43050
+    },
+    {
+      "epoch": 12.550993068903256,
+      "grad_norm": 0.3493344783782959,
+      "learning_rate": 0.00044965384615384615,
+      "loss": 3.3336,
+      "step": 43100
+    },
+    {
+      "epoch": 12.565554196516977,
+      "grad_norm": 0.3651737868785858,
+      "learning_rate": 0.00044947902097902095,
+      "loss": 3.3356,
+      "step": 43150
+    },
+    {
+      "epoch": 12.5801153241307,
+      "grad_norm": 0.35403475165367126,
+      "learning_rate": 0.0004493041958041958,
+      "loss": 3.3392,
+      "step": 43200
+    },
+    {
+      "epoch": 12.594676451744423,
+      "grad_norm": 0.35346367955207825,
+      "learning_rate": 0.00044912937062937055,
+      "loss": 3.3331,
+      "step": 43250
+    },
+    {
+      "epoch": 12.609237579358146,
+      "grad_norm": 0.3847574293613434,
+      "learning_rate": 0.0004489545454545454,
+      "loss": 3.3227,
+      "step": 43300
+    },
+    {
+      "epoch": 12.623798706971868,
+      "grad_norm": 0.32844939827919006,
+      "learning_rate": 0.0004487797202797202,
+      "loss": 3.3305,
+      "step": 43350
+    },
+    {
+      "epoch": 12.63835983458559,
+      "grad_norm": 0.3514168858528137,
+      "learning_rate": 0.00044860489510489506,
+      "loss": 3.3346,
+      "step": 43400
+    },
+    {
+      "epoch": 12.652920962199312,
+      "grad_norm": 0.3668827712535858,
+      "learning_rate": 0.00044843006993006986,
+      "loss": 3.3453,
+      "step": 43450
+    },
+    {
+      "epoch": 12.667482089813035,
+      "grad_norm": 0.35664016008377075,
+      "learning_rate": 0.0004482552447552447,
+      "loss": 3.3523,
+      "step": 43500
+    },
+    {
+      "epoch": 12.682043217426758,
+      "grad_norm": 0.3246038854122162,
+      "learning_rate": 0.0004480804195804195,
+      "loss": 3.3459,
+      "step": 43550
+    },
+    {
+      "epoch": 12.69660434504048,
+      "grad_norm": 0.39765822887420654,
+      "learning_rate": 0.00044790559440559437,
+      "loss": 3.3396,
+      "step": 43600
+    },
+    {
+      "epoch": 12.711165472654203,
+      "grad_norm": 0.404825896024704,
+      "learning_rate": 0.00044773076923076917,
+      "loss": 3.3458,
+      "step": 43650
+    },
+    {
+      "epoch": 12.725726600267924,
+      "grad_norm": 0.3610759675502777,
+      "learning_rate": 0.000447555944055944,
+      "loss": 3.344,
+      "step": 43700
+    },
+    {
+      "epoch": 12.740287727881647,
+      "grad_norm": 0.3398517668247223,
+      "learning_rate": 0.0004473811188811189,
+      "loss": 3.3499,
+      "step": 43750
+    },
+    {
+      "epoch": 12.75484885549537,
+      "grad_norm": 0.35025089979171753,
+      "learning_rate": 0.0004472062937062937,
+      "loss": 3.3479,
+      "step": 43800
+    },
+    {
+      "epoch": 12.769409983109092,
+      "grad_norm": 0.35315173864364624,
+      "learning_rate": 0.00044703146853146853,
+      "loss": 3.3462,
+      "step": 43850
+    },
+    {
+      "epoch": 12.783971110722815,
+      "grad_norm": 0.35999706387519836,
+      "learning_rate": 0.00044685664335664333,
+      "loss": 3.348,
+      "step": 43900
+    },
+    {
+      "epoch": 12.798532238336536,
+      "grad_norm": 0.33961042761802673,
+      "learning_rate": 0.0004466818181818182,
+      "loss": 3.3528,
+      "step": 43950
+    },
+    {
+      "epoch": 12.813093365950259,
+      "grad_norm": 0.34737616777420044,
+      "learning_rate": 0.00044650699300699293,
+      "loss": 3.3423,
+      "step": 44000
+    },
+    {
+      "epoch": 12.813093365950259,
+      "eval_accuracy": 0.3708158586314349,
+      "eval_loss": 3.547250986099243,
+      "eval_runtime": 180.8057,
+      "eval_samples_per_second": 92.066,
+      "eval_steps_per_second": 5.758,
+      "step": 44000
+    },
+    {
+      "epoch": 12.827654493563982,
+      "grad_norm": 0.372464656829834,
+      "learning_rate": 0.0004463321678321678,
+      "loss": 3.3532,
+      "step": 44050
+    },
+    {
+      "epoch": 12.842215621177704,
+      "grad_norm": 0.346667617559433,
+      "learning_rate": 0.0004461573426573426,
+      "loss": 3.3469,
+      "step": 44100
+    },
+    {
+      "epoch": 12.856776748791427,
+      "grad_norm": 0.3418535888195038,
+      "learning_rate": 0.00044598251748251744,
+      "loss": 3.3397,
+      "step": 44150
+    },
+    {
+      "epoch": 12.871337876405148,
+      "grad_norm": 0.3593897223472595,
+      "learning_rate": 0.00044580769230769224,
+      "loss": 3.3517,
+      "step": 44200
+    },
+    {
+      "epoch": 12.88589900401887,
+      "grad_norm": 0.3259468972682953,
+      "learning_rate": 0.0004456328671328671,
+      "loss": 3.3539,
+      "step": 44250
+    },
+    {
+      "epoch": 12.900460131632594,
+      "grad_norm": 0.3682238459587097,
+      "learning_rate": 0.0004454580419580419,
+      "loss": 3.3607,
+      "step": 44300
+    },
+    {
+      "epoch": 12.915021259246316,
+      "grad_norm": 0.33671849966049194,
+      "learning_rate": 0.00044528321678321674,
+      "loss": 3.3507,
+      "step": 44350
+    },
+    {
+      "epoch": 12.929582386860039,
+      "grad_norm": 0.3739355504512787,
+      "learning_rate": 0.0004451083916083916,
+      "loss": 3.3463,
+      "step": 44400
+    },
+    {
+      "epoch": 12.944143514473762,
+      "grad_norm": 0.3360922336578369,
+      "learning_rate": 0.0004449335664335664,
+      "loss": 3.3563,
+      "step": 44450
+    },
+    {
+      "epoch": 12.958704642087483,
+      "grad_norm": 0.38051101565361023,
+      "learning_rate": 0.00044475874125874125,
+      "loss": 3.3509,
+      "step": 44500
+    },
+    {
+      "epoch": 12.973265769701205,
+      "grad_norm": 0.35468021035194397,
+      "learning_rate": 0.00044458391608391605,
+      "loss": 3.3574,
+      "step": 44550
+    },
+    {
+      "epoch": 12.987826897314928,
+      "grad_norm": 0.33574724197387695,
+      "learning_rate": 0.0004444090909090909,
+      "loss": 3.3432,
+      "step": 44600
+    },
+    {
+      "epoch": 13.002329780418195,
+      "grad_norm": 0.3547254502773285,
+      "learning_rate": 0.0004442342657342657,
+      "loss": 3.3365,
+      "step": 44650
+    },
+    {
+      "epoch": 13.016890908031918,
+      "grad_norm": 0.3475710451602936,
+      "learning_rate": 0.00044405944055944056,
+      "loss": 3.232,
+      "step": 44700
+    },
+    {
+      "epoch": 13.031452035645641,
+      "grad_norm": 0.3456555902957916,
+      "learning_rate": 0.0004438846153846153,
+      "loss": 3.2463,
+      "step": 44750
+    },
+    {
+      "epoch": 13.046013163259364,
+      "grad_norm": 0.36879873275756836,
+      "learning_rate": 0.00044370979020979016,
+      "loss": 3.2534,
+      "step": 44800
+    },
+    {
+      "epoch": 13.060574290873085,
+      "grad_norm": 0.3598310649394989,
+      "learning_rate": 0.00044353496503496496,
+      "loss": 3.2438,
+      "step": 44850
+    },
+    {
+      "epoch": 13.075135418486807,
+      "grad_norm": 0.35868072509765625,
+      "learning_rate": 0.0004433601398601398,
+      "loss": 3.2632,
+      "step": 44900
+    },
+    {
+      "epoch": 13.08969654610053,
+      "grad_norm": 0.33635494112968445,
+      "learning_rate": 0.0004431853146853146,
+      "loss": 3.2603,
+      "step": 44950
+    },
+    {
+      "epoch": 13.104257673714253,
+      "grad_norm": 0.36905378103256226,
+      "learning_rate": 0.00044301048951048946,
+      "loss": 3.258,
+      "step": 45000
+    },
+    {
+      "epoch": 13.104257673714253,
+      "eval_accuracy": 0.37040074506478055,
+      "eval_loss": 3.5601999759674072,
+      "eval_runtime": 180.8794,
+      "eval_samples_per_second": 92.028,
+      "eval_steps_per_second": 5.755,
+      "step": 45000
+    },
+    {
+      "epoch": 13.118818801327976,
+      "grad_norm": 0.3864152729511261,
+      "learning_rate": 0.00044283566433566426,
+      "loss": 3.2666,
+      "step": 45050
+    },
+    {
+      "epoch": 13.133379928941697,
+      "grad_norm": 0.3702114522457123,
+      "learning_rate": 0.0004426608391608391,
+      "loss": 3.2676,
+      "step": 45100
+    },
+    {
+      "epoch": 13.14794105655542,
+      "grad_norm": 0.35562071204185486,
+      "learning_rate": 0.00044248601398601397,
+      "loss": 3.2743,
+      "step": 45150
+    },
+    {
+      "epoch": 13.162502184169142,
+      "grad_norm": 0.3449855148792267,
+      "learning_rate": 0.00044231118881118877,
+      "loss": 3.2757,
+      "step": 45200
+    },
+    {
+      "epoch": 13.177063311782865,
+      "grad_norm": 0.34758079051971436,
+      "learning_rate": 0.0004421363636363636,
+      "loss": 3.2819,
+      "step": 45250
+    },
+    {
+      "epoch": 13.191624439396588,
+      "grad_norm": 0.3789484202861786,
+      "learning_rate": 0.0004419615384615384,
+      "loss": 3.2785,
+      "step": 45300
+    },
+    {
+      "epoch": 13.206185567010309,
+      "grad_norm": 0.3660043478012085,
+      "learning_rate": 0.0004417867132867133,
+      "loss": 3.2718,
+      "step": 45350
+    },
+    {
+      "epoch": 13.220746694624031,
+      "grad_norm": 0.392027348279953,
+      "learning_rate": 0.0004416118881118881,
+      "loss": 3.273,
+      "step": 45400
+    },
+    {
+      "epoch": 13.235307822237754,
+      "grad_norm": 0.4098648130893707,
+      "learning_rate": 0.00044143706293706293,
+      "loss": 3.2767,
+      "step": 45450
+    },
+    {
+      "epoch": 13.249868949851477,
+      "grad_norm": 0.3648849129676819,
+      "learning_rate": 0.0004412622377622377,
+      "loss": 3.2936,
+      "step": 45500
+    },
+    {
+      "epoch": 13.2644300774652,
+      "grad_norm": 0.381013423204422,
+      "learning_rate": 0.00044108741258741253,
+      "loss": 3.2894,
+      "step": 45550
+    },
+    {
+      "epoch": 13.27899120507892,
+      "grad_norm": 0.344605416059494,
+      "learning_rate": 0.00044091258741258733,
+      "loss": 3.2903,
+      "step": 45600
+    },
+    {
+      "epoch": 13.293552332692643,
+      "grad_norm": 0.3393156826496124,
+      "learning_rate": 0.0004407377622377622,
+      "loss": 3.2873,
+      "step": 45650
+    },
+    {
+      "epoch": 13.308113460306366,
+      "grad_norm": 0.34138327836990356,
+      "learning_rate": 0.000440562937062937,
+      "loss": 3.2941,
+      "step": 45700
+    },
+    {
+      "epoch": 13.322674587920089,
+      "grad_norm": 0.35385775566101074,
+      "learning_rate": 0.00044038811188811184,
+      "loss": 3.2914,
+      "step": 45750
+    },
+    {
+      "epoch": 13.337235715533811,
+      "grad_norm": 0.39340725541114807,
+      "learning_rate": 0.0004402132867132867,
+      "loss": 3.2961,
+      "step": 45800
+    },
+    {
+      "epoch": 13.351796843147532,
+      "grad_norm": 0.38336238265037537,
+      "learning_rate": 0.0004400384615384615,
+      "loss": 3.3032,
+      "step": 45850
+    },
+    {
+      "epoch": 13.366357970761255,
+      "grad_norm": 0.376488596200943,
+      "learning_rate": 0.00043986363636363635,
+      "loss": 3.302,
+      "step": 45900
+    },
+    {
+      "epoch": 13.380919098374978,
+      "grad_norm": 0.38429391384124756,
+      "learning_rate": 0.00043968881118881115,
+      "loss": 3.3086,
+      "step": 45950
+    },
+    {
+      "epoch": 13.3954802259887,
+      "grad_norm": 0.3611532151699066,
+      "learning_rate": 0.000439513986013986,
+      "loss": 3.2948,
+      "step": 46000
+    },
+    {
+      "epoch": 13.3954802259887,
+      "eval_accuracy": 0.37078893679434516,
+      "eval_loss": 3.556919813156128,
+      "eval_runtime": 181.2381,
+      "eval_samples_per_second": 91.846,
+      "eval_steps_per_second": 5.744,
+      "step": 46000
+    },
+    {
+      "epoch": 13.410041353602423,
+      "grad_norm": 0.35340753197669983,
+      "learning_rate": 0.0004393391608391608,
+      "loss": 3.3028,
+      "step": 46050
+    },
+    {
+      "epoch": 13.424602481216146,
+      "grad_norm": 0.33229947090148926,
+      "learning_rate": 0.00043916433566433565,
+      "loss": 3.3226,
+      "step": 46100
+    },
+    {
+      "epoch": 13.439163608829867,
+      "grad_norm": 0.3614226281642914,
+      "learning_rate": 0.00043898951048951045,
+      "loss": 3.3084,
+      "step": 46150
+    },
+    {
+      "epoch": 13.45372473644359,
+      "grad_norm": 0.35992908477783203,
+      "learning_rate": 0.0004388146853146853,
+      "loss": 3.3103,
+      "step": 46200
+    },
+    {
+      "epoch": 13.468285864057313,
+      "grad_norm": 0.37572646141052246,
+      "learning_rate": 0.00043863986013986005,
+      "loss": 3.3319,
+      "step": 46250
+    },
+    {
+      "epoch": 13.482846991671035,
+      "grad_norm": 0.35673362016677856,
+      "learning_rate": 0.0004384650349650349,
+      "loss": 3.3052,
+      "step": 46300
+    },
+    {
+      "epoch": 13.497408119284758,
+      "grad_norm": 0.3684885501861572,
+      "learning_rate": 0.0004382902097902097,
+      "loss": 3.3061,
+      "step": 46350
+    },
+    {
+      "epoch": 13.51196924689848,
+      "grad_norm": 0.35558366775512695,
+      "learning_rate": 0.00043811538461538456,
+      "loss": 3.3148,
+      "step": 46400
+    },
+    {
+      "epoch": 13.526530374512202,
+      "grad_norm": 0.3368312120437622,
+      "learning_rate": 0.0004379405594405594,
+      "loss": 3.311,
+      "step": 46450
