diff --git "a/last_to_push_frequency_5039/checkpoint-100000/trainer_state.json" "b/last_to_push_frequency_5039/checkpoint-100000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/last_to_push_frequency_5039/checkpoint-100000/trainer_state.json"
@@ -0,0 +1,14943 @@
+{
+  "best_global_step": 89000,
+  "best_metric": 3.530644655227661,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/last_to_push_frequency_5039/checkpoint-40000",
+  "epoch": 29.120857359193895,
+  "eval_steps": 1000,
+  "global_step": 100000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014561127613722406,
+      "grad_norm": 0.7142499685287476,
+      "learning_rate": 0.000294,
+      "loss": 8.4587,
+      "step": 50
+    },
+    {
+      "epoch": 0.029122255227444813,
+      "grad_norm": 0.5678296685218811,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7387,
+      "step": 100
+    },
+    {
+      "epoch": 0.04368338284116722,
+      "grad_norm": 0.5606561303138733,
+      "learning_rate": 0.0005998286713286713,
+      "loss": 6.3655,
+      "step": 150
+    },
+    {
+      "epoch": 0.058244510454889625,
+      "grad_norm": 0.47408413887023926,
+      "learning_rate": 0.0005996538461538461,
+      "loss": 6.1262,
+      "step": 200
+    },
+    {
+      "epoch": 0.07280563806861204,
+      "grad_norm": 0.4732152223587036,
+      "learning_rate": 0.0005994790209790209,
+      "loss": 5.9972,
+      "step": 250
+    },
+    {
+      "epoch": 0.08736676568233444,
+      "grad_norm": 0.438856303691864,
+      "learning_rate": 0.0005993041958041958,
+      "loss": 5.8873,
+      "step": 300
+    },
+    {
+      "epoch": 0.10192789329605685,
+      "grad_norm": 0.4995385408401489,
+      "learning_rate": 0.0005991293706293705,
+      "loss": 5.7343,
+      "step": 350
+    },
+    {
+      "epoch": 0.11648902090977925,
+      "grad_norm": 0.4228087067604065,
+      "learning_rate": 0.0005989545454545454,
+      "loss": 5.63,
+      "step": 400
+    },
+    {
+      "epoch": 0.13105014852350166,
+      "grad_norm": 0.4598565399646759,
+      "learning_rate": 0.0005987797202797202,
+      "loss": 5.5141,
+      "step": 450
+    },
+    {
+      "epoch": 0.14561127613722408,
+      "grad_norm": 0.47586789727211,
+      "learning_rate": 0.000598604895104895,
+      "loss": 5.414,
+      "step": 500
+    },
+    {
+      "epoch": 0.16017240375094646,
+      "grad_norm": 0.477698415517807,
+      "learning_rate": 0.0005984300699300698,
+      "loss": 5.3322,
+      "step": 550
+    },
+    {
+      "epoch": 0.17473353136466888,
+      "grad_norm": 0.44980600476264954,
+      "learning_rate": 0.0005982552447552447,
+      "loss": 5.2471,
+      "step": 600
+    },
+    {
+      "epoch": 0.1892946589783913,
+      "grad_norm": 0.4625977873802185,
+      "learning_rate": 0.0005980804195804195,
+      "loss": 5.1807,
+      "step": 650
+    },
+    {
+      "epoch": 0.2038557865921137,
+      "grad_norm": 0.5372968912124634,
+      "learning_rate": 0.0005979055944055943,
+      "loss": 5.1209,
+      "step": 700
+    },
+    {
+      "epoch": 0.2184169142058361,
+      "grad_norm": 0.5401430726051331,
+      "learning_rate": 0.0005977307692307691,
+      "loss": 5.0708,
+      "step": 750
+    },
+    {
+      "epoch": 0.2329780418195585,
+      "grad_norm": 0.4346584379673004,
+      "learning_rate": 0.000597555944055944,
+      "loss": 5.0076,
+      "step": 800
+    },
+    {
+      "epoch": 0.24753916943328091,
+      "grad_norm": 0.460715115070343,
+      "learning_rate": 0.0005973811188811188,
+      "loss": 4.9693,
+      "step": 850
+    },
+    {
+      "epoch": 0.2621002970470033,
+      "grad_norm": 0.4268857538700104,
+      "learning_rate": 0.0005972062937062936,
+      "loss": 4.9309,
+      "step": 900
+    },
+    {
+      "epoch": 0.27666142466072574,
+      "grad_norm": 0.44858410954475403,
+      "learning_rate": 0.0005970314685314685,
+      "loss": 4.8723,
+      "step": 950
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "grad_norm": 0.4247923493385315,
+      "learning_rate": 0.0005968566433566433,
+      "loss": 4.8399,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "eval_accuracy": 0.2545978147932791,
+      "eval_loss": 4.757110118865967,
+      "eval_runtime": 177.2928,
+      "eval_samples_per_second": 93.89,
+      "eval_steps_per_second": 5.872,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30578367988817057,
+      "grad_norm": 0.47108718752861023,
+      "learning_rate": 0.0005966818181818181,
+      "loss": 4.7695,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3203448075018929,
+      "grad_norm": 0.4697106182575226,
+      "learning_rate": 0.0005965069930069929,
+      "loss": 4.7315,
+      "step": 1100
+    },
+    {
+      "epoch": 0.33490593511561534,
+      "grad_norm": 0.4338913559913635,
+      "learning_rate": 0.0005963321678321677,
+      "loss": 4.7008,
+      "step": 1150
+    },
+    {
+      "epoch": 0.34946706272933775,
+      "grad_norm": 0.47174957394599915,
+      "learning_rate": 0.0005961573426573425,
+      "loss": 4.6608,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36402819034306017,
+      "grad_norm": 0.4433532953262329,
+      "learning_rate": 0.0005959825174825174,
+      "loss": 4.632,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3785893179567826,
+      "grad_norm": 0.5037450194358826,
+      "learning_rate": 0.0005958076923076922,
+      "loss": 4.5994,
+      "step": 1300
+    },
+    {
+      "epoch": 0.393150445570505,
+      "grad_norm": 0.4283815622329712,
+      "learning_rate": 0.000595632867132867,
+      "loss": 4.567,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4077115731842274,
+      "grad_norm": 0.39518365263938904,
+      "learning_rate": 0.0005954580419580418,
+      "loss": 4.556,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4222727007979498,
+      "grad_norm": 0.435833215713501,
+      "learning_rate": 0.0005952832167832168,
+      "loss": 4.5333,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4368338284116722,
+      "grad_norm": 0.4726100265979767,
+      "learning_rate": 0.0005951083916083916,
+      "loss": 4.4928,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4513949560253946,
+      "grad_norm": 0.44213613867759705,
+      "learning_rate": 0.0005949335664335664,
+      "loss": 4.4892,
+      "step": 1550
+    },
+    {
+      "epoch": 0.465956083639117,
+      "grad_norm": 0.39726656675338745,
+      "learning_rate": 0.0005947587412587413,
+      "loss": 4.4738,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4805172112528394,
+      "grad_norm": 0.4227396547794342,
+      "learning_rate": 0.0005945839160839161,
+      "loss": 4.4391,
+      "step": 1650
+    },
+    {
+      "epoch": 0.49507833886656183,
+      "grad_norm": 0.4156210124492645,
+      "learning_rate": 0.0005944090909090909,
+      "loss": 4.4382,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5096394664802842,
+      "grad_norm": 0.4048631489276886,
+      "learning_rate": 0.0005942342657342657,
+      "loss": 4.4035,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5242005940940067,
+      "grad_norm": 0.3983406722545624,
+      "learning_rate": 0.0005940594405594406,
+      "loss": 4.3835,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5387617217077291,
+      "grad_norm": 0.4011361002922058,
+      "learning_rate": 0.0005938846153846153,
+      "loss": 4.3834,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5533228493214515,
+      "grad_norm": 0.4169325828552246,
+      "learning_rate": 0.0005937097902097902,
+      "loss": 4.3576,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5678839769351739,
+      "grad_norm": 0.4109379053115845,
+      "learning_rate": 0.000593534965034965,
+      "loss": 4.3456,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "grad_norm": 0.4288181960582733,
+      "learning_rate": 0.0005933601398601398,
+      "loss": 4.3272,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "eval_accuracy": 0.3003302568766484,
+      "eval_loss": 4.275340557098389,
+      "eval_runtime": 177.3788,
+      "eval_samples_per_second": 93.844,
+      "eval_steps_per_second": 5.869,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5970062321626187,
+      "grad_norm": 0.3974442183971405,
+      "learning_rate": 0.0005931853146853146,
+      "loss": 4.3343,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6115673597763411,
+      "grad_norm": 0.39755979180336,
+      "learning_rate": 0.0005930104895104895,
+      "loss": 4.3154,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6261284873900634,
+      "grad_norm": 0.3861689865589142,
+      "learning_rate": 0.0005928356643356643,
+      "loss": 4.2906,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6406896150037859,
+      "grad_norm": 0.3685368597507477,
+      "learning_rate": 0.0005926608391608391,
+      "loss": 4.2912,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6552507426175083,
+      "grad_norm": 0.4314819574356079,
+      "learning_rate": 0.000592486013986014,
+      "loss": 4.2757,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6698118702312307,
+      "grad_norm": 0.38772013783454895,
+      "learning_rate": 0.0005923111888111888,
+      "loss": 4.2653,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6843729978449531,
+      "grad_norm": 0.36975449323654175,
+      "learning_rate": 0.0005921363636363636,
+      "loss": 4.2463,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6989341254586755,
+      "grad_norm": 0.374622642993927,
+      "learning_rate": 0.0005919615384615384,
+      "loss": 4.2391,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7134952530723979,
+      "grad_norm": 0.3807830214500427,
+      "learning_rate": 0.0005917867132867133,
+      "loss": 4.2362,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7280563806861203,
+      "grad_norm": 0.35665392875671387,
+      "learning_rate": 0.0005916118881118881,
+      "loss": 4.2275,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7426175082998427,
+      "grad_norm": 0.39687132835388184,
+      "learning_rate": 0.0005914370629370629,
+      "loss": 4.2086,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7571786359135652,
+      "grad_norm": 0.3794260025024414,
+      "learning_rate": 0.0005912622377622377,
+      "loss": 4.2127,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7717397635272876,
+      "grad_norm": 0.40508922934532166,
+      "learning_rate": 0.0005910874125874125,
+      "loss": 4.2015,
+      "step": 2650
+    },
+    {
+      "epoch": 0.78630089114101,