+    },
+    {
+      "epoch": 13.541091502125925,
+      "grad_norm": 0.3515823483467102,
+      "learning_rate": 0.0004377657342657342,
+      "loss": 3.3151,
+      "step": 46500
+    },
+    {
+      "epoch": 13.555652629739647,
+      "grad_norm": 0.3697565495967865,
+      "learning_rate": 0.00043759090909090907,
+      "loss": 3.3236,
+      "step": 46550
+    },
+    {
+      "epoch": 13.57021375735337,
+      "grad_norm": 0.371623694896698,
+      "learning_rate": 0.00043741608391608387,
+      "loss": 3.3189,
+      "step": 46600
+    },
+    {
+      "epoch": 13.584774884967091,
+      "grad_norm": 0.36001795530319214,
+      "learning_rate": 0.0004372412587412587,
+      "loss": 3.3307,
+      "step": 46650
+    },
+    {
+      "epoch": 13.599336012580814,
+      "grad_norm": 0.3571346700191498,
+      "learning_rate": 0.0004370664335664335,
+      "loss": 3.3313,
+      "step": 46700
+    },
+    {
+      "epoch": 13.613897140194537,
+      "grad_norm": 0.34875962138175964,
+      "learning_rate": 0.0004368916083916084,
+      "loss": 3.328,
+      "step": 46750
+    },
+    {
+      "epoch": 13.62845826780826,
+      "grad_norm": 0.3460927903652191,
+      "learning_rate": 0.0004367167832167832,
+      "loss": 3.3221,
+      "step": 46800
+    },
+    {
+      "epoch": 13.643019395421982,
+      "grad_norm": 0.3462391495704651,
+      "learning_rate": 0.00043654195804195803,
+      "loss": 3.315,
+      "step": 46850
+    },
+    {
+      "epoch": 13.657580523035705,
+      "grad_norm": 0.33300238847732544,
+      "learning_rate": 0.00043636713286713283,
+      "loss": 3.3241,
+      "step": 46900
+    },
+    {
+      "epoch": 13.672141650649426,
+      "grad_norm": 0.35763484239578247,
+      "learning_rate": 0.0004361923076923077,
+      "loss": 3.3222,
+      "step": 46950
+    },
+    {
+      "epoch": 13.686702778263149,
+      "grad_norm": 0.3411622941493988,
+      "learning_rate": 0.00043601748251748243,
+      "loss": 3.3168,
+      "step": 47000
+    },
+    {
+      "epoch": 13.686702778263149,
+      "eval_accuracy": 0.3710402856489209,
+      "eval_loss": 3.5501530170440674,
+      "eval_runtime": 181.1346,
+      "eval_samples_per_second": 91.899,
+      "eval_steps_per_second": 5.747,
+      "step": 47000
+    },
+    {
+      "epoch": 13.701263905876871,
+      "grad_norm": 0.36353328824043274,
+      "learning_rate": 0.00043584265734265734,
+      "loss": 3.3267,
+      "step": 47050
+    },
+    {
+      "epoch": 13.715825033490594,
+      "grad_norm": 0.3539396822452545,
+      "learning_rate": 0.0004356678321678321,
+      "loss": 3.3319,
+      "step": 47100
+    },
+    {
+      "epoch": 13.730386161104317,
+      "grad_norm": 0.3500533103942871,
+      "learning_rate": 0.00043549300699300694,
+      "loss": 3.3324,
+      "step": 47150
+    },
+    {
+      "epoch": 13.744947288718038,
+      "grad_norm": 0.360314279794693,
+      "learning_rate": 0.0004353181818181818,
+      "loss": 3.3379,
+      "step": 47200
+    },
+    {
+      "epoch": 13.75950841633176,
+      "grad_norm": 0.365933358669281,
+      "learning_rate": 0.0004351433566433566,
+      "loss": 3.329,
+      "step": 47250
+    },
+    {
+      "epoch": 13.774069543945483,
+      "grad_norm": 0.3569656014442444,
+      "learning_rate": 0.00043496853146853144,
+      "loss": 3.3337,
+      "step": 47300
+    },
+    {
+      "epoch": 13.788630671559206,
+      "grad_norm": 0.35865193605422974,
+      "learning_rate": 0.00043479370629370624,
+      "loss": 3.3275,
+      "step": 47350
+    },
+    {
+      "epoch": 13.803191799172929,
+      "grad_norm": 0.342277467250824,
+      "learning_rate": 0.0004346188811188811,
+      "loss": 3.3288,
+      "step": 47400
+    },
+    {
+      "epoch": 13.81775292678665,
+      "grad_norm": 0.364913672208786,
+      "learning_rate": 0.0004344440559440559,
+      "loss": 3.3432,
+      "step": 47450
+    },
+    {
+      "epoch": 13.832314054400372,
+      "grad_norm": 0.361052006483078,
+      "learning_rate": 0.00043426923076923075,
+      "loss": 3.332,
+      "step": 47500
+    },
+    {
+      "epoch": 13.846875182014095,
+      "grad_norm": 0.3582068681716919,
+      "learning_rate": 0.00043409440559440555,
+      "loss": 3.3425,
+      "step": 47550
+    },
+    {
+      "epoch": 13.861436309627818,
+      "grad_norm": 0.34829556941986084,
+      "learning_rate": 0.0004339195804195804,
+      "loss": 3.3265,
+      "step": 47600
+    },
+    {
+      "epoch": 13.87599743724154,
+      "grad_norm": 0.35710349678993225,
+      "learning_rate": 0.0004337447552447552,
+      "loss": 3.3391,
+      "step": 47650
+    },
+    {
+      "epoch": 13.890558564855262,
+      "grad_norm": 0.3624763786792755,
+      "learning_rate": 0.00043356993006993006,
+      "loss": 3.3336,
+      "step": 47700
+    },
+    {
+      "epoch": 13.905119692468984,
+      "grad_norm": 0.338870644569397,
+      "learning_rate": 0.0004333951048951048,
+      "loss": 3.3439,
+      "step": 47750
+    },
+    {
+      "epoch": 13.919680820082707,
+      "grad_norm": 0.32162490487098694,
+      "learning_rate": 0.0004332202797202797,
+      "loss": 3.3331,
+      "step": 47800
+    },
+    {
+      "epoch": 13.93424194769643,
+      "grad_norm": 0.3730635344982147,
+      "learning_rate": 0.00043304545454545456,
+      "loss": 3.3267,
+      "step": 47850
+    },
+    {
+      "epoch": 13.948803075310153,
+      "grad_norm": 0.36320826411247253,
+      "learning_rate": 0.0004328706293706293,
+      "loss": 3.3336,
+      "step": 47900
+    },
+    {
+      "epoch": 13.963364202923874,
+      "grad_norm": 0.3524576723575592,
+      "learning_rate": 0.00043269580419580416,
+      "loss": 3.3353,
+      "step": 47950
+    },
+    {
+      "epoch": 13.977925330537596,
+      "grad_norm": 0.3460340201854706,
+      "learning_rate": 0.00043252097902097896,
+      "loss": 3.3278,
+      "step": 48000
+    },
+    {
+      "epoch": 13.977925330537596,
+      "eval_accuracy": 0.37145504652775313,
+      "eval_loss": 3.540445566177368,
+      "eval_runtime": 181.2081,
+      "eval_samples_per_second": 91.861,
+      "eval_steps_per_second": 5.745,
+      "step": 48000
+    },
+    {
+      "epoch": 13.992486458151319,
+      "grad_norm": 0.35427072644233704,
+      "learning_rate": 0.0004323461538461538,
+      "loss": 3.3319,
+      "step": 48050
+    },
+    {
+      "epoch": 14.006989341254586,
+      "grad_norm": 0.3449511229991913,
+      "learning_rate": 0.0004321713286713286,
+      "loss": 3.2881,
+      "step": 48100
+    },
+    {
+      "epoch": 14.021550468868309,
+      "grad_norm": 0.39801499247550964,
+      "learning_rate": 0.00043199650349650347,
+      "loss": 3.2324,
+      "step": 48150
+    },
+    {
+      "epoch": 14.036111596482032,
+      "grad_norm": 0.34631258249282837,
+      "learning_rate": 0.00043182167832167827,
+      "loss": 3.2373,
+      "step": 48200
+    },
+    {
+      "epoch": 14.050672724095755,
+      "grad_norm": 0.34527838230133057,
+      "learning_rate": 0.0004316468531468531,
+      "loss": 3.2396,
+      "step": 48250
+    },
+    {
+      "epoch": 14.065233851709475,
+      "grad_norm": 0.3679901659488678,
+      "learning_rate": 0.0004314720279720279,
+      "loss": 3.2423,
+      "step": 48300
+    },
+    {
+      "epoch": 14.079794979323198,
+      "grad_norm": 0.36143872141838074,
+      "learning_rate": 0.0004312972027972028,
+      "loss": 3.2434,
+      "step": 48350
+    },
+    {
+      "epoch": 14.094356106936921,
+      "grad_norm": 0.36590418219566345,
+      "learning_rate": 0.0004311223776223776,
+      "loss": 3.2398,
+      "step": 48400
+    },
+    {
+      "epoch": 14.108917234550644,
+      "grad_norm": 0.35469797253608704,
+      "learning_rate": 0.00043094755244755243,
+      "loss": 3.2428,
+      "step": 48450
+    },
+    {
+      "epoch": 14.123478362164366,
+      "grad_norm": 0.34602782130241394,
+      "learning_rate": 0.0004307727272727272,
+      "loss": 3.2579,
+      "step": 48500
+    },
+    {
+      "epoch": 14.13803948977809,
+      "grad_norm": 0.32153913378715515,
+      "learning_rate": 0.0004305979020979021,
+      "loss": 3.2568,
+      "step": 48550
+    },
+    {
+      "epoch": 14.15260061739181,
+      "grad_norm": 0.34756767749786377,
+      "learning_rate": 0.00043042307692307694,
+      "loss": 3.2628,
+      "step": 48600
+    },
+    {
+      "epoch": 14.167161745005533,
+      "grad_norm": 0.3779682517051697,
+      "learning_rate": 0.0004302482517482517,
+      "loss": 3.2529,
+      "step": 48650
+    },
+    {
+      "epoch": 14.181722872619256,
+      "grad_norm": 0.3358452320098877,
+      "learning_rate": 0.00043007342657342654,
+      "loss": 3.2635,
+      "step": 48700
+    },
+    {
+      "epoch": 14.196284000232978,
+      "grad_norm": 0.3383064568042755,
+      "learning_rate": 0.00042989860139860134,
+      "loss": 3.2613,
+      "step": 48750
+    },
+    {
+      "epoch": 14.210845127846701,
+      "grad_norm": 0.33334216475486755,
+      "learning_rate": 0.0004297237762237762,
+      "loss": 3.2725,
+      "step": 48800
+    },
+    {
+      "epoch": 14.225406255460422,
+      "grad_norm": 0.4189155697822571,
+      "learning_rate": 0.000429548951048951,
+      "loss": 3.271,
+      "step": 48850
+    },
+    {
+      "epoch": 14.239967383074145,
+      "grad_norm": 0.3933100402355194,
+      "learning_rate": 0.00042937412587412585,
+      "loss": 3.2692,
+      "step": 48900
+    },
+    {
+      "epoch": 14.254528510687868,
+      "grad_norm": 0.3877250850200653,
+      "learning_rate": 0.00042919930069930065,
+      "loss": 3.2711,
+      "step": 48950
+    },
+    {
+      "epoch": 14.26908963830159,
+      "grad_norm": 0.37484902143478394,
+      "learning_rate": 0.0004290244755244755,
+      "loss": 3.2799,
+      "step": 49000
+    },
+    {
+      "epoch": 14.26908963830159,
+      "eval_accuracy": 0.3709477638769138,
+      "eval_loss": 3.558025360107422,
+      "eval_runtime": 180.9464,
+      "eval_samples_per_second": 91.994,
+      "eval_steps_per_second": 5.753,
+      "step": 49000
+    },
+    {
+      "epoch": 14.283650765915313,
+      "grad_norm": 0.3490859270095825,
+      "learning_rate": 0.0004288496503496503,
+      "loss": 3.2762,
+      "step": 49050
+    },
+    {
+      "epoch": 14.298211893529034,
+      "grad_norm": 0.38379967212677,
+      "learning_rate": 0.00042867482517482515,
+      "loss": 3.2816,
+      "step": 49100
+    },
+    {
+      "epoch": 14.312773021142757,
+      "grad_norm": 0.366131454706192,
+      "learning_rate": 0.00042849999999999995,
+      "loss": 3.2791,
+      "step": 49150
+    },
+    {
+      "epoch": 14.32733414875648,
+      "grad_norm": 0.3447644114494324,
+      "learning_rate": 0.0004283251748251748,
+      "loss": 3.27,
+      "step": 49200
+    },
+    {
+      "epoch": 14.341895276370202,
+      "grad_norm": 0.3631740212440491,
+      "learning_rate": 0.00042815034965034966,
+      "loss": 3.2812,
+      "step": 49250
+    },
+    {
+      "epoch": 14.356456403983925,
+      "grad_norm": 0.37680432200431824,
+      "learning_rate": 0.00042797552447552446,
+      "loss": 3.2918,
+      "step": 49300
+    },
+    {
+      "epoch": 14.371017531597648,
+      "grad_norm": 0.348541796207428,
+      "learning_rate": 0.0004278006993006993,
+      "loss": 3.2858,
+      "step": 49350
+    },
+    {
+      "epoch": 14.385578659211369,
+      "grad_norm": 0.40237003564834595,
+      "learning_rate": 0.00042762587412587406,
+      "loss": 3.2925,
+      "step": 49400
+    },
+    {
+      "epoch": 14.400139786825092,
+      "grad_norm": 0.3444540798664093,
+      "learning_rate": 0.0004274510489510489,
+      "loss": 3.2888,
+      "step": 49450
+    },
+    {
+      "epoch": 14.414700914438814,
+      "grad_norm": 0.37795794010162354,
+      "learning_rate": 0.0004272762237762237,
+      "loss": 3.3037,
+      "step": 49500
+    },
+    {
+      "epoch": 14.429262042052537,
+      "grad_norm": 0.3671080768108368,
+      "learning_rate": 0.00042710139860139857,
+      "loss": 3.2885,
+      "step": 49550
+    },
+    {
+      "epoch": 14.44382316966626,
+      "grad_norm": 0.3939375877380371,
+      "learning_rate": 0.00042692657342657337,
+      "loss": 3.2869,
+      "step": 49600
+    },
+    {
+      "epoch": 14.45838429727998,
+      "grad_norm": 0.3599258065223694,
+      "learning_rate": 0.0004267517482517482,
+      "loss": 3.2999,
+      "step": 49650
+    },
+    {
+      "epoch": 14.472945424893704,
+      "grad_norm": 0.39820772409439087,
+      "learning_rate": 0.000426576923076923,
+      "loss": 3.302,
+      "step": 49700
+    },
+    {
+      "epoch": 14.487506552507426,
+      "grad_norm": 0.33834782242774963,
+      "learning_rate": 0.0004264020979020979,
+      "loss": 3.2965,
+      "step": 49750
+    },
+    {
+      "epoch": 14.502067680121149,
+      "grad_norm": 0.34589359164237976,
+      "learning_rate": 0.0004262272727272727,
+      "loss": 3.2905,
+      "step": 49800
+    },
+    {
+      "epoch": 14.516628807734872,
+      "grad_norm": 0.36442044377326965,
+      "learning_rate": 0.00042605244755244753,
+      "loss": 3.3112,
+      "step": 49850
+    },
+    {
+      "epoch": 14.531189935348593,
+      "grad_norm": 0.3587281107902527,
+      "learning_rate": 0.00042587762237762233,
+      "loss": 3.2954,
+      "step": 49900
+    },
+    {
+      "epoch": 14.545751062962315,
+      "grad_norm": 0.40609636902809143,
+      "learning_rate": 0.0004257027972027972,
+      "loss": 3.3017,
+      "step": 49950
+    },
+    {
+      "epoch": 14.560312190576038,
+      "grad_norm": 0.36938217282295227,
+      "learning_rate": 0.00042552797202797204,
+      "loss": 3.305,
+      "step": 50000
+    },
+    {
+      "epoch": 14.560312190576038,
+      "eval_accuracy": 0.3712948086938959,
+      "eval_loss": 3.549705982208252,
+      "eval_runtime": 180.9754,
+      "eval_samples_per_second": 91.979,
+      "eval_steps_per_second": 5.752,
+      "step": 50000
+    },
+    {
+      "epoch": 14.574873318189761,
+      "grad_norm": 0.3450993299484253,
+      "learning_rate": 0.00042535314685314684,
+      "loss": 3.3049,
+      "step": 50050
+    },
+    {
+      "epoch": 14.589434445803484,
+      "grad_norm": 0.3647083342075348,
+      "learning_rate": 0.0004251783216783217,
+      "loss": 3.2905,
+      "step": 50100
+    },
+    {
+      "epoch": 14.603995573417205,
+      "grad_norm": 0.374889999628067,
+      "learning_rate": 0.00042500349650349643,
+      "loss": 3.2949,
+      "step": 50150
+    },
+    {
+      "epoch": 14.618556701030927,
+      "grad_norm": 0.35723409056663513,
+      "learning_rate": 0.0004248286713286713,
+      "loss": 3.3063,
+      "step": 50200
+    },
+    {
+      "epoch": 14.63311782864465,
+      "grad_norm": 0.3582786023616791,
+      "learning_rate": 0.0004246538461538461,
+      "loss": 3.3007,
+      "step": 50250
+    },
+    {
+      "epoch": 14.647678956258373,
+      "grad_norm": 0.3550582528114319,
+      "learning_rate": 0.00042447902097902094,
+      "loss": 3.3112,
+      "step": 50300
+    },
+    {
+      "epoch": 14.662240083872096,
+      "grad_norm": 0.38724565505981445,
+      "learning_rate": 0.00042430419580419574,
+      "loss": 3.3092,
+      "step": 50350
+    },
+    {
+      "epoch": 14.676801211485817,
+      "grad_norm": 0.3551936745643616,
+      "learning_rate": 0.0004241293706293706,
+      "loss": 3.3052,
+      "step": 50400
+    },
+    {
+      "epoch": 14.69136233909954,
+      "grad_norm": 0.37477564811706543,
+      "learning_rate": 0.0004239545454545454,
+      "loss": 3.3014,
+      "step": 50450
+    },
+    {
+      "epoch": 14.705923466713262,
+      "grad_norm": 0.3632015287876129,
+      "learning_rate": 0.00042377972027972025,
+      "loss": 3.3086,
+      "step": 50500
+    },
+    {
+      "epoch": 14.720484594326985,
+      "grad_norm": 0.3654315769672394,
+      "learning_rate": 0.00042360489510489505,
+      "loss": 3.3075,
+      "step": 50550
+    },
+    {
+      "epoch": 14.735045721940708,
+      "grad_norm": 0.3743443191051483,
+      "learning_rate": 0.0004234300699300699,
+      "loss": 3.3103,
+      "step": 50600
+    },
+    {
+      "epoch": 14.749606849554429,
+      "grad_norm": 0.3471044898033142,
+      "learning_rate": 0.00042325524475524476,
+      "loss": 3.3226,
+      "step": 50650
+    },
+    {
+      "epoch": 14.764167977168151,
+      "grad_norm": 0.35494619607925415,
+      "learning_rate": 0.00042308041958041956,
+      "loss": 3.3218,
+      "step": 50700
+    },
+    {
+      "epoch": 14.778729104781874,
+      "grad_norm": 0.36895397305488586,
+      "learning_rate": 0.0004229055944055944,
+      "loss": 3.3032,
+      "step": 50750
+    },
+    {
+      "epoch": 14.793290232395597,
+      "grad_norm": 0.3416297733783722,
+      "learning_rate": 0.0004227307692307692,
+      "loss": 3.3163,
+      "step": 50800
+    },
+    {
+      "epoch": 14.80785136000932,
+      "grad_norm": 0.3781328797340393,
+      "learning_rate": 0.00042255594405594406,
+      "loss": 3.3142,
+      "step": 50850
+    },
+    {
+      "epoch": 14.822412487623042,
+      "grad_norm": 0.3623887002468109,
+      "learning_rate": 0.0004223811188811188,
+      "loss": 3.3257,
+      "step": 50900
+    },
+    {
+      "epoch": 14.836973615236763,
+      "grad_norm": 0.360993355512619,
+      "learning_rate": 0.00042220629370629366,
+      "loss": 3.3151,
+      "step": 50950
+    },
+    {
+      "epoch": 14.851534742850486,
+      "grad_norm": 0.3625052571296692,
+      "learning_rate": 0.00042203146853146846,
+      "loss": 3.3274,
+      "step": 51000
+    },
+    {
+      "epoch": 14.851534742850486,
+      "eval_accuracy": 0.37191624463649997,
+      "eval_loss": 3.5400757789611816,
+      "eval_runtime": 180.9103,
+      "eval_samples_per_second": 92.012,
+      "eval_steps_per_second": 5.754,
+      "step": 51000
+    },
+    {
+      "epoch": 14.866095870464209,
+      "grad_norm": 0.3725496232509613,
+      "learning_rate": 0.0004218566433566433,
+      "loss": 3.328,
+      "step": 51050
+    },
+    {
+      "epoch": 14.880656998077932,
+      "grad_norm": 0.3665943741798401,
+      "learning_rate": 0.0004216818181818181,
+      "loss": 3.3191,
+      "step": 51100
+    },
+    {
+      "epoch": 14.895218125691654,
+      "grad_norm": 0.3478486239910126,
+      "learning_rate": 0.00042150699300699297,
+      "loss": 3.3334,
+      "step": 51150
+    },
+    {
+      "epoch": 14.909779253305375,
+      "grad_norm": 0.3583490252494812,
+      "learning_rate": 0.00042133216783216777,
+      "loss": 3.3253,
+      "step": 51200
+    },
+    {
+      "epoch": 14.924340380919098,
+      "grad_norm": 0.3656362295150757,
+      "learning_rate": 0.0004211573426573426,
+      "loss": 3.3083,
+      "step": 51250
+    },
+    {
+      "epoch": 14.93890150853282,
+      "grad_norm": 0.37020862102508545,
+      "learning_rate": 0.0004209825174825175,
+      "loss": 3.3181,
+      "step": 51300
+    },
+    {
+      "epoch": 14.953462636146543,
+      "grad_norm": 0.3434571325778961,
+      "learning_rate": 0.0004208076923076923,
+      "loss": 3.3307,
+      "step": 51350
+    },
+    {
+      "epoch": 14.968023763760266,
+      "grad_norm": 0.3477000594139099,
+      "learning_rate": 0.00042063286713286713,
+      "loss": 3.3369,
+      "step": 51400
+    },
+    {
+      "epoch": 14.982584891373987,
+      "grad_norm": 0.3382193148136139,
+      "learning_rate": 0.00042045804195804193,
+      "loss": 3.3173,
+      "step": 51450
+    },
+    {
+      "epoch": 14.99714601898771,
+      "grad_norm": 0.3472067713737488,
+      "learning_rate": 0.0004202832167832168,
+      "loss": 3.3205,
+      "step": 51500
+    },
+    {
+      "epoch": 15.011648902090977,
+      "grad_norm": 0.3728955090045929,
+      "learning_rate": 0.0004201083916083916,
+      "loss": 3.2401,
+      "step": 51550
+    },
+    {
+      "epoch": 15.0262100297047,
+      "grad_norm": 0.3808425962924957,
+      "learning_rate": 0.00041993356643356644,
+      "loss": 3.2133,
+      "step": 51600
+    },
+    {
+      "epoch": 15.040771157318423,
+      "grad_norm": 0.3673969805240631,
+      "learning_rate": 0.0004197587412587412,
+      "loss": 3.2078,
+      "step": 51650
+    },
+    {
+      "epoch": 15.055332284932145,
+      "grad_norm": 0.37854793667793274,
+      "learning_rate": 0.00041958391608391604,
+      "loss": 3.2111,
+      "step": 51700
+    },
+    {
+      "epoch": 15.069893412545868,
+      "grad_norm": 0.36936038732528687,
+      "learning_rate": 0.00041940909090909084,
+      "loss": 3.2208,
+      "step": 51750
+    },
+    {
+      "epoch": 15.084454540159589,
+      "grad_norm": 0.36692115664482117,
+      "learning_rate": 0.0004192342657342657,
+      "loss": 3.229,
+      "step": 51800
+    },
+    {
+      "epoch": 15.099015667773312,
+      "grad_norm": 0.34014618396759033,
+      "learning_rate": 0.0004190594405594405,
+      "loss": 3.2281,
+      "step": 51850
+    },
+    {
+      "epoch": 15.113576795387035,
+      "grad_norm": 0.3525432348251343,
+      "learning_rate": 0.00041888461538461535,
+      "loss": 3.2413,
+      "step": 51900
+    },
+    {
+      "epoch": 15.128137923000757,
+      "grad_norm": 0.3579222857952118,
+      "learning_rate": 0.00041870979020979015,
+      "loss": 3.2443,
+      "step": 51950
+    },
+    {
+      "epoch": 15.14269905061448,
+      "grad_norm": 0.36283615231513977,
+      "learning_rate": 0.000418534965034965,
+      "loss": 3.2323,
+      "step": 52000
+    },
+    {
+      "epoch": 15.14269905061448,
+      "eval_accuracy": 0.3711156432802507,
+      "eval_loss": 3.5574755668640137,
+      "eval_runtime": 180.7629,
+      "eval_samples_per_second": 92.087,
+      "eval_steps_per_second": 5.759,
+      "step": 52000
+    },
+    {
+      "epoch": 15.157260178228203,
+      "grad_norm": 0.3686647415161133,
+      "learning_rate": 0.00041836013986013985,
+      "loss": 3.2459,
+      "step": 52050
+    },
+    {
+      "epoch": 15.171821305841924,
+      "grad_norm": 0.35007837414741516,
+      "learning_rate": 0.00041818531468531465,
+      "loss": 3.2433,
+      "step": 52100
+    },
+    {
+      "epoch": 15.186382433455647,
+      "grad_norm": 0.3874566853046417,
+      "learning_rate": 0.0004180104895104895,
+      "loss": 3.238,
+      "step": 52150
+    },
+    {
+      "epoch": 15.20094356106937,
+      "grad_norm": 0.35891878604888916,
+      "learning_rate": 0.0004178356643356643,
+      "loss": 3.2537,
+      "step": 52200
+    },
+    {
+      "epoch": 15.215504688683092,
+      "grad_norm": 0.35012251138687134,
+      "learning_rate": 0.00041766083916083916,
+      "loss": 3.2566,
+      "step": 52250
+    },
+    {
+      "epoch": 15.230065816296815,
+      "grad_norm": 0.37354549765586853,
+      "learning_rate": 0.00041748601398601396,
+      "loss": 3.2547,
+      "step": 52300
+    },
+    {
+      "epoch": 15.244626943910536,
+      "grad_norm": 0.3650526702404022,
+      "learning_rate": 0.0004173111888111888,
+      "loss": 3.2521,
+      "step": 52350
+    },
+    {
+      "epoch": 15.259188071524258,
+      "grad_norm": 0.43291184306144714,
+      "learning_rate": 0.00041713636363636356,
+      "loss": 3.253,
+      "step": 52400
+    },
+    {
+      "epoch": 15.273749199137981,
+      "grad_norm": 0.36537203192710876,
+      "learning_rate": 0.0004169615384615384,
+      "loss": 3.2628,
+      "step": 52450
+    },
+    {
+      "epoch": 15.288310326751704,
+      "grad_norm": 0.3516569137573242,
+      "learning_rate": 0.0004167867132867132,
+      "loss": 3.2648,
+      "step": 52500
+    },
+    {
+      "epoch": 15.302871454365427,
+      "grad_norm": 0.3715130090713501,
+      "learning_rate": 0.00041661188811188807,
+      "loss": 3.2721,
+      "step": 52550
+    },
+    {
+      "epoch": 15.317432581979148,
+      "grad_norm": 0.3508760631084442,
+      "learning_rate": 0.00041643706293706287,
+      "loss": 3.2753,
+      "step": 52600
+    },
+    {
+      "epoch": 15.33199370959287,
+      "grad_norm": 0.36405643820762634,
+      "learning_rate": 0.0004162622377622377,
+      "loss": 3.2591,
+      "step": 52650
+    },
+    {
+      "epoch": 15.346554837206593,
+      "grad_norm": 0.3771628737449646,
+      "learning_rate": 0.0004160874125874126,
+      "loss": 3.2705,
+      "step": 52700
+    },
+    {
+      "epoch": 15.361115964820316,
+      "grad_norm": 0.3789055347442627,
+      "learning_rate": 0.0004159125874125874,
+      "loss": 3.29,
+      "step": 52750
+    },
+    {
+      "epoch": 15.375677092434039,
+      "grad_norm": 0.35194921493530273,
+      "learning_rate": 0.00041573776223776223,
+      "loss": 3.28,
+      "step": 52800
+    },
+    {
+      "epoch": 15.39023822004776,
+      "grad_norm": 0.39530250430107117,
+      "learning_rate": 0.00041556293706293703,
+      "loss": 3.2666,
+      "step": 52850
+    },
+    {
+      "epoch": 15.404799347661482,
+      "grad_norm": 0.35562649369239807,
+      "learning_rate": 0.0004153881118881119,
+      "loss": 3.2888,
+      "step": 52900
+    },
+    {
+      "epoch": 15.419360475275205,
+      "grad_norm": 0.3992999792098999,
+      "learning_rate": 0.0004152132867132867,
+      "loss": 3.2757,
+      "step": 52950
+    },
+    {
+      "epoch": 15.433921602888928,
+      "grad_norm": 0.3915247321128845,
+      "learning_rate": 0.00041503846153846154,
+      "loss": 3.2795,
+      "step": 53000
+    },
+    {
+      "epoch": 15.433921602888928,
+      "eval_accuracy": 0.3716249244954154,
+      "eval_loss": 3.551257371902466,
+      "eval_runtime": 180.9509,
+      "eval_samples_per_second": 91.992,
+      "eval_steps_per_second": 5.753,
+      "step": 53000
+    },
+    {
+      "epoch": 15.44848273050265,
+      "grad_norm": 0.3508327305316925,
+      "learning_rate": 0.00041486363636363634,
+      "loss": 3.281,
+      "step": 53050
+    },
+    {
+      "epoch": 15.463043858116373,
+      "grad_norm": 0.37493276596069336,
+      "learning_rate": 0.0004146888111888112,
+      "loss": 3.2868,
+      "step": 53100
+    },
+    {
+      "epoch": 15.477604985730094,
+      "grad_norm": 0.38858386874198914,
+      "learning_rate": 0.00041451398601398593,
+      "loss": 3.2819,
+      "step": 53150
+    },
+    {
+      "epoch": 15.492166113343817,
+      "grad_norm": 0.36600261926651,