+      "grad_norm": 0.3626640737056732,
+      "learning_rate": 0.0005909125874125873,
+      "loss": 4.187,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8008620187547324,
+      "grad_norm": 0.36335116624832153,
+      "learning_rate": 0.0005907377622377622,
+      "loss": 4.173,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8154231463684548,
+      "grad_norm": 0.38054153323173523,
+      "learning_rate": 0.000590562937062937,
+      "loss": 4.1755,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8299842739821772,
+      "grad_norm": 0.3813888132572174,
+      "learning_rate": 0.0005903881118881118,
+      "loss": 4.171,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8445454015958996,
+      "grad_norm": 0.393034428358078,
+      "learning_rate": 0.0005902132867132867,
+      "loss": 4.154,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8591065292096219,
+      "grad_norm": 0.36573460698127747,
+      "learning_rate": 0.0005900384615384615,
+      "loss": 4.1445,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "grad_norm": 0.36031314730644226,
+      "learning_rate": 0.0005898636363636363,
+      "loss": 4.1337,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "eval_accuracy": 0.3159021295996076,
+      "eval_loss": 4.092033386230469,
+      "eval_runtime": 177.2002,
+      "eval_samples_per_second": 93.939,
+      "eval_steps_per_second": 5.875,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8882287844370668,
+      "grad_norm": 0.33011677861213684,
+      "learning_rate": 0.0005896888111888111,
+      "loss": 4.1267,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9027899120507892,
+      "grad_norm": 0.3441181480884552,
+      "learning_rate": 0.000589513986013986,
+      "loss": 4.1295,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9173510396645116,
+      "grad_norm": 0.3503768742084503,
+      "learning_rate": 0.0005893391608391608,
+      "loss": 4.1174,
+      "step": 3150
+    },
+    {
+      "epoch": 0.931912167278234,
+      "grad_norm": 0.36162590980529785,
+      "learning_rate": 0.0005891643356643356,
+      "loss": 4.1218,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9464732948919564,
+      "grad_norm": 0.3388279974460602,
+      "learning_rate": 0.0005889895104895104,
+      "loss": 4.1123,
+      "step": 3250
+    },
+    {
+      "epoch": 0.9610344225056788,
+      "grad_norm": 0.36532530188560486,
+      "learning_rate": 0.0005888146853146853,
+      "loss": 4.1076,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9755955501194012,
+      "grad_norm": 0.42103567719459534,
+      "learning_rate": 0.00058863986013986,
+      "loss": 4.0885,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9901566777331237,
+      "grad_norm": 0.354731947183609,
+      "learning_rate": 0.0005884650349650349,
+      "loss": 4.0847,
+      "step": 3400
+    },
+    {
+      "epoch": 1.004659560836391,
+      "grad_norm": 0.34602439403533936,
+      "learning_rate": 0.0005882902097902097,
+      "loss": 4.0677,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0192206884501136,
+      "grad_norm": 0.3426474332809448,
+      "learning_rate": 0.0005881153846153845,
+      "loss": 4.0154,
+      "step": 3500
+    },
+    {
+      "epoch": 1.033781816063836,
+      "grad_norm": 0.34196737408638,
+      "learning_rate": 0.0005879405594405594,
+      "loss": 4.0117,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0483429436775584,
+      "grad_norm": 0.35845261812210083,
+      "learning_rate": 0.0005877657342657342,
+      "loss": 4.0123,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0629040712912807,
+      "grad_norm": 0.3470519185066223,
+      "learning_rate": 0.000587590909090909,
+      "loss": 4.0003,
+      "step": 3650
+    },
+    {
+      "epoch": 1.0774651989050033,
+      "grad_norm": 0.3520037531852722,
+      "learning_rate": 0.0005874160839160838,
+      "loss": 3.989,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0920263265187256,
+      "grad_norm": 0.3449414074420929,
+      "learning_rate": 0.0005872412587412587,
+      "loss": 4.0093,
+      "step": 3750
+    },
+    {
+      "epoch": 1.106587454132448,
+      "grad_norm": 0.3322015106678009,
+      "learning_rate": 0.0005870664335664335,
+      "loss": 4.0085,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1211485817461704,
+      "grad_norm": 0.35296934843063354,
+      "learning_rate": 0.0005868916083916083,
+      "loss": 3.9958,
+      "step": 3850
+    },
+    {
+      "epoch": 1.135709709359893,
+      "grad_norm": 0.3486291766166687,
+      "learning_rate": 0.0005867167832167831,
+      "loss": 4.0075,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1502708369736152,
+      "grad_norm": 0.34043097496032715,
+      "learning_rate": 0.000586541958041958,
+      "loss": 3.9973,
+      "step": 3950
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "grad_norm": 0.33894768357276917,
+      "learning_rate": 0.0005863671328671328,
+      "loss": 3.9915,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "eval_accuracy": 0.32561280096909206,
+      "eval_loss": 3.9853665828704834,
+      "eval_runtime": 177.196,
+      "eval_samples_per_second": 93.941,
+      "eval_steps_per_second": 5.875,
+      "step": 4000
+    },
+    {
+      "epoch": 1.17939309220106,
+      "grad_norm": 0.3345903754234314,
+      "learning_rate": 0.0005861923076923076,
+      "loss": 3.9927,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1939542198147826,
+      "grad_norm": 0.3380124270915985,
+      "learning_rate": 0.0005860174825174824,
+      "loss": 3.9764,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2085153474285049,
+      "grad_norm": 0.3687117099761963,
+      "learning_rate": 0.0005858426573426573,
+      "loss": 3.985,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2230764750422272,
+      "grad_norm": 0.3383759558200836,
+      "learning_rate": 0.000585667832167832,
+      "loss": 3.967,
+      "step": 4200
+    },
+    {
+      "epoch": 1.2376376026559497,
+      "grad_norm": 0.3614034354686737,
+      "learning_rate": 0.000585493006993007,
+      "loss": 3.9671,
+      "step": 4250
+    },
+    {
+      "epoch": 1.2521987302696722,
+      "grad_norm": 0.3286982476711273,
+      "learning_rate": 0.0005853181818181817,
+      "loss": 3.9622,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2667598578833945,
+      "grad_norm": 0.34898343682289124,
+      "learning_rate": 0.0005851433566433565,
+      "loss": 3.9527,
+      "step": 4350
+    },
+    {
+      "epoch": 1.2813209854971168,
+      "grad_norm": 0.3299115300178528,
+      "learning_rate": 0.0005849685314685315,
+      "loss": 3.9661,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2958821131108393,
+      "grad_norm": 0.3476330637931824,
+      "learning_rate": 0.0005847937062937063,
+      "loss": 3.9757,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3104432407245616,
+      "grad_norm": 0.3423220217227936,
+      "learning_rate": 0.0005846188811188811,
+      "loss": 3.9664,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3250043683382842,
+      "grad_norm": 0.36620113253593445,
+      "learning_rate": 0.0005844440559440559,
+      "loss": 3.9626,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3395654959520065,
+      "grad_norm": 0.34613823890686035,
+      "learning_rate": 0.0005842692307692308,
+      "loss": 3.9469,
+      "step": 4600
+    },
+    {
+      "epoch": 1.354126623565729,
+      "grad_norm": 0.3371380865573883,
+      "learning_rate": 0.0005840944055944056,
+      "loss": 3.9431,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3686877511794513,
+      "grad_norm": 0.3557433485984802,
+      "learning_rate": 0.0005839195804195804,
+      "loss": 3.9371,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3832488787931738,
+      "grad_norm": 0.34661224484443665,
+      "learning_rate": 0.0005837447552447552,
+      "loss": 3.9475,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3978100064068961,
+      "grad_norm": 0.3363933563232422,
+      "learning_rate": 0.0005835699300699301,
+      "loss": 3.9525,
+      "step": 4800
+    },
+    {
+      "epoch": 1.4123711340206184,
+      "grad_norm": 0.3473567068576813,
+      "learning_rate": 0.0005833951048951048,
+      "loss": 3.9478,
+      "step": 4850
+    },
+    {
+      "epoch": 1.426932261634341,
+      "grad_norm": 0.35772213339805603,
+      "learning_rate": 0.0005832202797202797,
+      "loss": 3.9494,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4414933892480635,
+      "grad_norm": 0.3194500803947449,
+      "learning_rate": 0.0005830454545454546,
+      "loss": 3.9274,
+      "step": 4950
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "grad_norm": 0.3184325397014618,
+      "learning_rate": 0.0005828706293706293,
+      "loss": 3.91,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "eval_accuracy": 0.3322197019411703,
+      "eval_loss": 3.9121792316436768,
+      "eval_runtime": 177.1541,
+      "eval_samples_per_second": 93.963,
+      "eval_steps_per_second": 5.876,
+      "step": 5000
+    },
+    {
+      "epoch": 1.470615644475508,
+      "grad_norm": 0.334377646446228,
+      "learning_rate": 0.0005826958041958042,
+      "loss": 3.9292,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4851767720892306,
+      "grad_norm": 0.3505079746246338,
+      "learning_rate": 0.000582520979020979,
+      "loss": 3.9197,
+      "step": 5100
+    },
+    {
+      "epoch": 1.4997378997029531,
+      "grad_norm": 0.3423149585723877,
+      "learning_rate": 0.0005823461538461538,
+      "loss": 3.9216,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5142990273166754,
+      "grad_norm": 0.3242221772670746,
+      "learning_rate": 0.0005821713286713286,
+      "loss": 3.9264,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5288601549303977,
+      "grad_norm": 0.3360047936439514,
+      "learning_rate": 0.0005819965034965035,
+      "loss": 3.9075,
+      "step": 5250
+    },
+    {
+      "epoch": 1.5434212825441203,
+      "grad_norm": 0.3281499445438385,
+      "learning_rate": 0.0005818216783216783,
+      "loss": 3.9156,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5579824101578428,
+      "grad_norm": 0.3143256604671478,
+      "learning_rate": 0.0005816468531468531,
+      "loss": 3.9054,
+      "step": 5350
+    },
+    {
+      "epoch": 1.572543537771565,
+      "grad_norm": 0.34801602363586426,
+      "learning_rate": 0.0005814720279720279,
+      "loss": 3.9052,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5871046653852874,