+      "learning_rate": 0.0004143391608391608,
+      "loss": 3.2751,
+      "step": 53200
+    },
+    {
+      "epoch": 15.50672724095754,
+      "grad_norm": 0.38814976811408997,
+      "learning_rate": 0.0004141643356643356,
+      "loss": 3.2928,
+      "step": 53250
+    },
+    {
+      "epoch": 15.521288368571263,
+      "grad_norm": 0.35792145133018494,
+      "learning_rate": 0.00041398951048951044,
+      "loss": 3.2845,
+      "step": 53300
+    },
+    {
+      "epoch": 15.535849496184985,
+      "grad_norm": 0.37970036268234253,
+      "learning_rate": 0.00041381468531468524,
+      "loss": 3.2981,
+      "step": 53350
+    },
+    {
+      "epoch": 15.550410623798706,
+      "grad_norm": 0.3698292672634125,
+      "learning_rate": 0.0004136398601398601,
+      "loss": 3.2764,
+      "step": 53400
+    },
+    {
+      "epoch": 15.564971751412429,
+      "grad_norm": 0.3558374047279358,
+      "learning_rate": 0.00041346503496503495,
+      "loss": 3.2896,
+      "step": 53450
+    },
+    {
+      "epoch": 15.579532879026152,
+      "grad_norm": Infinity,
+      "learning_rate": 0.00041329020979020975,
+      "loss": 3.2946,
+      "step": 53500
+    },
+    {
+      "epoch": 15.594094006639875,
+      "grad_norm": 0.37755706906318665,
+      "learning_rate": 0.0004131153846153846,
+      "loss": 3.2922,
+      "step": 53550
+    },
+    {
+      "epoch": 15.608655134253597,
+      "grad_norm": 0.3373359143733978,
+      "learning_rate": 0.0004129405594405594,
+      "loss": 3.2973,
+      "step": 53600
+    },
+    {
+      "epoch": 15.623216261867318,
+      "grad_norm": 0.35794350504875183,
+      "learning_rate": 0.00041276573426573426,
+      "loss": 3.3028,
+      "step": 53650
+    },
+    {
+      "epoch": 15.637777389481041,
+      "grad_norm": 0.3629034757614136,
+      "learning_rate": 0.00041259090909090906,
+      "loss": 3.292,
+      "step": 53700
+    },
+    {
+      "epoch": 15.652338517094764,
+      "grad_norm": 0.3656299412250519,
+      "learning_rate": 0.0004124160839160839,
+      "loss": 3.2817,
+      "step": 53750
+    },
+    {
+      "epoch": 15.666899644708487,
+      "grad_norm": 0.37915611267089844,
+      "learning_rate": 0.0004122412587412587,
+      "loss": 3.3044,
+      "step": 53800
+    },
+    {
+      "epoch": 15.68146077232221,
+      "grad_norm": 0.38034093379974365,
+      "learning_rate": 0.00041206643356643356,
+      "loss": 3.2931,
+      "step": 53850
+    },
+    {
+      "epoch": 15.69602189993593,
+      "grad_norm": 0.39361733198165894,
+      "learning_rate": 0.0004118916083916083,
+      "loss": 3.2864,
+      "step": 53900
+    },
+    {
+      "epoch": 15.710583027549653,
+      "grad_norm": 0.40040066838264465,
+      "learning_rate": 0.00041171678321678316,
+      "loss": 3.2923,
+      "step": 53950
+    },
+    {
+      "epoch": 15.725144155163376,
+      "grad_norm": 0.33866026997566223,
+      "learning_rate": 0.00041154195804195796,
+      "loss": 3.3018,
+      "step": 54000
+    },
+    {
+      "epoch": 15.725144155163376,
+      "eval_accuracy": 0.3721960436420614,
+      "eval_loss": 3.541879653930664,
+      "eval_runtime": 180.9075,
+      "eval_samples_per_second": 92.014,
+      "eval_steps_per_second": 5.754,
+      "step": 54000
+    },
+    {
+      "epoch": 15.739705282777098,
+      "grad_norm": 0.36742034554481506,
+      "learning_rate": 0.0004113671328671328,
+      "loss": 3.2937,
+      "step": 54050
+    },
+    {
+      "epoch": 15.754266410390821,
+      "grad_norm": 0.3519640862941742,
+      "learning_rate": 0.00041119230769230767,
+      "loss": 3.3098,
+      "step": 54100
+    },
+    {
+      "epoch": 15.768827538004544,
+      "grad_norm": 0.356086790561676,
+      "learning_rate": 0.00041101748251748247,
+      "loss": 3.3104,
+      "step": 54150
+    },
+    {
+      "epoch": 15.783388665618265,
+      "grad_norm": 0.3600098490715027,
+      "learning_rate": 0.0004108426573426573,
+      "loss": 3.3047,
+      "step": 54200
+    },
+    {
+      "epoch": 15.797949793231988,
+      "grad_norm": 0.3581182062625885,
+      "learning_rate": 0.0004106678321678321,
+      "loss": 3.3121,
+      "step": 54250
+    },
+    {
+      "epoch": 15.81251092084571,
+      "grad_norm": 0.3515772521495819,
+      "learning_rate": 0.000410493006993007,
+      "loss": 3.3019,
+      "step": 54300
+    },
+    {
+      "epoch": 15.827072048459433,
+      "grad_norm": 0.40560662746429443,
+      "learning_rate": 0.0004103181818181818,
+      "loss": 3.3104,
+      "step": 54350
+    },
+    {
+      "epoch": 15.841633176073156,
+      "grad_norm": 0.3703998029232025,
+      "learning_rate": 0.00041014335664335663,
+      "loss": 3.3006,
+      "step": 54400
+    },
+    {
+      "epoch": 15.856194303686877,
+      "grad_norm": 0.35367172956466675,
+      "learning_rate": 0.00040996853146853143,
+      "loss": 3.2965,
+      "step": 54450
+    },
+    {
+      "epoch": 15.8707554313006,
+      "grad_norm": 0.3762366473674774,
+      "learning_rate": 0.0004097937062937063,
+      "loss": 3.2994,
+      "step": 54500
+    },
+    {
+      "epoch": 15.885316558914322,
+      "grad_norm": 0.3624609708786011,
+      "learning_rate": 0.0004096188811188811,
+      "loss": 3.3033,
+      "step": 54550
+    },
+    {
+      "epoch": 15.899877686528045,
+      "grad_norm": 0.36504390835762024,
+      "learning_rate": 0.00040944405594405594,
+      "loss": 3.3073,
+      "step": 54600
+    },
+    {
+      "epoch": 15.914438814141768,
+      "grad_norm": 0.36164841055870056,
+      "learning_rate": 0.0004092692307692307,
+      "loss": 3.311,
+      "step": 54650
+    },
+    {
+      "epoch": 15.928999941755489,
+      "grad_norm": 0.35146453976631165,
+      "learning_rate": 0.00040909440559440554,
+      "loss": 3.3255,
+      "step": 54700
+    },
+    {
+      "epoch": 15.943561069369212,
+      "grad_norm": 0.36682817339897156,
+      "learning_rate": 0.00040891958041958034,
+      "loss": 3.3056,
+      "step": 54750
+    },
+    {
+      "epoch": 15.958122196982934,
+      "grad_norm": 0.375960111618042,
+      "learning_rate": 0.0004087447552447552,
+      "loss": 3.2998,
+      "step": 54800
+    },
+    {
+      "epoch": 15.972683324596657,
+      "grad_norm": 0.37814897298812866,
+      "learning_rate": 0.00040856993006993005,
+      "loss": 3.3165,
+      "step": 54850
+    },
+    {
+      "epoch": 15.98724445221038,
+      "grad_norm": 0.3649648129940033,
+      "learning_rate": 0.00040839510489510485,
+      "loss": 3.305,
+      "step": 54900
+    },
+    {
+      "epoch": 16.001747335313645,
+      "grad_norm": 0.38327813148498535,
+      "learning_rate": 0.0004082202797202797,
+      "loss": 3.3029,
+      "step": 54950
+    },
+    {
+      "epoch": 16.01630846292737,
+      "grad_norm": 0.3484591841697693,
+      "learning_rate": 0.0004080454545454545,
+      "loss": 3.2195,
+      "step": 55000
+    },
+    {
+      "epoch": 16.01630846292737,
+      "eval_accuracy": 0.37155215324144797,
+      "eval_loss": 3.5542728900909424,
+      "eval_runtime": 180.9535,
+      "eval_samples_per_second": 91.99,
+      "eval_steps_per_second": 5.753,
+      "step": 55000
+    },
+    {
+      "epoch": 16.03086959054109,
+      "grad_norm": 0.37213021516799927,
+      "learning_rate": 0.00040787062937062935,
+      "loss": 3.1973,
+      "step": 55050
+    },
+    {
+      "epoch": 16.045430718154815,
+      "grad_norm": 0.3620660603046417,
+      "learning_rate": 0.00040769580419580415,
+      "loss": 3.2098,
+      "step": 55100
+    },
+    {
+      "epoch": 16.059991845768536,
+      "grad_norm": 0.39267122745513916,
+      "learning_rate": 0.000407520979020979,
+      "loss": 3.2128,
+      "step": 55150
+    },
+    {
+      "epoch": 16.074552973382257,
+      "grad_norm": 0.3540675938129425,
+      "learning_rate": 0.0004073461538461538,
+      "loss": 3.211,
+      "step": 55200
+    },
+    {
+      "epoch": 16.08911410099598,
+      "grad_norm": 0.35271748900413513,
+      "learning_rate": 0.00040717132867132866,
+      "loss": 3.2222,
+      "step": 55250
+    },
+    {
+      "epoch": 16.103675228609703,
+      "grad_norm": 0.3637480139732361,
+      "learning_rate": 0.00040699650349650346,
+      "loss": 3.223,
+      "step": 55300
+    },
+    {
+      "epoch": 16.118236356223427,
+      "grad_norm": 0.39908871054649353,
+      "learning_rate": 0.0004068216783216783,
+      "loss": 3.2375,
+      "step": 55350
+    },
+    {
+      "epoch": 16.132797483837148,
+      "grad_norm": 0.37379196286201477,
+      "learning_rate": 0.00040664685314685306,
+      "loss": 3.2193,
+      "step": 55400
+    },
+    {
+      "epoch": 16.14735861145087,
+      "grad_norm": 0.36621153354644775,
+      "learning_rate": 0.0004064720279720279,
+      "loss": 3.2247,
+      "step": 55450
+    },
+    {
+      "epoch": 16.161919739064594,
+      "grad_norm": 0.3897995352745056,
+      "learning_rate": 0.00040629720279720277,
+      "loss": 3.2235,
+      "step": 55500
+    },
+    {
+      "epoch": 16.176480866678315,
+      "grad_norm": 0.384102463722229,
+      "learning_rate": 0.00040612237762237757,
+      "loss": 3.2327,
+      "step": 55550
+    },
+    {
+      "epoch": 16.19104199429204,
+      "grad_norm": 0.3749612271785736,
+      "learning_rate": 0.0004059475524475524,
+      "loss": 3.2375,
+      "step": 55600
+    },
+    {
+      "epoch": 16.20560312190576,
+      "grad_norm": 0.37099507451057434,
+      "learning_rate": 0.0004057727272727272,
+      "loss": 3.2415,
+      "step": 55650
+    },
+    {
+      "epoch": 16.22016424951948,
+      "grad_norm": 0.39269089698791504,
+      "learning_rate": 0.0004055979020979021,
+      "loss": 3.2305,
+      "step": 55700
+    },
+    {
+      "epoch": 16.234725377133206,
+      "grad_norm": 0.38968995213508606,
+      "learning_rate": 0.0004054230769230769,
+      "loss": 3.2416,
+      "step": 55750
+    },
+    {
+      "epoch": 16.249286504746927,
+      "grad_norm": 0.3777758777141571,
+      "learning_rate": 0.00040524825174825173,
+      "loss": 3.244,
+      "step": 55800
+    },
+    {
+      "epoch": 16.26384763236065,
+      "grad_norm": 0.3755890429019928,
+      "learning_rate": 0.00040507342657342653,
+      "loss": 3.2563,
+      "step": 55850
+    },
+    {
+      "epoch": 16.278408759974372,
+      "grad_norm": 0.3695318102836609,
+      "learning_rate": 0.0004048986013986014,
+      "loss": 3.2446,
+      "step": 55900
+    },
+    {
+      "epoch": 16.292969887588093,
+      "grad_norm": 0.33629676699638367,
+      "learning_rate": 0.0004047237762237762,
+      "loss": 3.2445,
+      "step": 55950
+    },
+    {
+      "epoch": 16.307531015201818,
+      "grad_norm": 0.35609471797943115,
+      "learning_rate": 0.00040454895104895104,
+      "loss": 3.253,
+      "step": 56000
+    },
+    {
+      "epoch": 16.307531015201818,
+      "eval_accuracy": 0.37172932009076776,
+      "eval_loss": 3.5524020195007324,
+      "eval_runtime": 180.9583,
+      "eval_samples_per_second": 91.988,
+      "eval_steps_per_second": 5.753,
+      "step": 56000
+    },
+    {
+      "epoch": 16.32209214281554,
+      "grad_norm": 0.379630446434021,
+      "learning_rate": 0.00040437412587412583,
+      "loss": 3.2563,
+      "step": 56050
+    },
+    {
+      "epoch": 16.336653270429263,
+      "grad_norm": 0.42045262455940247,
+      "learning_rate": 0.0004041993006993007,
+      "loss": 3.267,
+      "step": 56100
+    },
+    {
+      "epoch": 16.351214398042984,
+      "grad_norm": 0.3641156554222107,
+      "learning_rate": 0.00040402447552447554,
+      "loss": 3.253,
+      "step": 56150
+    },
+    {
+      "epoch": 16.36577552565671,
+      "grad_norm": 0.3764376640319824,
+      "learning_rate": 0.0004038496503496503,
+      "loss": 3.2674,
+      "step": 56200
+    },
+    {
+      "epoch": 16.38033665327043,
+      "grad_norm": 0.38018858432769775,
+      "learning_rate": 0.00040367482517482514,
+      "loss": 3.2573,
+      "step": 56250
+    },
+    {
+      "epoch": 16.39489778088415,
+      "grad_norm": 0.3741246163845062,
+      "learning_rate": 0.00040349999999999994,
+      "loss": 3.2688,
+      "step": 56300
+    },
+    {
+      "epoch": 16.409458908497875,
+      "grad_norm": 0.36829307675361633,
+      "learning_rate": 0.0004033251748251748,
+      "loss": 3.2688,
+      "step": 56350
+    },
+    {
+      "epoch": 16.424020036111596,
+      "grad_norm": 0.37031373381614685,
+      "learning_rate": 0.0004031503496503496,
+      "loss": 3.2537,
+      "step": 56400
+    },
+    {
+      "epoch": 16.43858116372532,
+      "grad_norm": 0.4177717864513397,
+      "learning_rate": 0.00040297552447552445,
+      "loss": 3.2683,
+      "step": 56450
+    },
+    {
+      "epoch": 16.45314229133904,
+      "grad_norm": 0.3726315498352051,
+      "learning_rate": 0.00040280069930069925,
+      "loss": 3.261,
+      "step": 56500
+    },
+    {
+      "epoch": 16.467703418952762,
+      "grad_norm": 0.36388418078422546,
+      "learning_rate": 0.0004026258741258741,
+      "loss": 3.2684,
+      "step": 56550
+    },
+    {
+      "epoch": 16.482264546566487,
+      "grad_norm": 0.35378575325012207,