+      "grad_norm": 0.33646127581596375,
+      "learning_rate": 0.0005812972027972028,
+      "loss": 3.9049,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6016657929990097,
+      "grad_norm": 0.34947100281715393,
+      "learning_rate": 0.0005811223776223776,
+      "loss": 3.9141,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6162269206127322,
+      "grad_norm": 0.3342869281768799,
+      "learning_rate": 0.0005809475524475524,
+      "loss": 3.8802,
+      "step": 5550
+    },
+    {
+      "epoch": 1.6307880482264547,
+      "grad_norm": 0.3248964250087738,
+      "learning_rate": 0.0005807727272727272,
+      "loss": 3.9011,
+      "step": 5600
+    },
+    {
+      "epoch": 1.645349175840177,
+      "grad_norm": 0.3466763496398926,
+      "learning_rate": 0.0005805979020979021,
+      "loss": 3.8889,
+      "step": 5650
+    },
+    {
+      "epoch": 1.6599103034538993,
+      "grad_norm": 0.34471645951271057,
+      "learning_rate": 0.0005804230769230769,
+      "loss": 3.8915,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6744714310676219,
+      "grad_norm": 0.3128034770488739,
+      "learning_rate": 0.0005802482517482517,
+      "loss": 3.8916,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6890325586813444,
+      "grad_norm": 0.3492928445339203,
+      "learning_rate": 0.0005800734265734265,
+      "loss": 3.8693,
+      "step": 5800
+    },
+    {
+      "epoch": 1.7035936862950667,
+      "grad_norm": 0.3331936001777649,
+      "learning_rate": 0.0005798986013986013,
+      "loss": 3.8772,
+      "step": 5850
+    },
+    {
+      "epoch": 1.718154813908789,
+      "grad_norm": 0.3292458951473236,
+      "learning_rate": 0.0005797237762237762,
+      "loss": 3.874,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7327159415225115,
+      "grad_norm": 0.3334185779094696,
+      "learning_rate": 0.000579548951048951,
+      "loss": 3.878,
+      "step": 5950
+    },
+    {
+      "epoch": 1.747277069136234,
+      "grad_norm": 0.3228028118610382,
+      "learning_rate": 0.0005793741258741258,
+      "loss": 3.8822,
+      "step": 6000
+    },
+    {
+      "epoch": 1.747277069136234,
+      "eval_accuracy": 0.33703835809241034,
+      "eval_loss": 3.8543479442596436,
+      "eval_runtime": 177.1903,
+      "eval_samples_per_second": 93.944,
+      "eval_steps_per_second": 5.875,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7618381967499563,
+      "grad_norm": 0.35547763109207153,
+      "learning_rate": 0.0005791993006993006,
+      "loss": 3.8844,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7763993243636786,
+      "grad_norm": 0.33292338252067566,
+      "learning_rate": 0.0005790244755244755,
+      "loss": 3.8799,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7909604519774012,
+      "grad_norm": 0.3139176666736603,
+      "learning_rate": 0.0005788496503496503,
+      "loss": 3.869,
+      "step": 6150
+    },
+    {
+      "epoch": 1.8055215795911237,
+      "grad_norm": 0.3225059509277344,
+      "learning_rate": 0.0005786748251748251,
+      "loss": 3.8519,
+      "step": 6200
+    },
+    {
+      "epoch": 1.820082707204846,
+      "grad_norm": 0.3270617127418518,
+      "learning_rate": 0.0005784999999999999,
+      "loss": 3.8537,
+      "step": 6250
+    },
+    {
+      "epoch": 1.8346438348185683,
+      "grad_norm": 0.32356396317481995,
+      "learning_rate": 0.0005783251748251748,
+      "loss": 3.8637,
+      "step": 6300
+    },
+    {
+      "epoch": 1.8492049624322906,
+      "grad_norm": 0.34862807393074036,
+      "learning_rate": 0.0005781503496503496,
+      "loss": 3.8608,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8637660900460131,
+      "grad_norm": 0.3327373266220093,
+      "learning_rate": 0.0005779755244755244,
+      "loss": 3.8592,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8783272176597356,
+      "grad_norm": 0.323494553565979,
+      "learning_rate": 0.0005778006993006993,
+      "loss": 3.859,
+      "step": 6450
+    },
+    {
+      "epoch": 1.892888345273458,
+      "grad_norm": 0.3193258047103882,
+      "learning_rate": 0.000577625874125874,
+      "loss": 3.8555,
+      "step": 6500
+    },
+    {
+      "epoch": 1.9074494728871803,
+      "grad_norm": 0.34051451086997986,
+      "learning_rate": 0.0005774510489510489,
+      "loss": 3.8548,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9220106005009028,
+      "grad_norm": 0.34305649995803833,
+      "learning_rate": 0.0005772762237762237,
+      "loss": 3.8472,
+      "step": 6600
+    },
+    {
+      "epoch": 1.9365717281146253,
+      "grad_norm": 0.3282603323459625,
+      "learning_rate": 0.0005771013986013985,
+      "loss": 3.8501,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9511328557283476,
+      "grad_norm": 0.31083357334136963,
+      "learning_rate": 0.0005769265734265733,
+      "loss": 3.8399,
+      "step": 6700
+    },
+    {
+      "epoch": 1.96569398334207,
+      "grad_norm": 0.3474798798561096,
+      "learning_rate": 0.0005767517482517482,
+      "loss": 3.8562,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9802551109557924,
+      "grad_norm": 0.3274882435798645,
+      "learning_rate": 0.000576576923076923,
+      "loss": 3.8483,
+      "step": 6800
+    },
+    {
+      "epoch": 1.994816238569515,
+      "grad_norm": 0.33568403124809265,
+      "learning_rate": 0.0005764020979020978,
+      "loss": 3.841,
+      "step": 6850
+    },
+    {
+      "epoch": 2.009319121672782,
+      "grad_norm": 0.3249611556529999,
+      "learning_rate": 0.0005762272727272726,
+      "loss": 3.7798,
+      "step": 6900
+    },
+    {
+      "epoch": 2.023880249286505,
+      "grad_norm": 0.3431794047355652,
+      "learning_rate": 0.0005760524475524475,
+      "loss": 3.7367,
+      "step": 6950
+    },
+    {
+      "epoch": 2.038441376900227,
+      "grad_norm": 0.322765976190567,
+      "learning_rate": 0.0005758776223776223,
+      "loss": 3.7549,
+      "step": 7000
+    },
+    {
+      "epoch": 2.038441376900227,
+      "eval_accuracy": 0.3418397325403657,
+      "eval_loss": 3.810866355895996,
+      "eval_runtime": 177.2789,
+      "eval_samples_per_second": 93.897,
+      "eval_steps_per_second": 5.872,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0530025045139495,
+      "grad_norm": 0.3592188060283661,
+      "learning_rate": 0.0005757027972027971,
+      "loss": 3.7467,
+      "step": 7050
+    },
+    {
+      "epoch": 2.067563632127672,
+      "grad_norm": 0.3323240876197815,
+      "learning_rate": 0.000575527972027972,
+      "loss": 3.7268,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0821247597413945,
+      "grad_norm": 0.3217570185661316,
+      "learning_rate": 0.0005753531468531468,
+      "loss": 3.7491,
+      "step": 7150
+    },
+    {
+      "epoch": 2.096685887355117,
+      "grad_norm": 0.33970892429351807,
+      "learning_rate": 0.0005751783216783216,
+      "loss": 3.7353,
+      "step": 7200
+    },
+    {
+      "epoch": 2.111247014968839,
+      "grad_norm": 0.33313965797424316,
+      "learning_rate": 0.0005750034965034964,
+      "loss": 3.7521,
+      "step": 7250
+    },
+    {
+      "epoch": 2.1258081425825615,
+      "grad_norm": 0.31688767671585083,
+      "learning_rate": 0.0005748286713286712,
+      "loss": 3.7518,
+      "step": 7300
+    },
+    {
+      "epoch": 2.140369270196284,
+      "grad_norm": 0.32191213965415955,
+      "learning_rate": 0.000574653846153846,
+      "loss": 3.7579,
+      "step": 7350
+    },
+    {
+      "epoch": 2.1549303978100065,
+      "grad_norm": 0.3361343741416931,
+      "learning_rate": 0.000574479020979021,
+      "loss": 3.7556,
+      "step": 7400
+    },
+    {
+      "epoch": 2.169491525423729,
+      "grad_norm": 0.31816932559013367,
+      "learning_rate": 0.0005743041958041958,
+      "loss": 3.7599,
+      "step": 7450
+    },
+    {
+      "epoch": 2.184052653037451,
+      "grad_norm": 0.32364562153816223,
+      "learning_rate": 0.0005741293706293706,
+      "loss": 3.7605,
+      "step": 7500
+    },
+    {
+      "epoch": 2.198613780651174,
+      "grad_norm": 0.32209041714668274,
+      "learning_rate": 0.0005739545454545454,
+      "loss": 3.7569,
+      "step": 7550
+    },
+    {
+      "epoch": 2.213174908264896,
+      "grad_norm": 0.33622536063194275,
+      "learning_rate": 0.0005737797202797203,
+      "loss": 3.7571,
+      "step": 7600
+    },
+    {
+      "epoch": 2.2277360358786185,
+      "grad_norm": 0.34482723474502563,
+      "learning_rate": 0.0005736048951048951,
+      "loss": 3.7517,
+      "step": 7650
+    },
+    {
+      "epoch": 2.2422971634923408,
+      "grad_norm": 0.322247177362442,
+      "learning_rate": 0.0005734300699300699,
+      "loss": 3.7608,
+      "step": 7700
+    },
+    {
+      "epoch": 2.256858291106063,
+      "grad_norm": 0.3295249044895172,
+      "learning_rate": 0.0005732552447552448,
+      "loss": 3.7445,
+      "step": 7750
+    },
+    {
+      "epoch": 2.271419418719786,
+      "grad_norm": 0.3132839798927307,
+      "learning_rate": 0.0005730804195804196,
+      "loss": 3.7522,
+      "step": 7800
+    },
+    {
+      "epoch": 2.285980546333508,
+      "grad_norm": 0.31943801045417786,
+      "learning_rate": 0.0005729055944055944,
+      "loss": 3.7567,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3005416739472304,
+      "grad_norm": 0.3134506940841675,
+      "learning_rate": 0.0005727307692307692,
+      "loss": 3.7585,
+      "step": 7900
+    },
+    {
+      "epoch": 2.3151028015609527,
+      "grad_norm": 0.32046815752983093,
+      "learning_rate": 0.0005725559440559441,
+      "loss": 3.7384,
+      "step": 7950
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "grad_norm": 0.34398654103279114,
+      "learning_rate": 0.0005723811188811188,
+      "loss": 3.751,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "eval_accuracy": 0.34463184446561096,
+      "eval_loss": 3.7817397117614746,
+      "eval_runtime": 177.3233,
+      "eval_samples_per_second": 93.874,
+      "eval_steps_per_second": 5.871,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3442250567883978,
+      "grad_norm": 0.3400885760784149,
+      "learning_rate": 0.0005722062937062937,
+      "loss": 3.7609,
+      "step": 8050
+    },
+    {
+      "epoch": 2.35878618440212,
+      "grad_norm": 0.30969181656837463,
+      "learning_rate": 0.0005720314685314685,
+      "loss": 3.754,
+      "step": 8100
+    },
+    {
+      "epoch": 2.3733473120158424,
+      "grad_norm": 0.35174381732940674,
+      "learning_rate": 0.0005718566433566433,
+      "loss": 3.7636,
+      "step": 8150
+    },
+    {
+      "epoch": 2.387908439629565,
+      "grad_norm": 0.30494076013565063,
+      "learning_rate": 0.0005716818181818181,
+      "loss": 3.7505,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4024695672432874,
+      "grad_norm": 0.3317963480949402,
+      "learning_rate": 0.000571506993006993,
+      "loss": 3.7411,
+      "step": 8250
+    },
+    {
+      "epoch": 2.4170306948570097,
+      "grad_norm": 0.3330554962158203,
+      "learning_rate": 0.0005713321678321678,
+      "loss": 3.7478,
+      "step": 8300
+    },
+    {
+      "epoch": 2.431591822470732,
+      "grad_norm": 0.33387240767478943,
+      "learning_rate": 0.0005711573426573426,
+      "loss": 3.7595,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4461529500844543,
+      "grad_norm": 0.3251585066318512,
+      "learning_rate": 0.0005709825174825175,
+      "loss": 3.744,
+      "step": 8400
+    },
+    {
+      "epoch": 2.460714077698177,
+      "grad_norm": 0.3279932141304016,
+      "learning_rate": 0.0005708076923076923,
+      "loss": 3.75,
+      "step": 8450
+    },
+    {
+      "epoch": 2.4752752053118994,
+      "grad_norm": 0.317627876996994,
+      "learning_rate": 0.0005706328671328671,
+      "loss": 3.7403,
+      "step": 8500
+    },
+    {
+      "epoch": 2.4898363329256217,
+      "grad_norm": 0.33804240822792053,
+      "learning_rate": 0.0005704580419580419,
+      "loss": 3.7408,
+      "step": 8550
+    },
+    {
+      "epoch": 2.5043974605393444,
+      "grad_norm": 0.333448588848114,
+      "learning_rate": 0.0005702832167832168,
+      "loss": 3.7505,
+      "step": 8600
+    },
+    {
+      "epoch": 2.5189585881530667,
+      "grad_norm": 0.325022429227829,
+      "learning_rate": 0.0005701083916083916,
+      "loss": 3.7527,
+      "step": 8650
+    },
+    {
+      "epoch": 2.533519715766789,
+      "grad_norm": 0.3052625358104706,
+      "learning_rate": 0.0005699335664335664,
+      "loss": 3.7333,
+      "step": 8700
+    },
+    {
+      "epoch": 2.5480808433805113,
+      "grad_norm": 0.3203795850276947,
+      "learning_rate": 0.0005697587412587412,
+      "loss": 3.7428,
+      "step": 8750
+    },
+    {
+      "epoch": 2.5626419709942336,
+      "grad_norm": 0.30669328570365906,
+      "learning_rate": 0.000569583916083916,
+      "loss": 3.7437,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5772030986079564,
+      "grad_norm": 0.3302299976348877,
+      "learning_rate": 0.0005694090909090908,
+      "loss": 3.7527,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5917642262216787,
+      "grad_norm": 0.3199024498462677,
+      "learning_rate": 0.0005692342657342657,
+      "loss": 3.7376,
+      "step": 8900
+    },
+    {
+      "epoch": 2.606325353835401,
+      "grad_norm": 0.3255689740180969,
+      "learning_rate": 0.0005690594405594405,
+      "loss": 3.7342,
+      "step": 8950
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "grad_norm": 0.33225247263908386,
+      "learning_rate": 0.0005688846153846153,
+      "loss": 3.7396,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "eval_accuracy": 0.347539990684339,
+      "eval_loss": 3.750495672225952,
+      "eval_runtime": 177.3467,
+      "eval_samples_per_second": 93.861,
+      "eval_steps_per_second": 5.87,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6354476090628456,
+      "grad_norm": 0.31644511222839355,
+      "learning_rate": 0.0005687097902097901,
+      "loss": 3.737,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6500087366765683,
+      "grad_norm": 0.32711660861968994,
+      "learning_rate": 0.000568534965034965,
+      "loss": 3.7258,
+      "step": 9100
+    },
+    {
+      "epoch": 2.6645698642902906,
+      "grad_norm": 0.32171499729156494,
+      "learning_rate": 0.0005683601398601398,
+      "loss": 3.7459,
+      "step": 9150
+    },
+    {
+      "epoch": 2.679130991904013,
+      "grad_norm": 0.30605417490005493,
+      "learning_rate": 0.0005681853146853146,
+      "loss": 3.7252,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6936921195177357,
+      "grad_norm": 0.31815701723098755,
+      "learning_rate": 0.0005680104895104895,
+      "loss": 3.7312,
+      "step": 9250
+    },
+    {
+      "epoch": 2.708253247131458,
+      "grad_norm": 0.320014625787735,
+      "learning_rate": 0.0005678356643356643,
+      "loss": 3.7287,
+      "step": 9300
+    },
+    {
+      "epoch": 2.7228143747451803,
+      "grad_norm": 0.31022703647613525,
+      "learning_rate": 0.0005676608391608391,
+      "loss": 3.7285,
+      "step": 9350
+    },
+    {
+      "epoch": 2.7373755023589026,
+      "grad_norm": 0.33272796869277954,
+      "learning_rate": 0.0005674860139860139,
+      "loss": 3.7302,
+      "step": 9400
+    },
+    {
+      "epoch": 2.751936629972625,
+      "grad_norm": 0.336722731590271,
+      "learning_rate": 0.0005673111888111888,
+      "loss": 3.713,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7664977575863476,
+      "grad_norm": 0.3396904468536377,
+      "learning_rate": 0.0005671363636363635,
+      "loss": 3.7374,
+      "step": 9500
+    },
+    {
+      "epoch": 2.78105888520007,
+      "grad_norm": 0.3151884973049164,
+      "learning_rate": 0.0005669615384615384,
+      "loss": 3.7306,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7956200128137922,
+      "grad_norm": 0.31197404861450195,
+      "learning_rate": 0.0005667867132867132,
+      "loss": 3.7201,
+      "step": 9600
+    },
+    {
+      "epoch": 2.8101811404275145,
+      "grad_norm": 0.33173638582229614,
+      "learning_rate": 0.000566611888111888,
+      "loss": 3.7265,
+      "step": 9650
+    },
+    {
+      "epoch": 2.824742268041237,
+      "grad_norm": 0.31983959674835205,
+      "learning_rate": 0.0005664370629370628,
+      "loss": 3.7218,
+      "step": 9700
+    },
+    {
+      "epoch": 2.8393033956549596,
+      "grad_norm": 0.3080846071243286,
+      "learning_rate": 0.0005662622377622377,
+      "loss": 3.7198,
+      "step": 9750
+    },
+    {
+      "epoch": 2.853864523268682,
+      "grad_norm": 0.31331005692481995,
+      "learning_rate": 0.0005660874125874125,
+      "loss": 3.7078,
+      "step": 9800
+    },
+    {
+      "epoch": 2.868425650882404,
+      "grad_norm": 0.3520505428314209,
+      "learning_rate": 0.0005659125874125873,
+      "loss": 3.726,
+      "step": 9850
+    },
+    {
+      "epoch": 2.882986778496127,
+      "grad_norm": 0.3302844762802124,
+      "learning_rate": 0.0005657377622377622,
+      "loss": 3.7182,
+      "step": 9900
+    },
+    {
+      "epoch": 2.8975479061098492,
+      "grad_norm": 0.3250056803226471,
+      "learning_rate": 0.000565562937062937,
+      "loss": 3.7079,
+      "step": 9950
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "grad_norm": 0.3227342963218689,
+      "learning_rate": 0.0005653881118881118,
+      "loss": 3.7105,
+      "step": 10000
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "eval_accuracy": 0.3495654768468674,
+      "eval_loss": 3.724705457687378,
+      "eval_runtime": 177.4012,
+      "eval_samples_per_second": 93.833,
+      "eval_steps_per_second": 5.868,
+      "step": 10000
+    },
+    {
+      "epoch": 2.926670161337294,
+      "grad_norm": 0.32378628849983215,
+      "learning_rate": 0.0005652132867132866,
+      "loss": 3.7132,
+      "step": 10050
+    },
+    {
+      "epoch": 2.941231288951016,
+      "grad_norm": 0.3264535665512085,
+      "learning_rate": 0.0005650384615384615,
+      "loss": 3.72,
+      "step": 10100
+    },
+    {
+      "epoch": 2.955792416564739,
+      "grad_norm": 0.3335714042186737,
+      "learning_rate": 0.0005648636363636363,
+      "loss": 3.714,
+      "step": 10150
+    },
+    {
+      "epoch": 2.970353544178461,
+      "grad_norm": 0.3222605288028717,
+      "learning_rate": 0.0005646888111888111,
+      "loss": 3.7256,
+      "step": 10200
+    },
+    {
+      "epoch": 2.9849146717921835,
+      "grad_norm": 0.3324715495109558,
+      "learning_rate": 0.000564513986013986,
+      "loss": 3.7218,
+      "step": 10250
+    },
+    {
+      "epoch": 2.9994757994059063,
+      "grad_norm": 0.3224101960659027,
+      "learning_rate": 0.0005643391608391607,
+      "loss": 3.7147,
+      "step": 10300
+    },
+    {
+      "epoch": 3.0139786825091734,
+      "grad_norm": 0.3330100476741791,
+      "learning_rate": 0.0005641643356643355,
+      "loss": 3.6132,
+      "step": 10350
+    },
+    {
+      "epoch": 3.0285398101228957,
+      "grad_norm": 0.3183719515800476,
+      "learning_rate": 0.0005639895104895105,
+      "loss": 3.6057,
+      "step": 10400
+    },
+    {
+      "epoch": 3.0431009377366185,
+      "grad_norm": 0.31805160641670227,
+      "learning_rate": 0.0005638146853146853,
+      "loss": 3.616,
+      "step": 10450
+    },
+    {
+      "epoch": 3.057662065350341,
+      "grad_norm": 0.3436165153980255,
+      "learning_rate": 0.0005636398601398601,
+      "loss": 3.625,
+      "step": 10500
+    },
+    {
+      "epoch": 3.072223192964063,
+      "grad_norm": 0.32004937529563904,
+      "learning_rate": 0.000563465034965035,
+      "loss": 3.6283,
+      "step": 10550
+    },
+    {
+      "epoch": 3.0867843205777854,
+      "grad_norm": 0.3213531970977783,
+      "learning_rate": 0.0005632902097902098,
+      "loss": 3.6217,
+      "step": 10600
+    },
+    {
+      "epoch": 3.101345448191508,
+      "grad_norm": 0.31485000252723694,
+      "learning_rate": 0.0005631153846153846,
+      "loss": 3.6229,
+      "step": 10650
+    },
+    {
+      "epoch": 3.1159065758052304,
+      "grad_norm": 0.3263075649738312,
+      "learning_rate": 0.0005629405594405594,
+      "loss": 3.6317,
+      "step": 10700
+    },
+    {
+      "epoch": 3.1304677034189528,
+      "grad_norm": 0.34470370411872864,
+      "learning_rate": 0.0005627657342657343,
+      "loss": 3.6219,
+      "step": 10750
+    },
+    {
+      "epoch": 3.145028831032675,
+      "grad_norm": 0.33321434259414673,
+      "learning_rate": 0.0005625909090909091,
+      "loss": 3.6205,
+      "step": 10800
+    },
+    {
+      "epoch": 3.1595899586463974,
+      "grad_norm": 0.32217034697532654,
+      "learning_rate": 0.0005624160839160839,
+      "loss": 3.6309,
+      "step": 10850
+    },
+    {
+      "epoch": 3.17415108626012,
+      "grad_norm": 0.34395185112953186,
+      "learning_rate": 0.0005622412587412587,
+      "loss": 3.6366,
+      "step": 10900
+    },
+    {
+      "epoch": 3.1887122138738424,
+      "grad_norm": 0.3254980146884918,