+      "learning_rate": 0.0004024510489510489,
+      "loss": 3.2733,
+      "step": 56600
+    },
+    {
+      "epoch": 16.496825674180208,
+      "grad_norm": 0.3618945777416229,
+      "learning_rate": 0.00040227622377622376,
+      "loss": 3.2709,
+      "step": 56650
+    },
+    {
+      "epoch": 16.511386801793932,
+      "grad_norm": 0.4058282673358917,
+      "learning_rate": 0.00040210139860139856,
+      "loss": 3.2689,
+      "step": 56700
+    },
+    {
+      "epoch": 16.525947929407653,
+      "grad_norm": 0.3674890995025635,
+      "learning_rate": 0.0004019265734265734,
+      "loss": 3.2723,
+      "step": 56750
+    },
+    {
+      "epoch": 16.540509057021374,
+      "grad_norm": 0.3580159842967987,
+      "learning_rate": 0.0004017517482517482,
+      "loss": 3.2853,
+      "step": 56800
+    },
+    {
+      "epoch": 16.5550701846351,
+      "grad_norm": 0.39147987961769104,
+      "learning_rate": 0.00040157692307692306,
+      "loss": 3.2778,
+      "step": 56850
+    },
+    {
+      "epoch": 16.56963131224882,
+      "grad_norm": 0.3897418975830078,
+      "learning_rate": 0.0004014020979020979,
+      "loss": 3.276,
+      "step": 56900
+    },
+    {
+      "epoch": 16.584192439862544,
+      "grad_norm": 0.36115020513534546,
+      "learning_rate": 0.00040122727272727266,
+      "loss": 3.2842,
+      "step": 56950
+    },
+    {
+      "epoch": 16.598753567476265,
+      "grad_norm": 0.34980452060699463,
+      "learning_rate": 0.0004010524475524475,
+      "loss": 3.2781,
+      "step": 57000
+    },
+    {
+      "epoch": 16.598753567476265,
+      "eval_accuracy": 0.3722949137948669,
+      "eval_loss": 3.5432724952697754,
+      "eval_runtime": 180.9348,
+      "eval_samples_per_second": 92.0,
+      "eval_steps_per_second": 5.753,
+      "step": 57000
+    },
+    {
+      "epoch": 16.613314695089986,
+      "grad_norm": 0.3930431604385376,
+      "learning_rate": 0.0004008776223776223,
+      "loss": 3.2808,
+      "step": 57050
+    },
+    {
+      "epoch": 16.62787582270371,
+      "grad_norm": 0.377023309469223,
+      "learning_rate": 0.00040070279720279717,
+      "loss": 3.2813,
+      "step": 57100
+    },
+    {
+      "epoch": 16.642436950317432,
+      "grad_norm": 0.3362906277179718,
+      "learning_rate": 0.00040052797202797197,
+      "loss": 3.2755,
+      "step": 57150
+    },
+    {
+      "epoch": 16.656998077931156,
+      "grad_norm": 0.3678101599216461,
+      "learning_rate": 0.0004003531468531468,
+      "loss": 3.275,
+      "step": 57200
+    },
+    {
+      "epoch": 16.671559205544877,
+      "grad_norm": 0.3514406979084015,
+      "learning_rate": 0.0004001783216783216,
+      "loss": 3.2848,
+      "step": 57250
+    },
+    {
+      "epoch": 16.6861203331586,
+      "grad_norm": 0.34927234053611755,
+      "learning_rate": 0.0004000034965034965,
+      "loss": 3.2896,
+      "step": 57300
+    },
+    {
+      "epoch": 16.700681460772323,
+      "grad_norm": 0.35985633730888367,
+      "learning_rate": 0.0003998286713286713,
+      "loss": 3.2758,
+      "step": 57350
+    },
+    {
+      "epoch": 16.715242588386044,
+      "grad_norm": 0.38298559188842773,
+      "learning_rate": 0.00039965384615384613,
+      "loss": 3.2831,
+      "step": 57400
+    },
+    {
+      "epoch": 16.72980371599977,
+      "grad_norm": 0.39080435037612915,
+      "learning_rate": 0.00039947902097902093,
+      "loss": 3.2831,
+      "step": 57450
+    },
+    {
+      "epoch": 16.74436484361349,
+      "grad_norm": 0.35113897919654846,
+      "learning_rate": 0.0003993041958041958,
+      "loss": 3.284,
+      "step": 57500
+    },
+    {
+      "epoch": 16.75892597122721,
+      "grad_norm": 0.34621018171310425,
+      "learning_rate": 0.00039912937062937064,
+      "loss": 3.2853,
+      "step": 57550
+    },
+    {
+      "epoch": 16.773487098840935,
+      "grad_norm": 0.40347158908843994,
+      "learning_rate": 0.00039895454545454544,
+      "loss": 3.2903,
+      "step": 57600
+    },
+    {
+      "epoch": 16.788048226454656,
+      "grad_norm": 0.3585277199745178,
+      "learning_rate": 0.0003987797202797203,
+      "loss": 3.292,
+      "step": 57650
+    },
+    {
+      "epoch": 16.80260935406838,
+      "grad_norm": 0.3407650589942932,
+      "learning_rate": 0.00039860489510489504,
+      "loss": 3.2895,
+      "step": 57700
+    },
+    {
+      "epoch": 16.8171704816821,
+      "grad_norm": 0.36110949516296387,
+      "learning_rate": 0.0003984300699300699,
+      "loss": 3.2995,
+      "step": 57750
+    },
+    {
+      "epoch": 16.831731609295822,
+      "grad_norm": 0.3790327310562134,
+      "learning_rate": 0.0003982552447552447,
+      "loss": 3.2806,
+      "step": 57800
+    },
+    {
+      "epoch": 16.846292736909547,
+      "grad_norm": 0.3504062592983246,
+      "learning_rate": 0.00039808041958041955,
+      "loss": 3.2835,
+      "step": 57850
+    },
+    {
+      "epoch": 16.860853864523268,
+      "grad_norm": 0.34082862734794617,
+      "learning_rate": 0.00039790559440559435,
+      "loss": 3.2844,
+      "step": 57900
+    },
+    {
+      "epoch": 16.875414992136992,
+      "grad_norm": 0.4020354151725769,
+      "learning_rate": 0.0003977307692307692,
+      "loss": 3.2997,
+      "step": 57950
+    },
+    {
+      "epoch": 16.889976119750713,
+      "grad_norm": 0.3737473487854004,
+      "learning_rate": 0.000397555944055944,
+      "loss": 3.2999,
+      "step": 58000
+    },
+    {
+      "epoch": 16.889976119750713,
+      "eval_accuracy": 0.3727604617200867,
+      "eval_loss": 3.5354135036468506,
+      "eval_runtime": 180.934,
+      "eval_samples_per_second": 92.0,
+      "eval_steps_per_second": 5.753,
+      "step": 58000
+    },
+    {
+      "epoch": 16.904537247364434,
+      "grad_norm": 0.3592594563961029,
+      "learning_rate": 0.00039738111888111885,
+      "loss": 3.2993,
+      "step": 58050
+    },
+    {
+      "epoch": 16.91909837497816,
+      "grad_norm": 0.3586547076702118,
+      "learning_rate": 0.00039720629370629365,
+      "loss": 3.2973,
+      "step": 58100
+    },
+    {
+      "epoch": 16.93365950259188,
+      "grad_norm": 0.3590955138206482,
+      "learning_rate": 0.0003970314685314685,
+      "loss": 3.2927,
+      "step": 58150
+    },
+    {
+      "epoch": 16.948220630205604,
+      "grad_norm": 0.38664183020591736,
+      "learning_rate": 0.0003968566433566433,
+      "loss": 3.2935,
+      "step": 58200
+    },
+    {
+      "epoch": 16.962781757819325,
+      "grad_norm": 0.38259342312812805,
+      "learning_rate": 0.00039668181818181816,
+      "loss": 3.2941,
+      "step": 58250
+    },
+    {
+      "epoch": 16.977342885433046,
+      "grad_norm": 0.3696136474609375,
+      "learning_rate": 0.000396506993006993,
+      "loss": 3.2981,
+      "step": 58300
+    },
+    {
+      "epoch": 16.99190401304677,
+      "grad_norm": 0.3578000068664551,
+      "learning_rate": 0.0003963321678321678,
+      "loss": 3.292,
+      "step": 58350
+    },
+    {
+      "epoch": 17.006406896150036,
+      "grad_norm": 0.38829582929611206,
+      "learning_rate": 0.00039615734265734267,
+      "loss": 3.256,
+      "step": 58400
+    },
+    {
+      "epoch": 17.02096802376376,
+      "grad_norm": 0.3556952178478241,
+      "learning_rate": 0.0003959825174825174,
+      "loss": 3.1862,
+      "step": 58450
+    },
+    {
+      "epoch": 17.03552915137748,
+      "grad_norm": 0.36325693130493164,
+      "learning_rate": 0.00039580769230769227,
+      "loss": 3.1957,
+      "step": 58500
+    },
+    {
+      "epoch": 17.050090278991206,
+      "grad_norm": 0.3780340552330017,
+      "learning_rate": 0.00039563286713286707,
+      "loss": 3.1984,
+      "step": 58550
+    },
+    {
+      "epoch": 17.064651406604927,
+      "grad_norm": 0.3776744306087494,
+      "learning_rate": 0.0003954580419580419,
+      "loss": 3.2022,
+      "step": 58600
+    },
+    {
+      "epoch": 17.07921253421865,
+      "grad_norm": 0.37272578477859497,
+      "learning_rate": 0.0003952832167832167,
+      "loss": 3.2085,
+      "step": 58650
+    },
+    {
+      "epoch": 17.093773661832373,
+      "grad_norm": 0.3816814124584198,
+      "learning_rate": 0.0003951083916083916,
+      "loss": 3.2128,
+      "step": 58700
+    },
+    {
+      "epoch": 17.108334789446094,
+      "grad_norm": 0.3480782210826874,
+      "learning_rate": 0.0003949335664335664,
+      "loss": 3.2202,
+      "step": 58750
+    },
+    {
+      "epoch": 17.122895917059818,
+      "grad_norm": 0.4132950007915497,
+      "learning_rate": 0.00039475874125874123,
+      "loss": 3.2165,
+      "step": 58800
+    },
+    {
+      "epoch": 17.13745704467354,
+      "grad_norm": 0.3595176935195923,
+      "learning_rate": 0.00039458391608391603,
+      "loss": 3.22,
+      "step": 58850
+    },
+    {
+      "epoch": 17.152018172287264,
+      "grad_norm": 0.3662016689777374,
+      "learning_rate": 0.0003944090909090909,
+      "loss": 3.2138,
+      "step": 58900
+    },
+    {
+      "epoch": 17.166579299900985,
+      "grad_norm": 0.36622393131256104,
+      "learning_rate": 0.00039423426573426573,
+      "loss": 3.2161,
+      "step": 58950
+    },
+    {
+      "epoch": 17.181140427514705,
+      "grad_norm": 0.37797629833221436,
+      "learning_rate": 0.00039405944055944053,
+      "loss": 3.2191,
+      "step": 59000
+    },
+    {
+      "epoch": 17.181140427514705,
+      "eval_accuracy": 0.3717471896070893,
+      "eval_loss": 3.555344820022583,
+      "eval_runtime": 180.7239,
+      "eval_samples_per_second": 92.107,
+      "eval_steps_per_second": 5.76,
+      "step": 59000
+    },
+    {
+      "epoch": 17.19570155512843,
+      "grad_norm": 0.39831972122192383,
+      "learning_rate": 0.0003938846153846154,
+      "loss": 3.2431,
+      "step": 59050
+    },
+    {
+      "epoch": 17.21026268274215,
+      "grad_norm": 0.3561229109764099,
+      "learning_rate": 0.0003937097902097902,
+      "loss": 3.2428,
+      "step": 59100
+    },
+    {
+      "epoch": 17.224823810355876,
+      "grad_norm": 0.35875722765922546,
+      "learning_rate": 0.00039353496503496504,
+      "loss": 3.2266,
+      "step": 59150
+    },
+    {
+      "epoch": 17.239384937969596,
+      "grad_norm": 0.4089879095554352,
+      "learning_rate": 0.0003933601398601398,
+      "loss": 3.2269,
+      "step": 59200
+    },
+    {
+      "epoch": 17.253946065583317,
+      "grad_norm": 0.37490785121917725,
+      "learning_rate": 0.00039318531468531464,
+      "loss": 3.2422,
+      "step": 59250
+    },
+    {
+      "epoch": 17.268507193197042,
+      "grad_norm": 0.36489033699035645,
+      "learning_rate": 0.00039301048951048944,
+      "loss": 3.2319,
+      "step": 59300
+    },
+    {
+      "epoch": 17.283068320810763,
+      "grad_norm": 0.4006343185901642,
+      "learning_rate": 0.0003928356643356643,
+      "loss": 3.2394,
+      "step": 59350
+    },
+    {
+      "epoch": 17.297629448424487,
+      "grad_norm": 0.4032607078552246,
+      "learning_rate": 0.0003926608391608391,
+      "loss": 3.2427,
+      "step": 59400
+    },
+    {
+      "epoch": 17.31219057603821,
+      "grad_norm": 0.3809773623943329,
+      "learning_rate": 0.00039248601398601395,
+      "loss": 3.2293,
+      "step": 59450
+    },
+    {
+      "epoch": 17.32675170365193,
+      "grad_norm": 0.3603288531303406,
+      "learning_rate": 0.00039231118881118875,
+      "loss": 3.2487,
+      "step": 59500
+    },
+    {
+      "epoch": 17.341312831265654,
+      "grad_norm": 0.3441549837589264,
+      "learning_rate": 0.0003921363636363636,
+      "loss": 3.2395,
+      "step": 59550
+    },
+    {
+      "epoch": 17.355873958879375,
+      "grad_norm": 0.37875792384147644,
+      "learning_rate": 0.00039196153846153846,
+      "loss": 3.2598,
+      "step": 59600
+    },
+    {
+      "epoch": 17.3704350864931,
+      "grad_norm": 0.38208144903182983,
+      "learning_rate": 0.00039178671328671326,
+      "loss": 3.2508,
+      "step": 59650
+    },
+    {
+      "epoch": 17.38499621410682,
+      "grad_norm": 0.3884451389312744,
+      "learning_rate": 0.0003916118881118881,
+      "loss": 3.2471,
+      "step": 59700
+    },
+    {
+      "epoch": 17.39955734172054,
+      "grad_norm": 0.3753611445426941,
+      "learning_rate": 0.0003914370629370629,
+      "loss": 3.2523,
+      "step": 59750
+    },
+    {
+      "epoch": 17.414118469334266,
+      "grad_norm": 0.37695708870887756,
+      "learning_rate": 0.00039126223776223776,
+      "loss": 3.2556,
+      "step": 59800
+    },
+    {
+      "epoch": 17.428679596947987,
+      "grad_norm": 0.3882795572280884,
+      "learning_rate": 0.00039108741258741256,
+      "loss": 3.2442,
+      "step": 59850
+    },
+    {
+      "epoch": 17.44324072456171,
+      "grad_norm": 0.35586661100387573,
+      "learning_rate": 0.0003909125874125874,
+      "loss": 3.2433,
+      "step": 59900
+    },
+    {
+      "epoch": 17.457801852175432,
+      "grad_norm": 0.38463085889816284,
+      "learning_rate": 0.00039073776223776216,
+      "loss": 3.2651,
+      "step": 59950
+    },
+    {
+      "epoch": 17.472362979789153,
+      "grad_norm": 0.38579466938972473,
+      "learning_rate": 0.000390562937062937,
+      "loss": 3.2596,
+      "step": 60000
+    },
+    {
+      "epoch": 17.472362979789153,
+      "eval_accuracy": 0.37229173960446765,
+      "eval_loss": 3.5483200550079346,
+      "eval_runtime": 180.7914,
+      "eval_samples_per_second": 92.073,
+      "eval_steps_per_second": 5.758,
+      "step": 60000
+    },
+    {
+      "epoch": 17.486924107402878,
+      "grad_norm": 0.39465898275375366,
+      "learning_rate": 0.0003903881118881118,
+      "loss": 3.2515,
+      "step": 60050
+    },
+    {
+      "epoch": 17.5014852350166,
+      "grad_norm": 0.37305083870887756,
+      "learning_rate": 0.00039021328671328667,
+      "loss": 3.2657,
+      "step": 60100
+    },
+    {
+      "epoch": 17.516046362630323,
+      "grad_norm": 0.3921613097190857,
+      "learning_rate": 0.00039003846153846147,
+      "loss": 3.2809,
+      "step": 60150
+    },
+    {
+      "epoch": 17.530607490244044,
+      "grad_norm": 0.35129493474960327,
+      "learning_rate": 0.0003898636363636363,
+      "loss": 3.259,
+      "step": 60200
+    },
+    {
+      "epoch": 17.545168617857765,
+      "grad_norm": 0.3839664161205292,
+      "learning_rate": 0.0003896888111888111,
+      "loss": 3.2613,
+      "step": 60250
+    },
+    {
+      "epoch": 17.55972974547149,
+      "grad_norm": 0.3718804717063904,
+      "learning_rate": 0.000389513986013986,
+      "loss": 3.2626,
+      "step": 60300
+    },
+    {
+      "epoch": 17.57429087308521,
+      "grad_norm": 0.4159618318080902,
+      "learning_rate": 0.00038933916083916083,
+      "loss": 3.2682,
+      "step": 60350
+    },
+    {
+      "epoch": 17.588852000698935,
+      "grad_norm": 0.37494173645973206,
+      "learning_rate": 0.00038916433566433563,
+      "loss": 3.2774,
+      "step": 60400
+    },
+    {
+      "epoch": 17.603413128312656,
+      "grad_norm": 0.38625165820121765,
+      "learning_rate": 0.0003889895104895105,
+      "loss": 3.2691,
+      "step": 60450
+    },
+    {
+      "epoch": 17.617974255926377,
+      "grad_norm": 0.39727920293807983,
+      "learning_rate": 0.0003888146853146853,
+      "loss": 3.2739,
+      "step": 60500
+    },
+    {
+      "epoch": 17.6325353835401,
+      "grad_norm": 0.3708936274051666,
+      "learning_rate": 0.00038863986013986014,
+      "loss": 3.2578,
+      "step": 60550
+    },
+    {
+      "epoch": 17.647096511153823,
+      "grad_norm": 0.36158886551856995,
+      "learning_rate": 0.00038846503496503494,
+      "loss": 3.271,
+      "step": 60600
+    },
+    {
+      "epoch": 17.661657638767547,
+      "grad_norm": 0.3670588731765747,
+      "learning_rate": 0.0003882902097902098,
+      "loss": 3.2752,
+      "step": 60650
+    },
+    {
+      "epoch": 17.676218766381268,
+      "grad_norm": 0.3502790629863739,
+      "learning_rate": 0.00038811538461538454,
+      "loss": 3.262,
+      "step": 60700
+    },
+    {
+      "epoch": 17.690779893994993,
+      "grad_norm": 0.38418471813201904,
+      "learning_rate": 0.0003879405594405594,
+      "loss": 3.2687,
+      "step": 60750
+    },
+    {
+      "epoch": 17.705341021608714,
+      "grad_norm": 0.3664897680282593,
+      "learning_rate": 0.0003877657342657342,
+      "loss": 3.2597,
+      "step": 60800
+    },
+    {
+      "epoch": 17.719902149222435,
+      "grad_norm": 0.40013816952705383,
+      "learning_rate": 0.00038759090909090905,
+      "loss": 3.2732,
+      "step": 60850
+    },
+    {
+      "epoch": 17.73446327683616,
+      "grad_norm": 0.3857571482658386,
+      "learning_rate": 0.00038741608391608384,
+      "loss": 3.2733,
+      "step": 60900
+    },
+    {
+      "epoch": 17.74902440444988,
+      "grad_norm": 0.38301730155944824,
+      "learning_rate": 0.0003872412587412587,
+      "loss": 3.2715,
+      "step": 60950
+    },
+    {
+      "epoch": 17.763585532063605,
+      "grad_norm": 0.37607985734939575,
+      "learning_rate": 0.00038706643356643355,
+      "loss": 3.2635,
+      "step": 61000
+    },
+    {
+      "epoch": 17.763585532063605,
+      "eval_accuracy": 0.3731001000928039,
+      "eval_loss": 3.5363667011260986,
+      "eval_runtime": 181.354,
+      "eval_samples_per_second": 91.787,
+      "eval_steps_per_second": 5.74,
+      "step": 61000
+    },
+    {
+      "epoch": 17.778146659677326,
+      "grad_norm": 0.3776940703392029,
+      "learning_rate": 0.00038689160839160835,
+      "loss": 3.2758,
+      "step": 61050
+    },
+    {
+      "epoch": 17.792707787291047,
+      "grad_norm": 0.39629730582237244,
+      "learning_rate": 0.0003867167832167832,
+      "loss": 3.2631,
+      "step": 61100
+    },
+    {
+      "epoch": 17.80726891490477,
+      "grad_norm": 0.3491132855415344,
+      "learning_rate": 0.000386541958041958,
+      "loss": 3.2839,
+      "step": 61150
+    },
+    {
+      "epoch": 17.821830042518492,
+      "grad_norm": 0.37949714064598083,
+      "learning_rate": 0.00038636713286713286,
+      "loss": 3.2783,
+      "step": 61200
+    },
+    {
+      "epoch": 17.836391170132217,
+      "grad_norm": 0.3643863797187805,
+      "learning_rate": 0.00038619230769230766,
+      "loss": 3.2805,
+      "step": 61250
+    },
+    {
+      "epoch": 17.850952297745938,
+      "grad_norm": 0.3382050096988678,
+      "learning_rate": 0.0003860174825174825,
+      "loss": 3.2723,
+      "step": 61300
+    },
+    {
+      "epoch": 17.86551342535966,
+      "grad_norm": 0.3952576816082001,
+      "learning_rate": 0.0003858426573426573,
+      "loss": 3.2847,
+      "step": 61350
+    },
+    {
+      "epoch": 17.880074552973383,
+      "grad_norm": 0.38078320026397705,
+      "learning_rate": 0.00038566783216783217,
+      "loss": 3.2854,
+      "step": 61400
+    },
+    {
+      "epoch": 17.894635680587104,
+      "grad_norm": 0.3551877439022064,
+      "learning_rate": 0.0003854930069930069,
+      "loss": 3.2797,
+      "step": 61450
+    },
+    {
+      "epoch": 17.90919680820083,
+      "grad_norm": 0.3975968360900879,
+      "learning_rate": 0.00038531818181818177,
+      "loss": 3.2771,
+      "step": 61500
+    },
+    {
+      "epoch": 17.92375793581455,
+      "grad_norm": 0.3528117835521698,
+      "learning_rate": 0.00038514335664335657,
+      "loss": 3.2777,
+      "step": 61550
+    },
+    {
+      "epoch": 17.93831906342827,
+      "grad_norm": 0.3516587018966675,
+      "learning_rate": 0.0003849685314685314,
+      "loss": 3.2816,
+      "step": 61600
+    },
+    {
+      "epoch": 17.952880191041995,
+      "grad_norm": 0.37779533863067627,
+      "learning_rate": 0.0003847937062937062,
+      "loss": 3.2731,
+      "step": 61650
+    },
+    {
+      "epoch": 17.967441318655716,
+      "grad_norm": 0.37921205163002014,
+      "learning_rate": 0.0003846188811188811,
+      "loss": 3.2815,
+      "step": 61700
+    },
+    {
+      "epoch": 17.98200244626944,
+      "grad_norm": 0.40624403953552246,
+      "learning_rate": 0.00038444405594405593,
+      "loss": 3.2829,
+      "step": 61750
+    },
+    {
+      "epoch": 17.99656357388316,
+      "grad_norm": 0.3792065680027008,
+      "learning_rate": 0.00038426923076923073,
+      "loss": 3.2762,
+      "step": 61800
+    },
+    {
+      "epoch": 18.01106645698643,
+      "grad_norm": 0.38736584782600403,
+      "learning_rate": 0.0003840944055944056,
+      "loss": 3.2056,
+      "step": 61850
+    },
+    {
+      "epoch": 18.02562758460015,
+      "grad_norm": 0.3663954436779022,
+      "learning_rate": 0.0003839195804195804,
+      "loss": 3.1867,
+      "step": 61900
+    },
+    {
+      "epoch": 18.040188712213872,
+      "grad_norm": 0.36250391602516174,
+      "learning_rate": 0.00038374475524475523,
+      "loss": 3.1847,
+      "step": 61950
+    },
+    {
+      "epoch": 18.054749839827597,
+      "grad_norm": 0.3424190580844879,
+      "learning_rate": 0.00038356993006993003,
+      "loss": 3.1872,
+      "step": 62000
+    },
+    {
+      "epoch": 18.054749839827597,
+      "eval_accuracy": 0.3726784030201364,
+      "eval_loss": 3.5482542514801025,
+      "eval_runtime": 181.0633,
+      "eval_samples_per_second": 91.935,
+      "eval_steps_per_second": 5.749,
+      "step": 62000
+    },
+    {
+      "epoch": 18.069310967441318,
+      "grad_norm": 0.3808369040489197,
+      "learning_rate": 0.0003833951048951049,
+      "loss": 3.1775,
+      "step": 62050
+    },
+    {
+      "epoch": 18.083872095055042,
+      "grad_norm": 0.41652339696884155,
+      "learning_rate": 0.0003832202797202797,
+      "loss": 3.1956,
+      "step": 62100
+    },
+    {
+      "epoch": 18.098433222668763,
+      "grad_norm": 0.3494375944137573,
+      "learning_rate": 0.00038304545454545454,
+      "loss": 3.1895,
+      "step": 62150
+    },
+    {
+      "epoch": 18.112994350282484,
+      "grad_norm": 0.3790935277938843,
+      "learning_rate": 0.0003828706293706293,
+      "loss": 3.1971,
+      "step": 62200
+    },
+    {
+      "epoch": 18.12755547789621,
+      "grad_norm": 0.3747474253177643,
+      "learning_rate": 0.00038269580419580414,
+      "loss": 3.2032,
+      "step": 62250
+    },
+    {
+      "epoch": 18.14211660550993,
+      "grad_norm": 0.38967105746269226,
+      "learning_rate": 0.00038252097902097894,
+      "loss": 3.2032,
+      "step": 62300
+    },
+    {
+      "epoch": 18.156677733123654,
+      "grad_norm": 0.3795710504055023,
+      "learning_rate": 0.0003823461538461538,
+      "loss": 3.2167,
+      "step": 62350
+    },
+    {
+      "epoch": 18.171238860737375,
+      "grad_norm": 0.392669677734375,
+      "learning_rate": 0.00038217132867132865,
+      "loss": 3.209,
+      "step": 62400
+    },
+    {
+      "epoch": 18.185799988351096,
+      "grad_norm": 0.38014981150627136,
+      "learning_rate": 0.00038199650349650345,
+      "loss": 3.2065,
+      "step": 62450
+    },
+    {
+      "epoch": 18.20036111596482,
+      "grad_norm": 0.37931060791015625,
+      "learning_rate": 0.0003818216783216783,
+      "loss": 3.2128,
+      "step": 62500
+    },
+    {
+      "epoch": 18.214922243578542,
+      "grad_norm": 0.3978460729122162,
+      "learning_rate": 0.0003816468531468531,
+      "loss": 3.2153,
+      "step": 62550
+    },
+    {
+      "epoch": 18.229483371192266,
+      "grad_norm": 0.3716878592967987,
+      "learning_rate": 0.00038147202797202796,
+      "loss": 3.2101,
+      "step": 62600
+    },
+    {
+      "epoch": 18.244044498805987,
+      "grad_norm": 0.3965378403663635,
+      "learning_rate": 0.00038129720279720276,
+      "loss": 3.2217,
+      "step": 62650
+    },
+    {
+      "epoch": 18.25860562641971,
+      "grad_norm": 0.4007912576198578,
+      "learning_rate": 0.0003811223776223776,
+      "loss": 3.2219,
+      "step": 62700
+    },
+    {
+      "epoch": 18.273166754033433,
+      "grad_norm": 0.3940446078777313,
+      "learning_rate": 0.0003809475524475524,
+      "loss": 3.2261,
+      "step": 62750
+    },
+    {
+      "epoch": 18.287727881647154,
+      "grad_norm": 0.4010846018791199,
+      "learning_rate": 0.00038077272727272726,
+      "loss": 3.2194,
+      "step": 62800
+    },
+    {
+      "epoch": 18.30228900926088,
+      "grad_norm": 0.39300212264060974,
+      "learning_rate": 0.00038059790209790206,
+      "loss": 3.2288,
+      "step": 62850
+    },
+    {
+      "epoch": 18.3168501368746,
+      "grad_norm": 0.3691205084323883,
+      "learning_rate": 0.0003804230769230769,
+      "loss": 3.2333,
+      "step": 62900
+    },
+    {
+      "epoch": 18.33141126448832,
+      "grad_norm": 0.35388556122779846,
+      "learning_rate": 0.00038024825174825166,
+      "loss": 3.2303,
+      "step": 62950
+    },
+    {
+      "epoch": 18.345972392102045,
+      "grad_norm": 0.3913156986236572,
+      "learning_rate": 0.0003800734265734265,
+      "loss": 3.2398,
+      "step": 63000
+    },
+    {
+      "epoch": 18.345972392102045,
+      "eval_accuracy": 0.3726364331693022,
+      "eval_loss": 3.547825336456299,
+      "eval_runtime": 181.4413,
+      "eval_samples_per_second": 91.743,
+      "eval_steps_per_second": 5.737,
+      "step": 63000
+    },
+    {
+      "epoch": 18.360533519715766,
+      "grad_norm": 0.4170674681663513,
+      "learning_rate": 0.0003798986013986013,
+      "loss": 3.2439,
+      "step": 63050
+    },
+    {
+      "epoch": 18.37509464732949,
+      "grad_norm": 0.3595086932182312,
+      "learning_rate": 0.00037972377622377617,
+      "loss": 3.2187,
+      "step": 63100
+    },
+    {
+      "epoch": 18.38965577494321,
+      "grad_norm": 0.34484726190567017,
+      "learning_rate": 0.000379548951048951,
+      "loss": 3.224,
+      "step": 63150
+    },
+    {
+      "epoch": 18.404216902556932,
+      "grad_norm": 0.3886369466781616,
+      "learning_rate": 0.0003793741258741258,
+      "loss": 3.235,
+      "step": 63200
+    },
+    {
+      "epoch": 18.418778030170657,
+      "grad_norm": 0.3723530173301697,
+      "learning_rate": 0.0003791993006993007,
+      "loss": 3.2362,
+      "step": 63250
+    },
+    {
+      "epoch": 18.433339157784378,
+      "grad_norm": 0.3875845968723297,
+      "learning_rate": 0.0003790244755244755,
+      "loss": 3.2425,
+      "step": 63300
+    },
+    {
+      "epoch": 18.447900285398102,