+      "learning_rate": 0.0005620664335664336,
+      "loss": 3.6369,
+      "step": 10950
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "grad_norm": 0.338420033454895,
+      "learning_rate": 0.0005618916083916083,
+      "loss": 3.6418,
+      "step": 11000
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "eval_accuracy": 0.3518970960390101,
+      "eval_loss": 3.711723566055298,
+      "eval_runtime": 177.6335,
+      "eval_samples_per_second": 93.71,
+      "eval_steps_per_second": 5.86,
+      "step": 11000
+    },
+    {
+      "epoch": 3.217834469101287,
+      "grad_norm": 0.3291839361190796,
+      "learning_rate": 0.0005617167832167832,
+      "loss": 3.6324,
+      "step": 11050
+    },
+    {
+      "epoch": 3.2323955967150098,
+      "grad_norm": 0.3438150882720947,
+      "learning_rate": 0.000561541958041958,
+      "loss": 3.6253,
+      "step": 11100
+    },
+    {
+      "epoch": 3.246956724328732,
+      "grad_norm": 0.36098095774650574,
+      "learning_rate": 0.0005613671328671328,
+      "loss": 3.6401,
+      "step": 11150
+    },
+    {
+      "epoch": 3.2615178519424544,
+      "grad_norm": 0.351468026638031,
+      "learning_rate": 0.0005611923076923077,
+      "loss": 3.6395,
+      "step": 11200
+    },
+    {
+      "epoch": 3.2760789795561767,
+      "grad_norm": 0.3300299644470215,
+      "learning_rate": 0.0005610174825174825,
+      "loss": 3.644,
+      "step": 11250
+    },
+    {
+      "epoch": 3.2906401071698994,
+      "grad_norm": 0.3039533495903015,
+      "learning_rate": 0.0005608426573426573,
+      "loss": 3.6416,
+      "step": 11300
+    },
+    {
+      "epoch": 3.3052012347836217,
+      "grad_norm": 0.35001346468925476,
+      "learning_rate": 0.0005606678321678321,
+      "loss": 3.6403,
+      "step": 11350
+    },
+    {
+      "epoch": 3.319762362397344,
+      "grad_norm": 0.3169427514076233,
+      "learning_rate": 0.000560493006993007,
+      "loss": 3.6419,
+      "step": 11400
+    },
+    {
+      "epoch": 3.3343234900110663,
+      "grad_norm": 0.3217058777809143,
+      "learning_rate": 0.0005603181818181818,
+      "loss": 3.6448,
+      "step": 11450
+    },
+    {
+      "epoch": 3.3488846176247886,
+      "grad_norm": 0.3203667402267456,
+      "learning_rate": 0.0005601433566433566,
+      "loss": 3.6387,
+      "step": 11500
+    },
+    {
+      "epoch": 3.3634457452385114,
+      "grad_norm": 0.3251610994338989,
+      "learning_rate": 0.0005599685314685314,
+      "loss": 3.6272,
+      "step": 11550
+    },
+    {
+      "epoch": 3.3780068728522337,
+      "grad_norm": 0.329180508852005,
+      "learning_rate": 0.0005597937062937063,
+      "loss": 3.6358,
+      "step": 11600
+    },
+    {
+      "epoch": 3.392568000465956,
+      "grad_norm": 0.31434664130210876,
+      "learning_rate": 0.0005596188811188811,
+      "loss": 3.6503,
+      "step": 11650
+    },
+    {
+      "epoch": 3.4071291280796787,
+      "grad_norm": 0.3283628225326538,
+      "learning_rate": 0.0005594440559440559,
+      "loss": 3.6348,
+      "step": 11700
+    },
+    {
+      "epoch": 3.421690255693401,
+      "grad_norm": 0.3240419626235962,
+      "learning_rate": 0.0005592692307692307,
+      "loss": 3.6482,
+      "step": 11750
+    },
+    {
+      "epoch": 3.4362513833071233,
+      "grad_norm": 0.32653725147247314,
+      "learning_rate": 0.0005590944055944055,
+      "loss": 3.6465,
+      "step": 11800
+    },
+    {
+      "epoch": 3.4508125109208456,
+      "grad_norm": 0.3058265447616577,
+      "learning_rate": 0.0005589195804195803,
+      "loss": 3.6452,
+      "step": 11850
+    },
+    {
+      "epoch": 3.465373638534568,
+      "grad_norm": 0.3274167776107788,
+      "learning_rate": 0.0005587447552447552,
+      "loss": 3.6373,
+      "step": 11900
+    },
+    {
+      "epoch": 3.4799347661482907,
+      "grad_norm": 0.34979650378227234,
+      "learning_rate": 0.00055856993006993,
+      "loss": 3.6462,
+      "step": 11950
+    },
+    {
+      "epoch": 3.494495893762013,
+      "grad_norm": 0.3270816206932068,
+      "learning_rate": 0.0005583951048951048,
+      "loss": 3.6293,
+      "step": 12000
+    },
+    {
+      "epoch": 3.494495893762013,
+      "eval_accuracy": 0.3533076122023403,
+      "eval_loss": 3.6936569213867188,
+      "eval_runtime": 177.1638,
+      "eval_samples_per_second": 93.958,
+      "eval_steps_per_second": 5.876,
+      "step": 12000
+    },
+    {
+      "epoch": 3.5090570213757353,
+      "grad_norm": 0.3215574026107788,
+      "learning_rate": 0.0005582202797202797,
+      "loss": 3.632,
+      "step": 12050
+    },
+    {
+      "epoch": 3.523618148989458,
+      "grad_norm": 0.326255738735199,
+      "learning_rate": 0.0005580454545454545,
+      "loss": 3.6415,
+      "step": 12100
+    },
+    {
+      "epoch": 3.53817927660318,
+      "grad_norm": 0.33679527044296265,
+      "learning_rate": 0.0005578706293706293,
+      "loss": 3.6456,
+      "step": 12150
+    },
+    {
+      "epoch": 3.5527404042169026,
+      "grad_norm": 0.3434588313102722,
+      "learning_rate": 0.0005576958041958041,
+      "loss": 3.6466,
+      "step": 12200
+    },
+    {
+      "epoch": 3.567301531830625,
+      "grad_norm": 0.31146910786628723,
+      "learning_rate": 0.000557520979020979,
+      "loss": 3.6367,
+      "step": 12250
+    },
+    {
+      "epoch": 3.5818626594443472,
+      "grad_norm": 0.3134685158729553,
+      "learning_rate": 0.0005573461538461538,
+      "loss": 3.6253,
+      "step": 12300
+    },
+    {
+      "epoch": 3.59642378705807,
+      "grad_norm": 0.31741416454315186,
+      "learning_rate": 0.0005571713286713286,
+      "loss": 3.6344,
+      "step": 12350
+    },
+    {
+      "epoch": 3.6109849146717923,
+      "grad_norm": 0.3213517963886261,
+      "learning_rate": 0.0005569965034965034,
+      "loss": 3.6461,
+      "step": 12400
+    },
+    {
+      "epoch": 3.6255460422855146,
+      "grad_norm": 0.3236750662326813,
+      "learning_rate": 0.0005568216783216783,
+      "loss": 3.6363,
+      "step": 12450
+    },
+    {
+      "epoch": 3.640107169899237,
+      "grad_norm": 0.3350413143634796,
+      "learning_rate": 0.000556646853146853,
+      "loss": 3.6341,
+      "step": 12500
+    },
+    {
+      "epoch": 3.654668297512959,
+      "grad_norm": 0.3210690915584564,
+      "learning_rate": 0.0005564720279720279,
+      "loss": 3.6341,
+      "step": 12550
+    },
+    {
+      "epoch": 3.669229425126682,
+      "grad_norm": 0.31280940771102905,
+      "learning_rate": 0.0005562972027972027,
+      "loss": 3.6406,
+      "step": 12600
+    },
+    {
+      "epoch": 3.6837905527404042,
+      "grad_norm": 0.3478114902973175,
+      "learning_rate": 0.0005561223776223775,
+      "loss": 3.6437,
+      "step": 12650
+    },
+    {
+      "epoch": 3.6983516803541265,
+      "grad_norm": 0.3223908245563507,
+      "learning_rate": 0.0005559475524475524,
+      "loss": 3.6346,
+      "step": 12700
+    },
+    {
+      "epoch": 3.7129128079678493,
+      "grad_norm": 0.30570903420448303,
+      "learning_rate": 0.0005557727272727272,
+      "loss": 3.6475,
+      "step": 12750
+    },
+    {
+      "epoch": 3.7274739355815716,
+      "grad_norm": 0.3094755709171295,
+      "learning_rate": 0.000555597902097902,
+      "loss": 3.6307,
+      "step": 12800
+    },
+    {
+      "epoch": 3.742035063195294,
+      "grad_norm": 0.2975589334964752,
+      "learning_rate": 0.0005554230769230768,
+      "loss": 3.6396,
+      "step": 12850
+    },
+    {
+      "epoch": 3.756596190809016,
+      "grad_norm": 0.3289732038974762,
+      "learning_rate": 0.0005552482517482517,
+      "loss": 3.6434,
+      "step": 12900
+    },
+    {
+      "epoch": 3.7711573184227385,
+      "grad_norm": 0.32117950916290283,
+      "learning_rate": 0.0005550734265734265,
+      "loss": 3.6423,
+      "step": 12950
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "grad_norm": 0.32686787843704224,
+      "learning_rate": 0.0005548986013986013,
+      "loss": 3.642,
+      "step": 13000
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "eval_accuracy": 0.35558350671858546,
+      "eval_loss": 3.676309823989868,
+      "eval_runtime": 177.0944,
+      "eval_samples_per_second": 93.995,
+      "eval_steps_per_second": 5.878,
+      "step": 13000
+    },
+    {
+      "epoch": 3.8002795736501835,
+      "grad_norm": 0.3189132809638977,
+      "learning_rate": 0.0005547237762237761,
+      "loss": 3.6439,
+      "step": 13050
+    },
+    {
+      "epoch": 3.814840701263906,
+      "grad_norm": 0.33861854672431946,
+      "learning_rate": 0.000554548951048951,
+      "loss": 3.6343,
+      "step": 13100
+    },
+    {
+      "epoch": 3.829401828877628,
+      "grad_norm": 0.3242887854576111,
+      "learning_rate": 0.0005543741258741258,
+      "loss": 3.64,
+      "step": 13150
+    },
+    {
+      "epoch": 3.8439629564913504,
+      "grad_norm": 0.3283834457397461,
+      "learning_rate": 0.0005541993006993006,
+      "loss": 3.6236,
+      "step": 13200
+    },
+    {
+      "epoch": 3.858524084105073,
+      "grad_norm": 0.3189845085144043,
+      "learning_rate": 0.0005540244755244756,
+      "loss": 3.6343,
+      "step": 13250
+    },
+    {
+      "epoch": 3.8730852117187955,
+      "grad_norm": 0.3211795687675476,
+      "learning_rate": 0.0005538496503496502,
+      "loss": 3.6512,
+      "step": 13300
+    },
+    {
+      "epoch": 3.887646339332518,
+      "grad_norm": 0.3205946087837219,
+      "learning_rate": 0.0005536748251748252,
+      "loss": 3.63,
+      "step": 13350
+    },
+    {
+      "epoch": 3.9022074669462405,
+      "grad_norm": 0.3182278275489807,
+      "learning_rate": 0.0005535,
+      "loss": 3.6192,
+      "step": 13400
+    },
+    {
+      "epoch": 3.916768594559963,
+      "grad_norm": 0.3093646466732025,
+      "learning_rate": 0.0005533251748251748,
+      "loss": 3.629,
+      "step": 13450
+    },
+    {
+      "epoch": 3.931329722173685,
+      "grad_norm": 0.325284481048584,
+      "learning_rate": 0.0005531503496503496,
+      "loss": 3.6445,
+      "step": 13500
+    },
+    {
+      "epoch": 3.9458908497874075,
+      "grad_norm": 0.34685397148132324,
+      "learning_rate": 0.0005529755244755245,
+      "loss": 3.6269,
+      "step": 13550
+    },
+    {
+      "epoch": 3.9604519774011298,
+      "grad_norm": 0.3556680977344513,
+      "learning_rate": 0.0005528006993006993,
+      "loss": 3.6371,
+      "step": 13600
+    },
+    {
+      "epoch": 3.9750131050148525,
+      "grad_norm": 0.3256296217441559,
+      "learning_rate": 0.0005526258741258741,