+      "grad_norm": 0.3876391053199768,
+      "learning_rate": 0.00037884965034965033,
+      "loss": 3.2526,
+      "step": 63350
+    },
+    {
+      "epoch": 18.462461413011823,
+      "grad_norm": 0.355400025844574,
+      "learning_rate": 0.00037867482517482513,
+      "loss": 3.2482,
+      "step": 63400
+    },
+    {
+      "epoch": 18.477022540625548,
+      "grad_norm": 0.3940849006175995,
+      "learning_rate": 0.0003785,
+      "loss": 3.239,
+      "step": 63450
+    },
+    {
+      "epoch": 18.49158366823927,
+      "grad_norm": 0.3690098226070404,
+      "learning_rate": 0.0003783251748251748,
+      "loss": 3.2422,
+      "step": 63500
+    },
+    {
+      "epoch": 18.50614479585299,
+      "grad_norm": 0.35771867632865906,
+      "learning_rate": 0.00037815034965034964,
+      "loss": 3.2373,
+      "step": 63550
+    },
+    {
+      "epoch": 18.520705923466714,
+      "grad_norm": 0.36510664224624634,
+      "learning_rate": 0.00037797552447552444,
+      "loss": 3.2422,
+      "step": 63600
+    },
+    {
+      "epoch": 18.535267051080435,
+      "grad_norm": 0.3902657628059387,
+      "learning_rate": 0.0003778006993006993,
+      "loss": 3.2532,
+      "step": 63650
+    },
+    {
+      "epoch": 18.54982817869416,
+      "grad_norm": 0.36882349848747253,
+      "learning_rate": 0.00037762587412587404,
+      "loss": 3.2641,
+      "step": 63700
+    },
+    {
+      "epoch": 18.56438930630788,
+      "grad_norm": 0.3985002934932709,
+      "learning_rate": 0.0003774510489510489,
+      "loss": 3.2609,
+      "step": 63750
+    },
+    {
+      "epoch": 18.5789504339216,
+      "grad_norm": 0.35080111026763916,
+      "learning_rate": 0.0003772762237762238,
+      "loss": 3.2476,
+      "step": 63800
+    },
+    {
+      "epoch": 18.593511561535326,
+      "grad_norm": 0.369186669588089,
+      "learning_rate": 0.00037710139860139854,
+      "loss": 3.2379,
+      "step": 63850
+    },
+    {
+      "epoch": 18.608072689149047,
+      "grad_norm": 0.36628544330596924,
+      "learning_rate": 0.0003769265734265734,
+      "loss": 3.2485,
+      "step": 63900
+    },
+    {
+      "epoch": 18.62263381676277,
+      "grad_norm": 0.38443052768707275,
+      "learning_rate": 0.0003767517482517482,
+      "loss": 3.2553,
+      "step": 63950
+    },
+    {
+      "epoch": 18.637194944376493,
+      "grad_norm": 0.37169235944747925,
+      "learning_rate": 0.00037657692307692305,
+      "loss": 3.2587,
+      "step": 64000
+    },
+    {
+      "epoch": 18.637194944376493,
+      "eval_accuracy": 0.3731584111460638,
+      "eval_loss": 3.541902780532837,
+      "eval_runtime": 180.7907,
+      "eval_samples_per_second": 92.073,
+      "eval_steps_per_second": 5.758,
+      "step": 64000
+    },
+    {
+      "epoch": 18.651756071990214,
+      "grad_norm": 0.3672039806842804,
+      "learning_rate": 0.00037640209790209785,
+      "loss": 3.2603,
+      "step": 64050
+    },
+    {
+      "epoch": 18.666317199603938,
+      "grad_norm": 0.45815590023994446,
+      "learning_rate": 0.0003762272727272727,
+      "loss": 3.2663,
+      "step": 64100
+    },
+    {
+      "epoch": 18.68087832721766,
+      "grad_norm": 0.3884202837944031,
+      "learning_rate": 0.0003760524475524475,
+      "loss": 3.2654,
+      "step": 64150
+    },
+    {
+      "epoch": 18.695439454831384,
+      "grad_norm": 0.3832554221153259,
+      "learning_rate": 0.00037587762237762236,
+      "loss": 3.2624,
+      "step": 64200
+    },
+    {
+      "epoch": 18.710000582445105,
+      "grad_norm": 0.36669617891311646,
+      "learning_rate": 0.00037570279720279716,
+      "loss": 3.2656,
+      "step": 64250
+    },
+    {
+      "epoch": 18.724561710058826,
+      "grad_norm": 0.38056427240371704,
+      "learning_rate": 0.000375527972027972,
+      "loss": 3.2587,
+      "step": 64300
+    },
+    {
+      "epoch": 18.73912283767255,
+      "grad_norm": 0.38066598773002625,
+      "learning_rate": 0.0003753531468531468,
+      "loss": 3.2681,
+      "step": 64350
+    },
+    {
+      "epoch": 18.75368396528627,
+      "grad_norm": 0.3981202244758606,
+      "learning_rate": 0.00037517832167832167,
+      "loss": 3.2619,
+      "step": 64400
+    },
+    {
+      "epoch": 18.768245092899996,
+      "grad_norm": 0.4338530898094177,
+      "learning_rate": 0.0003750034965034965,
+      "loss": 3.2692,
+      "step": 64450
+    },
+    {
+      "epoch": 18.782806220513717,
+      "grad_norm": 0.3539597988128662,
+      "learning_rate": 0.00037482867132867127,
+      "loss": 3.2738,
+      "step": 64500
+    },
+    {
+      "epoch": 18.797367348127437,
+      "grad_norm": 0.37372642755508423,
+      "learning_rate": 0.0003746538461538462,
+      "loss": 3.2636,
+      "step": 64550
+    },
+    {
+      "epoch": 18.811928475741162,
+      "grad_norm": 0.4146757423877716,
+      "learning_rate": 0.0003744790209790209,
+      "loss": 3.2714,
+      "step": 64600
+    },
+    {
+      "epoch": 18.826489603354883,
+      "grad_norm": 0.38235408067703247,
+      "learning_rate": 0.0003743041958041958,
+      "loss": 3.2688,
+      "step": 64650
+    },
+    {
+      "epoch": 18.841050730968607,
+      "grad_norm": 0.368381530046463,
+      "learning_rate": 0.0003741293706293706,
+      "loss": 3.2786,
+      "step": 64700
+    },
+    {
+      "epoch": 18.85561185858233,
+      "grad_norm": 0.37366122007369995,
+      "learning_rate": 0.0003739545454545454,
+      "loss": 3.2658,
+      "step": 64750
+    },
+    {
+      "epoch": 18.87017298619605,
+      "grad_norm": 0.39176511764526367,
+      "learning_rate": 0.0003737797202797202,
+      "loss": 3.2802,
+      "step": 64800
+    },
+    {
+      "epoch": 18.884734113809774,
+      "grad_norm": 0.36762627959251404,
+      "learning_rate": 0.0003736048951048951,
+      "loss": 3.2696,
+      "step": 64850
+    },
+    {
+      "epoch": 18.899295241423495,
+      "grad_norm": 0.36578455567359924,
+      "learning_rate": 0.0003734300699300699,
+      "loss": 3.2606,
+      "step": 64900
+    },
+    {
+      "epoch": 18.91385636903722,
+      "grad_norm": 0.37496402859687805,
+      "learning_rate": 0.00037325524475524473,
+      "loss": 3.2774,
+      "step": 64950
+    },
+    {
+      "epoch": 18.92841749665094,
+      "grad_norm": 0.4118943512439728,
+      "learning_rate": 0.00037308041958041953,
+      "loss": 3.2748,
+      "step": 65000
+    },
+    {
+      "epoch": 18.92841749665094,
+      "eval_accuracy": 0.37342116357355526,
+      "eval_loss": 3.5314598083496094,
+      "eval_runtime": 181.076,
+      "eval_samples_per_second": 91.928,
+      "eval_steps_per_second": 5.749,
+      "step": 65000
+    },
+    {
+      "epoch": 18.94297862426466,
+      "grad_norm": 0.3894543945789337,
+      "learning_rate": 0.0003729055944055944,
+      "loss": 3.2743,
+      "step": 65050
+    },
+    {
+      "epoch": 18.957539751878386,
+      "grad_norm": 0.3797741234302521,
+      "learning_rate": 0.0003727307692307692,
+      "loss": 3.2648,
+      "step": 65100
+    },
+    {
+      "epoch": 18.972100879492107,
+      "grad_norm": 0.40677353739738464,
+      "learning_rate": 0.00037255594405594404,
+      "loss": 3.273,
+      "step": 65150
+    },
+    {
+      "epoch": 18.98666200710583,
+      "grad_norm": 0.34352266788482666,
+      "learning_rate": 0.0003723811188811189,
+      "loss": 3.2839,
+      "step": 65200
+    },
+    {
+      "epoch": 19.001164890209097,
+      "grad_norm": 0.3892625570297241,
+      "learning_rate": 0.00037220629370629364,
+      "loss": 3.2697,
+      "step": 65250
+    },
+    {
+      "epoch": 19.01572601782282,
+      "grad_norm": 0.3817722499370575,
+      "learning_rate": 0.00037203146853146855,
+      "loss": 3.1613,
+      "step": 65300
+    },
+    {
+      "epoch": 19.030287145436542,
+      "grad_norm": 0.3764374256134033,
+      "learning_rate": 0.0003718566433566433,
+      "loss": 3.1725,
+      "step": 65350
+    },
+    {
+      "epoch": 19.044848273050263,
+      "grad_norm": 0.3804764747619629,
+      "learning_rate": 0.00037168181818181815,
+      "loss": 3.1711,
+      "step": 65400
+    },
+    {
+      "epoch": 19.059409400663988,
+      "grad_norm": 0.3937837481498718,
+      "learning_rate": 0.00037150699300699295,
+      "loss": 3.175,
+      "step": 65450
+    },
+    {
+      "epoch": 19.07397052827771,
+      "grad_norm": 0.3940994441509247,
+      "learning_rate": 0.0003713321678321678,
+      "loss": 3.1962,
+      "step": 65500
+    },
+    {
+      "epoch": 19.088531655891433,
+      "grad_norm": 0.3652641177177429,
+      "learning_rate": 0.0003711573426573426,
+      "loss": 3.1868,
+      "step": 65550
+    },
+    {
+      "epoch": 19.103092783505154,
+      "grad_norm": 0.34368589520454407,
+      "learning_rate": 0.00037098251748251746,
+      "loss": 3.192,
+      "step": 65600
+    },
+    {
+      "epoch": 19.11765391111888,
+      "grad_norm": 0.388906329870224,
+      "learning_rate": 0.00037080769230769226,
+      "loss": 3.1941,
+      "step": 65650
+    },
+    {
+      "epoch": 19.1322150387326,
+      "grad_norm": 0.40207478404045105,
+      "learning_rate": 0.0003706328671328671,
+      "loss": 3.2025,
+      "step": 65700
+    },
+    {
+      "epoch": 19.14677616634632,
+      "grad_norm": 0.3887316882610321,
+      "learning_rate": 0.0003704580419580419,
+      "loss": 3.2037,
+      "step": 65750
+    },
+    {
+      "epoch": 19.161337293960045,
+      "grad_norm": 0.38904979825019836,
+      "learning_rate": 0.00037028321678321676,
+      "loss": 3.1945,
+      "step": 65800
+    },
+    {
+      "epoch": 19.175898421573766,
+      "grad_norm": 0.3794829845428467,
+      "learning_rate": 0.0003701083916083916,
+      "loss": 3.2061,
+      "step": 65850
+    },
+    {
+      "epoch": 19.19045954918749,
+      "grad_norm": 0.4314369261264801,
+      "learning_rate": 0.0003699335664335664,
+      "loss": 3.2102,
+      "step": 65900
+    },
+    {
+      "epoch": 19.20502067680121,
+      "grad_norm": 0.40267378091812134,
+      "learning_rate": 0.00036975874125874127,
+      "loss": 3.2061,
+      "step": 65950
+    },
+    {
+      "epoch": 19.219581804414933,
+      "grad_norm": 0.3804206848144531,
+      "learning_rate": 0.00036958391608391607,
+      "loss": 3.1828,
+      "step": 66000
+    },
+    {
+      "epoch": 19.219581804414933,
+      "eval_accuracy": 0.3726852216513643,
+      "eval_loss": 3.5499932765960693,
+      "eval_runtime": 180.7468,
+      "eval_samples_per_second": 92.096,
+      "eval_steps_per_second": 5.759,
+      "step": 66000
+    },
+    {
+      "epoch": 19.234142932028657,
+      "grad_norm": 0.4092799425125122,
+      "learning_rate": 0.0003694090909090909,
+      "loss": 3.2022,
+      "step": 66050
+    },
+    {
+      "epoch": 19.248704059642378,
+      "grad_norm": 0.3658771216869354,
+      "learning_rate": 0.00036923426573426567,
+      "loss": 3.2156,
+      "step": 66100
+    },
+    {
+      "epoch": 19.263265187256103,
+      "grad_norm": 0.3707059621810913,
+      "learning_rate": 0.0003690594405594405,
+      "loss": 3.2101,
+      "step": 66150
+    },
+    {
+      "epoch": 19.277826314869824,
+      "grad_norm": 0.3689354360103607,
+      "learning_rate": 0.0003688846153846153,
+      "loss": 3.2297,
+      "step": 66200
+    },
+    {
+      "epoch": 19.292387442483545,
+      "grad_norm": 0.4154335558414459,
+      "learning_rate": 0.0003687097902097902,
+      "loss": 3.2106,
+      "step": 66250
+    },
+    {
+      "epoch": 19.30694857009727,
+      "grad_norm": 0.3923240303993225,
+      "learning_rate": 0.000368534965034965,
+      "loss": 3.2092,
+      "step": 66300
+    },
+    {
+      "epoch": 19.32150969771099,
+      "grad_norm": 0.3821892738342285,
+      "learning_rate": 0.00036836013986013983,
+      "loss": 3.2283,
+      "step": 66350
+    },
+    {
+      "epoch": 19.336070825324715,
+      "grad_norm": 0.4162690341472626,
+      "learning_rate": 0.00036818531468531463,
+      "loss": 3.2196,
+      "step": 66400
+    },
+    {
+      "epoch": 19.350631952938436,
+      "grad_norm": 0.3843232989311218,
+      "learning_rate": 0.0003680104895104895,
+      "loss": 3.2142,
+      "step": 66450
+    },
+    {
+      "epoch": 19.365193080552157,
+      "grad_norm": 0.37635594606399536,
+      "learning_rate": 0.0003678356643356643,
+      "loss": 3.2152,
+      "step": 66500
+    },
+    {
+      "epoch": 19.37975420816588,
+      "grad_norm": 0.3948022425174713,
+      "learning_rate": 0.00036766083916083914,
+      "loss": 3.2298,
+      "step": 66550
+    },
+    {
+      "epoch": 19.394315335779602,
+      "grad_norm": 0.39074137806892395,
+      "learning_rate": 0.000367486013986014,
+      "loss": 3.2166,
+      "step": 66600
+    },
+    {
+      "epoch": 19.408876463393327,
+      "grad_norm": 0.38576585054397583,
+      "learning_rate": 0.0003673111888111888,
+      "loss": 3.2184,
+      "step": 66650
+    },
+    {
+      "epoch": 19.423437591007048,
+      "grad_norm": 0.38614434003829956,
+      "learning_rate": 0.00036713636363636365,
+      "loss": 3.2252,
+      "step": 66700
+    },
+    {
+      "epoch": 19.43799871862077,
+      "grad_norm": 0.406095027923584,