+      "loss": 3.6367,
+      "step": 13650
+    },
+    {
+      "epoch": 3.989574232628575,
+      "grad_norm": 0.31314557790756226,
+      "learning_rate": 0.0005524510489510489,
+      "loss": 3.6367,
+      "step": 13700
+    },
+    {
+      "epoch": 4.004077115731842,
+      "grad_norm": 0.34284132719039917,
+      "learning_rate": 0.0005522762237762238,
+      "loss": 3.6039,
+      "step": 13750
+    },
+    {
+      "epoch": 4.018638243345564,
+      "grad_norm": 0.31531134247779846,
+      "learning_rate": 0.0005521013986013986,
+      "loss": 3.5281,
+      "step": 13800
+    },
+    {
+      "epoch": 4.033199370959287,
+      "grad_norm": 0.329609751701355,
+      "learning_rate": 0.0005519265734265734,
+      "loss": 3.5438,
+      "step": 13850
+    },
+    {
+      "epoch": 4.04776049857301,
+      "grad_norm": 0.3221539855003357,
+      "learning_rate": 0.0005517517482517482,
+      "loss": 3.5235,
+      "step": 13900
+    },
+    {
+      "epoch": 4.062321626186732,
+      "grad_norm": 0.3120812177658081,
+      "learning_rate": 0.0005515769230769231,
+      "loss": 3.5343,
+      "step": 13950
+    },
+    {
+      "epoch": 4.076882753800454,
+      "grad_norm": 0.32255053520202637,
+      "learning_rate": 0.0005514020979020979,
+      "loss": 3.543,
+      "step": 14000
+    },
+    {
+      "epoch": 4.076882753800454,
+      "eval_accuracy": 0.35604047257346666,
+      "eval_loss": 3.672227382659912,
+      "eval_runtime": 177.2502,
+      "eval_samples_per_second": 93.912,
+      "eval_steps_per_second": 5.873,
+      "step": 14000
+    },
+    {
+      "epoch": 4.091443881414177,
+      "grad_norm": 0.33779409527778625,
+      "learning_rate": 0.0005512272727272727,
+      "loss": 3.5375,
+      "step": 14050
+    },
+    {
+      "epoch": 4.106005009027899,
+      "grad_norm": 0.30933794379234314,
+      "learning_rate": 0.0005510524475524475,
+      "loss": 3.5311,
+      "step": 14100
+    },
+    {
+      "epoch": 4.120566136641622,
+      "grad_norm": 0.3431943356990814,
+      "learning_rate": 0.0005508776223776223,
+      "loss": 3.532,
+      "step": 14150
+    },
+    {
+      "epoch": 4.135127264255344,
+      "grad_norm": 0.31620094180107117,
+      "learning_rate": 0.0005507027972027972,
+      "loss": 3.5503,
+      "step": 14200
+    },
+    {
+      "epoch": 4.149688391869066,
+      "grad_norm": 0.3254013657569885,
+      "learning_rate": 0.000550527972027972,
+      "loss": 3.5538,
+      "step": 14250
+    },
+    {
+      "epoch": 4.164249519482789,
+      "grad_norm": 0.33322015404701233,
+      "learning_rate": 0.0005503531468531468,
+      "loss": 3.5521,
+      "step": 14300
+    },
+    {
+      "epoch": 4.178810647096511,
+      "grad_norm": 0.3261924386024475,
+      "learning_rate": 0.0005501783216783216,
+      "loss": 3.5439,
+      "step": 14350
+    },
+    {
+      "epoch": 4.193371774710234,
+      "grad_norm": 0.32681429386138916,
+      "learning_rate": 0.0005500034965034965,
+      "loss": 3.5543,
+      "step": 14400
+    },
+    {
+      "epoch": 4.207932902323956,
+      "grad_norm": 0.3240799307823181,
+      "learning_rate": 0.0005498286713286713,
+      "loss": 3.5622,
+      "step": 14450
+    },
+    {
+      "epoch": 4.222494029937678,
+      "grad_norm": 0.32154789566993713,
+      "learning_rate": 0.0005496538461538461,
+      "loss": 3.5425,
+      "step": 14500
+    },
+    {
+      "epoch": 4.237055157551401,
+      "grad_norm": 0.3319743275642395,
+      "learning_rate": 0.0005494790209790209,
+      "loss": 3.5755,
+      "step": 14550
+    },
+    {
+      "epoch": 4.251616285165123,
+      "grad_norm": 0.32192280888557434,
+      "learning_rate": 0.0005493041958041958,
+      "loss": 3.5474,
+      "step": 14600
+    },
+    {
+      "epoch": 4.266177412778846,
+      "grad_norm": 0.3264658451080322,
+      "learning_rate": 0.0005491293706293706,
+      "loss": 3.5525,
+      "step": 14650
+    },
+    {
+      "epoch": 4.280738540392568,
+      "grad_norm": 0.30953332781791687,
+      "learning_rate": 0.0005489545454545454,
+      "loss": 3.5686,
+      "step": 14700
+    },
+    {
+      "epoch": 4.29529966800629,
+      "grad_norm": 0.32248425483703613,
+      "learning_rate": 0.0005487797202797203,
+      "loss": 3.5576,
+      "step": 14750
+    },
+    {
+      "epoch": 4.309860795620013,
+      "grad_norm": 0.3248192071914673,
+      "learning_rate": 0.000548604895104895,
+      "loss": 3.5652,
+      "step": 14800
+    },
+    {
+      "epoch": 4.324421923233735,
+      "grad_norm": 0.3153315484523773,
+      "learning_rate": 0.0005484300699300699,
+      "loss": 3.5583,
+      "step": 14850
+    },
+    {
+      "epoch": 4.338983050847458,
+      "grad_norm": 0.33788663148880005,
+      "learning_rate": 0.0005482552447552447,
+      "loss": 3.5586,
+      "step": 14900
+    },
+    {
+      "epoch": 4.35354417846118,
+      "grad_norm": 0.3236865699291229,
+      "learning_rate": 0.0005480804195804195,
+      "loss": 3.5692,
+      "step": 14950
+    },
+    {
+      "epoch": 4.368105306074902,
+      "grad_norm": 0.3407519459724426,
+      "learning_rate": 0.0005479055944055943,
+      "loss": 3.5577,
+      "step": 15000
+    },
+    {
+      "epoch": 4.368105306074902,
+      "eval_accuracy": 0.3575495062017802,
+      "eval_loss": 3.658498764038086,
+      "eval_runtime": 176.9515,
+      "eval_samples_per_second": 94.071,
+      "eval_steps_per_second": 5.883,
+      "step": 15000
+    },
+    {
+      "epoch": 4.382666433688625,
+      "grad_norm": 0.3334459364414215,
+      "learning_rate": 0.0005477307692307692,
+      "loss": 3.5686,
+      "step": 15050
+    },
+    {
+      "epoch": 4.397227561302348,
+      "grad_norm": 0.32275843620300293,
+      "learning_rate": 0.000547555944055944,
+      "loss": 3.5695,
+      "step": 15100
+    },
+    {
+      "epoch": 4.41178868891607,
+      "grad_norm": 0.3123950660228729,
+      "learning_rate": 0.0005473811188811188,
+      "loss": 3.5732,
+      "step": 15150
+    },
+    {
+      "epoch": 4.426349816529792,
+      "grad_norm": 0.32868602871894836,
+      "learning_rate": 0.0005472062937062936,
+      "loss": 3.5836,
+      "step": 15200
+    },
+    {
+      "epoch": 4.440910944143514,
+      "grad_norm": 0.3395336866378784,
+      "learning_rate": 0.0005470314685314685,
+      "loss": 3.5748,
+      "step": 15250
+    },
+    {
+      "epoch": 4.455472071757237,
+      "grad_norm": 0.3110513687133789,
+      "learning_rate": 0.0005468566433566433,
+      "loss": 3.5758,
+      "step": 15300
+    },
+    {
+      "epoch": 4.47003319937096,
+      "grad_norm": 0.3269290328025818,
+      "learning_rate": 0.0005466818181818181,
+      "loss": 3.5562,
+      "step": 15350
+    },
+    {
+      "epoch": 4.4845943269846815,
+      "grad_norm": 0.32525432109832764,
+      "learning_rate": 0.000546506993006993,
+      "loss": 3.5786,
+      "step": 15400
+    },
+    {
+      "epoch": 4.499155454598404,
+      "grad_norm": 0.33424168825149536,
+      "learning_rate": 0.0005463321678321678,
+      "loss": 3.577,
+      "step": 15450
+    },
+    {
+      "epoch": 4.513716582212126,
+      "grad_norm": 0.31379204988479614,
+      "learning_rate": 0.0005461573426573426,
+      "loss": 3.5716,
+      "step": 15500
+    },
+    {
+      "epoch": 4.528277709825849,
+      "grad_norm": 0.33003076910972595,
+      "learning_rate": 0.0005459825174825174,
+      "loss": 3.5673,
+      "step": 15550
+    },
+    {
+      "epoch": 4.542838837439572,
+      "grad_norm": 0.31797513365745544,
+      "learning_rate": 0.0005458076923076922,
+      "loss": 3.5634,
+      "step": 15600
+    },
+    {
+      "epoch": 4.5573999650532935,
+      "grad_norm": 0.3288755416870117,
+      "learning_rate": 0.000545632867132867,
+      "loss": 3.5642,
+      "step": 15650
+    },
+    {
+      "epoch": 4.571961092667016,
+      "grad_norm": 0.3203357756137848,
+      "learning_rate": 0.0005454580419580419,
+      "loss": 3.5713,
+      "step": 15700
+    },
+    {
+      "epoch": 4.586522220280738,
+      "grad_norm": 0.31958380341529846,
+      "learning_rate": 0.0005452832167832167,
+      "loss": 3.5633,
+      "step": 15750
+    },
+    {
+      "epoch": 4.601083347894461,
+      "grad_norm": 0.33480215072631836,
+      "learning_rate": 0.0005451083916083915,
+      "loss": 3.5648,
+      "step": 15800
+    },
+    {
+      "epoch": 4.615644475508184,
+      "grad_norm": 0.31152933835983276,
+      "learning_rate": 0.0005449335664335663,
+      "loss": 3.5824,
+      "step": 15850
+    },
+    {
+      "epoch": 4.630205603121905,
+      "grad_norm": 0.31223437190055847,
+      "learning_rate": 0.0005447587412587412,
+      "loss": 3.5717,
+      "step": 15900
+    },
+    {
+      "epoch": 4.644766730735628,
+      "grad_norm": 0.32843029499053955,
+      "learning_rate": 0.000544583916083916,
+      "loss": 3.561,
+      "step": 15950
+    },
+    {
+      "epoch": 4.659327858349351,
+      "grad_norm": 0.3143811523914337,
+      "learning_rate": 0.0005444090909090908,
+      "loss": 3.5824,
+      "step": 16000
+    },
+    {
+      "epoch": 4.659327858349351,
+      "eval_accuracy": 0.35877027631680114,
+      "eval_loss": 3.646221399307251,
+      "eval_runtime": 176.8737,
+      "eval_samples_per_second": 94.112,
+      "eval_steps_per_second": 5.886,
+      "step": 16000
+    },
+    {
+      "epoch": 4.673888985963073,
+      "grad_norm": 0.32724830508232117,
+      "learning_rate": 0.0005442342657342657,
+      "loss": 3.5791,
+      "step": 16050
+    },
+    {
+      "epoch": 4.6884501135767955,
+      "grad_norm": 0.3239496648311615,
+      "learning_rate": 0.0005440594405594405,
+      "loss": 3.5755,
+      "step": 16100
+    },
+    {
+      "epoch": 4.703011241190518,
+      "grad_norm": 0.317526251077652,
+      "learning_rate": 0.0005438846153846153,
+      "loss": 3.5594,
+      "step": 16150
+    },
+    {
+      "epoch": 4.71757236880424,
+      "grad_norm": 0.32207709550857544,
+      "learning_rate": 0.0005437097902097901,
+      "loss": 3.566,
+      "step": 16200
+    },
+    {
+      "epoch": 4.732133496417963,
+      "grad_norm": 0.32053446769714355,
+      "learning_rate": 0.0005435349650349651,
+      "loss": 3.5819,
+      "step": 16250
+    },
+    {
+      "epoch": 4.746694624031685,
+      "grad_norm": 0.3298093378543854,
+      "learning_rate": 0.0005433601398601397,
+      "loss": 3.576,
+      "step": 16300
+    },
+    {
+      "epoch": 4.7612557516454075,
+      "grad_norm": 0.32744932174682617,
+      "learning_rate": 0.0005431853146853147,
+      "loss": 3.5798,
+      "step": 16350
+    },
+    {
+      "epoch": 4.77581687925913,
+      "grad_norm": 0.3389306664466858,
+      "learning_rate": 0.0005430104895104895,
+      "loss": 3.5866,
+      "step": 16400
+    },
+    {
+      "epoch": 4.790378006872852,
+      "grad_norm": 0.3112584352493286,
+      "learning_rate": 0.0005428356643356643,
+      "loss": 3.5773,
+      "step": 16450
+    },
+    {
+      "epoch": 4.804939134486575,
+      "grad_norm": 0.32686665654182434,
+      "learning_rate": 0.0005426608391608391,
+      "loss": 3.5815,
+      "step": 16500
+    },
+    {
+      "epoch": 4.819500262100297,
+      "grad_norm": 0.3334883749485016,
+      "learning_rate": 0.000542486013986014,
+      "loss": 3.5664,
+      "step": 16550
+    },
+    {
+      "epoch": 4.834061389714019,
+      "grad_norm": 0.3258337378501892,
+      "learning_rate": 0.0005423111888111888,
+      "loss": 3.5747,
+      "step": 16600
+    },
+    {
+      "epoch": 4.848622517327742,
+      "grad_norm": 0.29793718457221985,
+      "learning_rate": 0.0005421363636363636,
+      "loss": 3.5803,
+      "step": 16650
+    },
+    {
+      "epoch": 4.863183644941464,
+      "grad_norm": 0.3122703433036804,
+      "learning_rate": 0.0005419615384615385,
+      "loss": 3.5709,
+      "step": 16700
+    },
+    {
+      "epoch": 4.877744772555187,
+      "grad_norm": 0.31609809398651123,
+      "learning_rate": 0.0005417867132867133,
+      "loss": 3.5709,
+      "step": 16750
+    },
+    {
+      "epoch": 4.892305900168909,
+      "grad_norm": 0.3263167142868042,
+      "learning_rate": 0.0005416118881118881,
+      "loss": 3.5827,
+      "step": 16800
+    },
+    {
+      "epoch": 4.906867027782631,
+      "grad_norm": 0.3147113025188446,
+      "learning_rate": 0.0005414370629370629,
+      "loss": 3.5783,
+      "step": 16850
+    },
+    {
+      "epoch": 4.921428155396354,
+      "grad_norm": 0.3340342938899994,
+      "learning_rate": 0.0005412622377622378,
+      "loss": 3.5829,
+      "step": 16900
+    },
+    {
+      "epoch": 4.935989283010076,
+      "grad_norm": 0.3196568489074707,
+      "learning_rate": 0.0005410874125874126,
+      "loss": 3.5705,
+      "step": 16950
+    },
+    {
+      "epoch": 4.950550410623799,
+      "grad_norm": 0.3239845931529999,
+      "learning_rate": 0.0005409125874125874,
+      "loss": 3.5754,
+      "step": 17000
+    },
+    {
+      "epoch": 4.950550410623799,
+      "eval_accuracy": 0.3598982895346002,
+      "eval_loss": 3.6311697959899902,
+      "eval_runtime": 176.9446,
+      "eval_samples_per_second": 94.075,
+      "eval_steps_per_second": 5.883,
+      "step": 17000
+    },
+    {
+      "epoch": 4.9651115382375215,
+      "grad_norm": 0.3366827368736267,
+      "learning_rate": 0.0005407377622377622,
+      "loss": 3.5755,
+      "step": 17050
+    },
+    {
+      "epoch": 4.979672665851243,
+      "grad_norm": 0.3341779112815857,
+      "learning_rate": 0.000540562937062937,
+      "loss": 3.5782,
+      "step": 17100
+    },
+    {
+      "epoch": 4.994233793464966,
+      "grad_norm": 0.3099096119403839,
+      "learning_rate": 0.0005403881118881118,
+      "loss": 3.5812,
+      "step": 17150
+    },
+    {
+      "epoch": 5.008736676568233,
+      "grad_norm": 0.3530953824520111,
+      "learning_rate": 0.0005402132867132867,
+      "loss": 3.4902,
+      "step": 17200
+    },
+    {
+      "epoch": 5.023297804181956,
+      "grad_norm": 0.3362038731575012,
+      "learning_rate": 0.0005400384615384615,
+      "loss": 3.4618,
+      "step": 17250
+    },
+    {
+      "epoch": 5.037858931795678,
+      "grad_norm": 0.35798731446266174,
+      "learning_rate": 0.0005398636363636363,
+      "loss": 3.4807,
+      "step": 17300
+    },
+    {
+      "epoch": 5.052420059409401,
+      "grad_norm": 0.3367793560028076,
+      "learning_rate": 0.0005396888111888111,
+      "loss": 3.4656,
+      "step": 17350
+    },
+    {
+      "epoch": 5.066981187023123,
+      "grad_norm": 0.3349305987358093,
+      "learning_rate": 0.000539513986013986,
+      "loss": 3.4776,
+      "step": 17400
+    },
+    {
+      "epoch": 5.081542314636845,
+      "grad_norm": 0.31668785214424133,
+      "learning_rate": 0.0005393391608391608,
+      "loss": 3.4841,
+      "step": 17450
+    },
+    {
+      "epoch": 5.096103442250568,
+      "grad_norm": 0.32060810923576355,
+      "learning_rate": 0.0005391643356643356,
+      "loss": 3.4835,
+      "step": 17500
+    },
+    {
+      "epoch": 5.110664569864291,
+      "grad_norm": 0.31662890315055847,
+      "learning_rate": 0.0005389895104895105,
+      "loss": 3.4778,
+      "step": 17550
+    },
+    {
+      "epoch": 5.125225697478013,
+      "grad_norm": 0.3131362795829773,
+      "learning_rate": 0.0005388146853146853,
+      "loss": 3.4767,
+      "step": 17600
+    },
+    {
+      "epoch": 5.139786825091735,
+      "grad_norm": 0.32624542713165283,
+      "learning_rate": 0.0005386398601398601,
+      "loss": 3.4872,
+      "step": 17650
+    },
+    {
+      "epoch": 5.154347952705457,
+      "grad_norm": 0.3383287787437439,
+      "learning_rate": 0.0005384650349650349,
+      "loss": 3.4867,
+      "step": 17700
+    },
+    {
+      "epoch": 5.16890908031918,
+      "grad_norm": 0.3286052644252777,
+      "learning_rate": 0.0005382902097902098,
+      "loss": 3.4852,
+      "step": 17750
+    },
+    {
+      "epoch": 5.183470207932903,
+      "grad_norm": 0.31900569796562195,
+      "learning_rate": 0.0005381153846153845,
+      "loss": 3.4994,
+      "step": 17800
+    },
+    {
+      "epoch": 5.1980313355466246,
+      "grad_norm": 0.3237035572528839,
+      "learning_rate": 0.0005379405594405594,
+      "loss": 3.4957,
+      "step": 17850
+    },
+    {
+      "epoch": 5.212592463160347,
+      "grad_norm": 0.3051255941390991,
+      "learning_rate": 0.0005377657342657342,
+      "loss": 3.5044,
+      "step": 17900
+    },
+    {
+      "epoch": 5.227153590774069,
+      "grad_norm": 0.33518698811531067,
+      "learning_rate": 0.000537590909090909,
+      "loss": 3.4936,
+      "step": 17950
+    },
+    {
+      "epoch": 5.241714718387792,
+      "grad_norm": 0.3128952980041504,
+      "learning_rate": 0.0005374160839160838,
+      "loss": 3.5,
+      "step": 18000
+    },
+    {
+      "epoch": 5.241714718387792,
+      "eval_accuracy": 0.36015245989175304,
+      "eval_loss": 3.6348447799682617,
+      "eval_runtime": 176.8036,
+      "eval_samples_per_second": 94.15,
+      "eval_steps_per_second": 5.888,
+      "step": 18000
+    },
+    {
+      "epoch": 5.256275846001515,
+      "grad_norm": 0.33034732937812805,
+      "learning_rate": 0.0005372412587412587,
+      "loss": 3.509,
+      "step": 18050
+    },
+    {
+      "epoch": 5.2708369736152365,
+      "grad_norm": 0.3423735499382019,
+      "learning_rate": 0.0005370664335664335,
+      "loss": 3.5055,
+      "step": 18100
+    },
+    {
+      "epoch": 5.285398101228959,
+      "grad_norm": 0.3269577920436859,
+      "learning_rate": 0.0005368916083916083,
+      "loss": 3.5036,
+      "step": 18150
+    },
+    {
+      "epoch": 5.299959228842681,
+      "grad_norm": 0.3377140462398529,
+      "learning_rate": 0.0005367167832167832,
+      "loss": 3.5139,
+      "step": 18200
+    },
+    {
+      "epoch": 5.314520356456404,
+      "grad_norm": 0.3444065749645233,
+      "learning_rate": 0.000536541958041958,
+      "loss": 3.5031,
+      "step": 18250
+    },
+    {
+      "epoch": 5.329081484070127,
+      "grad_norm": 0.3035888671875,
+      "learning_rate": 0.0005363671328671328,
+      "loss": 3.4989,
+      "step": 18300
+    },
+    {
+      "epoch": 5.3436426116838485,
+      "grad_norm": 0.3229525089263916,
+      "learning_rate": 0.0005361923076923076,
+      "loss": 3.5053,
+      "step": 18350
+    },
+    {
+      "epoch": 5.358203739297571,
+      "grad_norm": 0.3133704960346222,
+      "learning_rate": 0.0005360174825174825,
+      "loss": 3.5075,
+      "step": 18400
+    },
+    {
+      "epoch": 5.372764866911294,
+      "grad_norm": 0.31471043825149536,
+      "learning_rate": 0.0005358426573426573,
+      "loss": 3.5161,
+      "step": 18450
+    },
+    {
+      "epoch": 5.387325994525016,
+      "grad_norm": 0.3169424533843994,
+      "learning_rate": 0.0005356678321678321,
+      "loss": 3.5031,
+      "step": 18500
+    },
+    {
+      "epoch": 5.401887122138739,
+      "grad_norm": 0.3456338047981262,
+      "learning_rate": 0.0005354930069930069,
+      "loss": 3.5103,
+      "step": 18550
+    },
+    {
+      "epoch": 5.41644824975246,
+      "grad_norm": 0.3145761489868164,
+      "learning_rate": 0.0005353181818181817,
+      "loss": 3.5202,
+      "step": 18600
+    },
+    {
+      "epoch": 5.431009377366183,
+      "grad_norm": 0.3344932198524475,
+      "learning_rate": 0.0005351433566433565,
+      "loss": 3.5115,
+      "step": 18650
+    },
+    {
+      "epoch": 5.445570504979906,
+      "grad_norm": 0.32802441716194153,
+      "learning_rate": 0.0005349685314685314,
+      "loss": 3.5108,
+      "step": 18700
+    },
+    {
+      "epoch": 5.460131632593628,
+      "grad_norm": 0.33433371782302856,
+      "learning_rate": 0.0005347937062937062,
+      "loss": 3.5175,
+      "step": 18750
+    },
+    {
+      "epoch": 5.4746927602073505,
+      "grad_norm": 0.3225354552268982,
+      "learning_rate": 0.000534618881118881,
+      "loss": 3.5294,
+      "step": 18800
+    },
+    {
+      "epoch": 5.489253887821073,
+      "grad_norm": 0.3271482586860657,
+      "learning_rate": 0.0005344440559440559,
+      "loss": 3.515,
+      "step": 18850
+    },
+    {
+      "epoch": 5.503815015434795,
+      "grad_norm": 0.32893088459968567,
+      "learning_rate": 0.0005342692307692307,
+      "loss": 3.517,
+      "step": 18900
+    },
+    {
+      "epoch": 5.518376143048518,
+      "grad_norm": 0.3370042145252228,
+      "learning_rate": 0.0005340944055944055,
+      "loss": 3.5241,
+      "step": 18950
+    },
+    {
+      "epoch": 5.53293727066224,
+      "grad_norm": 0.32803529500961304,
+      "learning_rate": 0.0005339195804195803,
+      "loss": 3.5244,
+      "step": 19000
+    },
+    {
+      "epoch": 5.53293727066224,
+      "eval_accuracy": 0.36118689327407866,
+      "eval_loss": 3.623746395111084,
+      "eval_runtime": 177.0774,
+      "eval_samples_per_second": 94.004,
+      "eval_steps_per_second": 5.879,
+      "step": 19000
+    },
+    {
+      "epoch": 5.5474983982759625,
+      "grad_norm": 0.325192391872406,
+      "learning_rate": 0.0005337447552447552,