+      "learning_rate": 0.00036696153846153844,
+      "loss": 3.2202,
+      "step": 66750
+    },
+    {
+      "epoch": 19.452559846234493,
+      "grad_norm": 0.37972956895828247,
+      "learning_rate": 0.0003667867132867133,
+      "loss": 3.2255,
+      "step": 66800
+    },
+    {
+      "epoch": 19.467120973848214,
+      "grad_norm": 0.38185039162635803,
+      "learning_rate": 0.00036661188811188804,
+      "loss": 3.2353,
+      "step": 66850
+    },
+    {
+      "epoch": 19.48168210146194,
+      "grad_norm": 0.39274418354034424,
+      "learning_rate": 0.0003664370629370629,
+      "loss": 3.2396,
+      "step": 66900
+    },
+    {
+      "epoch": 19.49624322907566,
+      "grad_norm": 0.39396706223487854,
+      "learning_rate": 0.0003662622377622377,
+      "loss": 3.2468,
+      "step": 66950
+    },
+    {
+      "epoch": 19.51080435668938,
+      "grad_norm": 0.37209567427635193,
+      "learning_rate": 0.00036608741258741255,
+      "loss": 3.2364,
+      "step": 67000
+    },
+    {
+      "epoch": 19.51080435668938,
+      "eval_accuracy": 0.3731644068390401,
+      "eval_loss": 3.5449118614196777,
+      "eval_runtime": 181.5019,
+      "eval_samples_per_second": 91.713,
+      "eval_steps_per_second": 5.735,
+      "step": 67000
+    },
+    {
+      "epoch": 19.525365484303105,
+      "grad_norm": 0.36750879883766174,
+      "learning_rate": 0.00036591258741258735,
+      "loss": 3.2415,
+      "step": 67050
+    },
+    {
+      "epoch": 19.539926611916826,
+      "grad_norm": 0.3621958792209625,
+      "learning_rate": 0.0003657377622377622,
+      "loss": 3.2354,
+      "step": 67100
+    },
+    {
+      "epoch": 19.55448773953055,
+      "grad_norm": 0.3803771734237671,
+      "learning_rate": 0.000365562937062937,
+      "loss": 3.241,
+      "step": 67150
+    },
+    {
+      "epoch": 19.56904886714427,
+      "grad_norm": 0.3699823319911957,
+      "learning_rate": 0.00036538811188811186,
+      "loss": 3.2311,
+      "step": 67200
+    },
+    {
+      "epoch": 19.583609994757992,
+      "grad_norm": 0.4162828028202057,
+      "learning_rate": 0.0003652132867132867,
+      "loss": 3.2361,
+      "step": 67250
+    },
+    {
+      "epoch": 19.598171122371717,
+      "grad_norm": 0.3639504909515381,
+      "learning_rate": 0.0003650384615384615,
+      "loss": 3.2508,
+      "step": 67300
+    },
+    {
+      "epoch": 19.612732249985438,
+      "grad_norm": 0.4054574966430664,
+      "learning_rate": 0.00036486363636363637,
+      "loss": 3.2472,
+      "step": 67350
+    },
+    {
+      "epoch": 19.627293377599162,
+      "grad_norm": 0.41714179515838623,
+      "learning_rate": 0.00036468881118881117,
+      "loss": 3.2421,
+      "step": 67400
+    },
+    {
+      "epoch": 19.641854505212883,
+      "grad_norm": 0.3846038281917572,
+      "learning_rate": 0.000364513986013986,
+      "loss": 3.2516,
+      "step": 67450
+    },
+    {
+      "epoch": 19.656415632826604,
+      "grad_norm": 0.35238131880760193,
+      "learning_rate": 0.0003643391608391608,
+      "loss": 3.2442,
+      "step": 67500
+    },
+    {
+      "epoch": 19.67097676044033,
+      "grad_norm": 0.36373382806777954,
+      "learning_rate": 0.0003641643356643357,
+      "loss": 3.2489,
+      "step": 67550
+    },
+    {
+      "epoch": 19.68553788805405,
+      "grad_norm": 0.3687552511692047,
+      "learning_rate": 0.0003639895104895104,
+      "loss": 3.2356,
+      "step": 67600
+    },
+    {
+      "epoch": 19.700099015667774,
+      "grad_norm": 0.3758412003517151,
+      "learning_rate": 0.0003638146853146853,
+      "loss": 3.2561,
+      "step": 67650
+    },
+    {
+      "epoch": 19.714660143281495,
+      "grad_norm": 0.37209826707839966,
+      "learning_rate": 0.00036363986013986007,
+      "loss": 3.2522,
+      "step": 67700
+    },
+    {
+      "epoch": 19.729221270895216,
+      "grad_norm": 0.38578248023986816,
+      "learning_rate": 0.0003634650349650349,
+      "loss": 3.2484,
+      "step": 67750
+    },
+    {
+      "epoch": 19.74378239850894,
+      "grad_norm": 0.3813081383705139,
+      "learning_rate": 0.0003632902097902097,
+      "loss": 3.253,
+      "step": 67800
+    },
+    {
+      "epoch": 19.758343526122662,
+      "grad_norm": 0.3634510040283203,
+      "learning_rate": 0.0003631153846153846,
+      "loss": 3.2554,
+      "step": 67850
+    },
+    {
+      "epoch": 19.772904653736386,
+      "grad_norm": 0.37720900774002075,
+      "learning_rate": 0.00036294055944055943,
+      "loss": 3.2575,
+      "step": 67900
+    },
+    {
+      "epoch": 19.787465781350107,
+      "grad_norm": 0.39338457584381104,
+      "learning_rate": 0.00036276573426573423,
+      "loss": 3.254,
+      "step": 67950
+    },
+    {
+      "epoch": 19.802026908963832,
+      "grad_norm": 0.40323713421821594,
+      "learning_rate": 0.0003625909090909091,
+      "loss": 3.271,
+      "step": 68000
+    },
+    {
+      "epoch": 19.802026908963832,
+      "eval_accuracy": 0.3733834259765867,
+      "eval_loss": 3.5359480381011963,
+      "eval_runtime": 181.6023,
+      "eval_samples_per_second": 91.662,
+      "eval_steps_per_second": 5.732,
+      "step": 68000
+    },
+    {
+      "epoch": 19.816588036577553,
+      "grad_norm": 0.3981825113296509,
+      "learning_rate": 0.0003624160839160839,
+      "loss": 3.2556,
+      "step": 68050
+    },
+    {
+      "epoch": 19.831149164191274,
+      "grad_norm": 0.3592379689216614,
+      "learning_rate": 0.00036224125874125874,
+      "loss": 3.2478,
+      "step": 68100
+    },
+    {
+      "epoch": 19.845710291805,
+      "grad_norm": 0.36781877279281616,
+      "learning_rate": 0.00036206643356643354,
+      "loss": 3.2517,
+      "step": 68150
+    },
+    {
+      "epoch": 19.86027141941872,
+      "grad_norm": 0.38581395149230957,
+      "learning_rate": 0.0003618916083916084,
+      "loss": 3.2555,
+      "step": 68200
+    },
+    {
+      "epoch": 19.874832547032444,
+      "grad_norm": 0.39512038230895996,
+      "learning_rate": 0.0003617167832167832,
+      "loss": 3.2689,
+      "step": 68250
+    },
+    {
+      "epoch": 19.889393674646165,
+      "grad_norm": 0.35403746366500854,
+      "learning_rate": 0.00036154195804195805,
+      "loss": 3.2671,
+      "step": 68300
+    },
+    {
+      "epoch": 19.903954802259886,
+      "grad_norm": 0.4128151535987854,
+      "learning_rate": 0.0003613671328671328,
+      "loss": 3.27,
+      "step": 68350
+    },
+    {
+      "epoch": 19.91851592987361,
+      "grad_norm": 0.4067288339138031,
+      "learning_rate": 0.00036119230769230765,
+      "loss": 3.2533,
+      "step": 68400
+    },
+    {
+      "epoch": 19.93307705748733,
+      "grad_norm": 0.36504778265953064,
+      "learning_rate": 0.00036101748251748245,
+      "loss": 3.2567,
+      "step": 68450
+    },
+    {
+      "epoch": 19.947638185101056,
+      "grad_norm": 0.3819653391838074,
+      "learning_rate": 0.0003608426573426573,
+      "loss": 3.265,
+      "step": 68500
+    },
+    {
+      "epoch": 19.962199312714777,
+      "grad_norm": 0.37546461820602417,
+      "learning_rate": 0.0003606678321678321,
+      "loss": 3.2596,
+      "step": 68550
+    },
+    {
+      "epoch": 19.976760440328498,
+      "grad_norm": 0.38692423701286316,
+      "learning_rate": 0.00036049300699300696,
+      "loss": 3.2577,
+      "step": 68600
+    },
+    {
+      "epoch": 19.991321567942222,
+      "grad_norm": 0.3578813672065735,
+      "learning_rate": 0.0003603181818181818,
+      "loss": 3.2568,
+      "step": 68650
+    },
+    {
+      "epoch": 20.005824451045488,
+      "grad_norm": 0.4185604751110077,
+      "learning_rate": 0.0003601433566433566,
+      "loss": 3.2322,
+      "step": 68700
+    },
+    {
+      "epoch": 20.020385578659212,
+      "grad_norm": 0.4104021191596985,
+      "learning_rate": 0.00035996853146853146,
+      "loss": 3.1446,
+      "step": 68750
+    },
+    {
+      "epoch": 20.034946706272933,
+      "grad_norm": 0.38897705078125,
+      "learning_rate": 0.00035979370629370626,
+      "loss": 3.1556,
+      "step": 68800
+    },
+    {
+      "epoch": 20.049507833886658,
+      "grad_norm": 0.40106043219566345,
+      "learning_rate": 0.0003596188811188811,
+      "loss": 3.1617,
+      "step": 68850
+    },
+    {
+      "epoch": 20.06406896150038,
+      "grad_norm": 0.4021047055721283,
+      "learning_rate": 0.0003594440559440559,
+      "loss": 3.1718,
+      "step": 68900
+    },
+    {
+      "epoch": 20.0786300891141,
+      "grad_norm": 0.3868863880634308,
+      "learning_rate": 0.00035926923076923077,
+      "loss": 3.1777,
+      "step": 68950
+    },
+    {
+      "epoch": 20.093191216727824,
+      "grad_norm": 0.368656188249588,
+      "learning_rate": 0.00035909440559440557,
+      "loss": 3.1782,
+      "step": 69000
+    },
+    {
+      "epoch": 20.093191216727824,
+      "eval_accuracy": 0.3730581302419697,
+      "eval_loss": 3.5493500232696533,
+      "eval_runtime": 181.0447,
+      "eval_samples_per_second": 91.944,
+      "eval_steps_per_second": 5.75,
+      "step": 69000
+    },
+    {
+      "epoch": 20.107752344341545,
+      "grad_norm": 0.37952375411987305,
+      "learning_rate": 0.0003589195804195804,
+      "loss": 3.1841,
+      "step": 69050
+    },
+    {
+      "epoch": 20.12231347195527,
+      "grad_norm": 0.3818630874156952,
+      "learning_rate": 0.00035874475524475517,
+      "loss": 3.1845,
+      "step": 69100
+    },
+    {
+      "epoch": 20.13687459956899,
+      "grad_norm": 0.4041067361831665,
+      "learning_rate": 0.00035856993006993,
+      "loss": 3.1926,
+      "step": 69150
+    },
+    {
+      "epoch": 20.15143572718271,
+      "grad_norm": 0.4008336365222931,
+      "learning_rate": 0.0003583951048951048,
+      "loss": 3.1876,
+      "step": 69200
+    },
+    {
+      "epoch": 20.165996854796436,
+      "grad_norm": 0.37791797518730164,
+      "learning_rate": 0.0003582202797202797,
+      "loss": 3.1929,
+      "step": 69250
+    },
+    {
+      "epoch": 20.180557982410157,
+      "grad_norm": 0.3881707191467285,
+      "learning_rate": 0.00035804545454545453,
+      "loss": 3.1894,
+      "step": 69300
+    },
+    {
+      "epoch": 20.19511911002388,
+      "grad_norm": 0.41837161779403687,
+      "learning_rate": 0.00035787062937062933,
+      "loss": 3.1876,
+      "step": 69350
+    },
+    {
+      "epoch": 20.209680237637603,
+      "grad_norm": 0.3870464563369751,
+      "learning_rate": 0.0003576958041958042,
+      "loss": 3.1971,
+      "step": 69400
+    },
+    {
+      "epoch": 20.224241365251324,
+      "grad_norm": 0.3730372190475464,
+      "learning_rate": 0.000357520979020979,
+      "loss": 3.2083,
+      "step": 69450
+    },
+    {
+      "epoch": 20.238802492865048,
+      "grad_norm": 0.3917742371559143,
+      "learning_rate": 0.00035734615384615384,
+      "loss": 3.1931,
+      "step": 69500
+    },
+    {
+      "epoch": 20.25336362047877,
+      "grad_norm": 0.39877134561538696,
+      "learning_rate": 0.00035717132867132864,
+      "loss": 3.201,
+      "step": 69550
+    },
+    {
+      "epoch": 20.267924748092494,
+      "grad_norm": 0.4186917543411255,
+      "learning_rate": 0.0003569965034965035,
+      "loss": 3.2039,
+      "step": 69600
+    },
+    {
+      "epoch": 20.282485875706215,
+      "grad_norm": 0.3791273832321167,
+      "learning_rate": 0.0003568216783216783,
+      "loss": 3.2102,
+      "step": 69650
+    },
+    {
+      "epoch": 20.297047003319935,
+      "grad_norm": 0.3736315667629242,
+      "learning_rate": 0.00035664685314685314,
+      "loss": 3.1941,
+      "step": 69700
+    },
+    {
+      "epoch": 20.31160813093366,
+      "grad_norm": 0.40645095705986023,
+      "learning_rate": 0.00035647202797202794,
+      "loss": 3.2125,
+      "step": 69750
+    },
+    {
+      "epoch": 20.32616925854738,
+      "grad_norm": 0.4175493121147156,
+      "learning_rate": 0.0003562972027972028,
+      "loss": 3.2107,
+      "step": 69800
+    },
+    {
+      "epoch": 20.340730386161106,
+      "grad_norm": 0.392880916595459,
+      "learning_rate": 0.00035612237762237754,
+      "loss": 3.2139,
+      "step": 69850
+    },
+    {
+      "epoch": 20.355291513774826,
+      "grad_norm": 0.3980697691440582,
+      "learning_rate": 0.0003559475524475524,
+      "loss": 3.2024,
+      "step": 69900
+    },
+    {
+      "epoch": 20.369852641388547,
+      "grad_norm": 0.37479138374328613,
+      "learning_rate": 0.0003557727272727272,
+      "loss": 3.2134,
+      "step": 69950
+    },
+    {
+      "epoch": 20.384413769002272,
+      "grad_norm": 0.3928723931312561,
+      "learning_rate": 0.00035559790209790205,
+      "loss": 3.227,
+      "step": 70000
+    },
+    {
+      "epoch": 20.384413769002272,
+      "eval_accuracy": 0.37295220633272147,
+      "eval_loss": 3.5466055870056152,
+      "eval_runtime": 181.2725,
+      "eval_samples_per_second": 91.829,
+      "eval_steps_per_second": 5.743,
+      "step": 70000
+    }
+  ],
+  "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": 5
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4631308623872e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}