+      "loss": 3.5244,
+      "step": 19050
+    },
+    {
+      "epoch": 5.562059525889685,
+      "grad_norm": 0.3339932858943939,
+      "learning_rate": 0.00053356993006993,
+      "loss": 3.5122,
+      "step": 19100
+    },
+    {
+      "epoch": 5.576620653503407,
+      "grad_norm": 0.3150177597999573,
+      "learning_rate": 0.0005333951048951048,
+      "loss": 3.5213,
+      "step": 19150
+    },
+    {
+      "epoch": 5.59118178111713,
+      "grad_norm": 0.3161846697330475,
+      "learning_rate": 0.0005332202797202796,
+      "loss": 3.5178,
+      "step": 19200
+    },
+    {
+      "epoch": 5.605742908730852,
+      "grad_norm": 0.3393925130367279,
+      "learning_rate": 0.0005330454545454546,
+      "loss": 3.5075,
+      "step": 19250
+    },
+    {
+      "epoch": 5.620304036344574,
+      "grad_norm": 0.3091968595981598,
+      "learning_rate": 0.0005328706293706292,
+      "loss": 3.5195,
+      "step": 19300
+    },
+    {
+      "epoch": 5.634865163958297,
+      "grad_norm": 0.3470529317855835,
+      "learning_rate": 0.0005326958041958042,
+      "loss": 3.5365,
+      "step": 19350
+    },
+    {
+      "epoch": 5.649426291572019,
+      "grad_norm": 0.3210427165031433,
+      "learning_rate": 0.000532520979020979,
+      "loss": 3.5194,
+      "step": 19400
+    },
+    {
+      "epoch": 5.663987419185742,
+      "grad_norm": 0.3233715295791626,
+      "learning_rate": 0.0005323461538461538,
+      "loss": 3.5358,
+      "step": 19450
+    },
+    {
+      "epoch": 5.6785485467994645,
+      "grad_norm": 0.3228999376296997,
+      "learning_rate": 0.0005321713286713287,
+      "loss": 3.5195,
+      "step": 19500
+    },
+    {
+      "epoch": 5.693109674413186,
+      "grad_norm": 0.32254138588905334,
+      "learning_rate": 0.0005319965034965035,
+      "loss": 3.5227,
+      "step": 19550
+    },
+    {
+      "epoch": 5.707670802026909,
+      "grad_norm": 0.3360702097415924,
+      "learning_rate": 0.0005318216783216783,
+      "loss": 3.5264,
+      "step": 19600
+    },
+    {
+      "epoch": 5.722231929640631,
+      "grad_norm": 0.32106834650039673,
+      "learning_rate": 0.0005316468531468531,
+      "loss": 3.5294,
+      "step": 19650
+    },
+    {
+      "epoch": 5.736793057254354,
+      "grad_norm": 0.3194294273853302,
+      "learning_rate": 0.000531472027972028,
+      "loss": 3.5303,
+      "step": 19700
+    },
+    {
+      "epoch": 5.7513541848680765,
+      "grad_norm": 0.31817537546157837,
+      "learning_rate": 0.0005312972027972028,
+      "loss": 3.5289,
+      "step": 19750
+    },
+    {
+      "epoch": 5.765915312481798,
+      "grad_norm": 0.32823848724365234,
+      "learning_rate": 0.0005311223776223776,
+      "loss": 3.542,
+      "step": 19800
+    },
+    {
+      "epoch": 5.780476440095521,
+      "grad_norm": 0.33044731616973877,
+      "learning_rate": 0.0005309475524475524,
+      "loss": 3.5308,
+      "step": 19850
+    },
+    {
+      "epoch": 5.795037567709244,
+      "grad_norm": 0.30992332100868225,
+      "learning_rate": 0.0005307727272727273,
+      "loss": 3.5285,
+      "step": 19900
+    },
+    {
+      "epoch": 5.809598695322966,
+      "grad_norm": 0.3125179409980774,
+      "learning_rate": 0.0005305979020979021,
+      "loss": 3.516,
+      "step": 19950
+    },
+    {
+      "epoch": 5.824159822936688,
+      "grad_norm": 0.32367056608200073,
+      "learning_rate": 0.0005304230769230769,
+      "loss": 3.5255,
+      "step": 20000
+    },
+    {
+      "epoch": 5.824159822936688,
+      "eval_accuracy": 0.36226388432027534,
+      "eval_loss": 3.613421678543091,
+      "eval_runtime": 176.9052,
+      "eval_samples_per_second": 94.096,
+      "eval_steps_per_second": 5.885,
+      "step": 20000
+    },
+    {
+      "epoch": 5.83872095055041,
+      "grad_norm": 0.3150468170642853,
+      "learning_rate": 0.0005302482517482517,
+      "loss": 3.5362,
+      "step": 20050
+    },
+    {
+      "epoch": 5.853282078164133,
+      "grad_norm": 0.2990545332431793,
+      "learning_rate": 0.0005300734265734265,
+      "loss": 3.5295,
+      "step": 20100
+    },
+    {
+      "epoch": 5.867843205777856,
+      "grad_norm": 0.29398471117019653,
+      "learning_rate": 0.0005298986013986013,
+      "loss": 3.5227,
+      "step": 20150
+    },
+    {
+      "epoch": 5.882404333391578,
+      "grad_norm": 0.30902978777885437,
+      "learning_rate": 0.0005297237762237762,
+      "loss": 3.5248,
+      "step": 20200
+    },
+    {
+      "epoch": 5.8969654610053,
+      "grad_norm": 0.3113495409488678,
+      "learning_rate": 0.000529548951048951,
+      "loss": 3.538,
+      "step": 20250
+    },
+    {
+      "epoch": 5.911526588619022,
+      "grad_norm": 0.31345126032829285,
+      "learning_rate": 0.0005293741258741258,
+      "loss": 3.5334,
+      "step": 20300
+    },
+    {
+      "epoch": 5.926087716232745,
+      "grad_norm": 0.29870733618736267,
+      "learning_rate": 0.0005291993006993007,
+      "loss": 3.5317,
+      "step": 20350
+    },
+    {
+      "epoch": 5.940648843846468,
+      "grad_norm": 0.32265445590019226,
+      "learning_rate": 0.0005290244755244755,
+      "loss": 3.5114,
+      "step": 20400
+    },
+    {
+      "epoch": 5.95520997146019,
+      "grad_norm": 0.34477177262306213,
+      "learning_rate": 0.0005288496503496503,
+      "loss": 3.5371,
+      "step": 20450
+    },
+    {
+      "epoch": 5.969771099073912,
+      "grad_norm": 0.3217330574989319,
+      "learning_rate": 0.0005286748251748251,
+      "loss": 3.541,
+      "step": 20500
+    },
+    {
+      "epoch": 5.984332226687634,
+      "grad_norm": 0.3138887882232666,
+      "learning_rate": 0.0005285,
+      "loss": 3.5157,
+      "step": 20550
+    },
+    {
+      "epoch": 5.998893354301357,
+      "grad_norm": 0.34451672434806824,
+      "learning_rate": 0.0005283251748251748,
+      "loss": 3.5283,
+      "step": 20600
+    },
+    {
+      "epoch": 6.013396237404625,
+      "grad_norm": 0.3342776894569397,
+      "learning_rate": 0.0005281503496503496,
+      "loss": 3.4197,
+      "step": 20650
+    },
+    {
+      "epoch": 6.027957365018347,
+      "grad_norm": 0.31999471783638,
+      "learning_rate": 0.0005279755244755244,
+      "loss": 3.4206,
+      "step": 20700
+    },
+    {
+      "epoch": 6.04251849263207,
+      "grad_norm": 0.3338138461112976,
+      "learning_rate": 0.0005278006993006993,
+      "loss": 3.4196,
+      "step": 20750
+    },
+    {
+      "epoch": 6.0570796202457915,
+      "grad_norm": 0.3524891436100006,
+      "learning_rate": 0.000527625874125874,
+      "loss": 3.4353,
+      "step": 20800
+    },
+    {
+      "epoch": 6.071640747859514,
+      "grad_norm": 0.3189772367477417,
+      "learning_rate": 0.0005274510489510489,
+      "loss": 3.4282,
+      "step": 20850
+    },
+    {
+      "epoch": 6.086201875473237,
+      "grad_norm": 0.3258211314678192,
+      "learning_rate": 0.0005272762237762238,
+      "loss": 3.4414,
+      "step": 20900
+    },
+    {
+      "epoch": 6.100763003086959,
+      "grad_norm": 0.33525657653808594,
+      "learning_rate": 0.0005271013986013985,
+      "loss": 3.4436,
+      "step": 20950
+    },
+    {
+      "epoch": 6.115324130700682,
+      "grad_norm": 0.3542822599411011,
+      "learning_rate": 0.0005269265734265734,
+      "loss": 3.4384,
+      "step": 21000
+    },
+    {
+      "epoch": 6.115324130700682,
+      "eval_accuracy": 0.3627670522798564,
+      "eval_loss": 3.6149725914001465,
+      "eval_runtime": 176.9568,
+      "eval_samples_per_second": 94.068,
+      "eval_steps_per_second": 5.883,
+      "step": 21000
+    },
+    {
+      "epoch": 6.1298852583144035,
+      "grad_norm": 0.3309166431427002,
+      "learning_rate": 0.0005267517482517482,
+      "loss": 3.4315,
+      "step": 21050
+    },
+    {
+      "epoch": 6.144446385928126,
+      "grad_norm": 0.3391399681568146,
+      "learning_rate": 0.000526576923076923,
+      "loss": 3.4495,
+      "step": 21100
+    },
+    {
+      "epoch": 6.159007513541849,
+      "grad_norm": 0.3149496018886566,
+      "learning_rate": 0.0005264020979020978,
+      "loss": 3.4587,
+      "step": 21150
+    },
+    {
+      "epoch": 6.173568641155571,
+      "grad_norm": 0.32553061842918396,
+      "learning_rate": 0.0005262272727272727,
+      "loss": 3.4494,
+      "step": 21200
+    },
+    {
+      "epoch": 6.1881297687692935,
+      "grad_norm": 0.33138251304626465,
+      "learning_rate": 0.0005260524475524475,
+      "loss": 3.4513,
+      "step": 21250
+    },
+    {
+      "epoch": 6.202690896383016,
+      "grad_norm": 0.32948899269104004,
+      "learning_rate": 0.0005258776223776223,
+      "loss": 3.4544,
+      "step": 21300
+    },
+    {
+      "epoch": 6.217252023996738,
+      "grad_norm": 0.341678261756897,
+      "learning_rate": 0.0005257027972027971,
+      "loss": 3.442,
+      "step": 21350
+    },
+    {
+      "epoch": 6.231813151610461,
+      "grad_norm": 0.33684396743774414,
+      "learning_rate": 0.000525527972027972,
+      "loss": 3.4379,
+      "step": 21400
+    },
+    {
+      "epoch": 6.246374279224183,
+      "grad_norm": 0.34825167059898376,
+      "learning_rate": 0.0005253531468531468,
+      "loss": 3.4584,
+      "step": 21450
+    },
+    {
+      "epoch": 6.2609354068379055,
+      "grad_norm": 0.335224449634552,
+      "learning_rate": 0.0005251783216783216,
+      "loss": 3.4535,
+      "step": 21500
+    },
+    {
+      "epoch": 6.275496534451628,
+      "grad_norm": 0.3297387659549713,
+      "learning_rate": 0.0005250034965034965,
+      "loss": 3.4562,
+      "step": 21550
+    },
+    {
+      "epoch": 6.29005766206535,
+      "grad_norm": 0.3344549238681793,
+      "learning_rate": 0.0005248286713286712,
+      "loss": 3.4579,
+      "step": 21600
+    },
+    {
+      "epoch": 6.304618789679073,
+      "grad_norm": 0.3421954810619354,
+      "learning_rate": 0.0005246538461538461,
+      "loss": 3.4619,
+      "step": 21650
+    },
+    {
+      "epoch": 6.319179917292795,
+      "grad_norm": 0.3300471007823944,
+      "learning_rate": 0.0005244790209790209,
+      "loss": 3.4587,
+      "step": 21700
+    },
+    {
+      "epoch": 6.3337410449065175,
+      "grad_norm": 0.3263568878173828,
+      "learning_rate": 0.0005243041958041957,
+      "loss": 3.4679,
+      "step": 21750
+    },
+    {
+      "epoch": 6.34830217252024,
+      "grad_norm": 0.33946892619132996,
+      "learning_rate": 0.0005241293706293705,
+      "loss": 3.4643,
+      "step": 21800
+    },
+    {
+      "epoch": 6.362863300133962,
+      "grad_norm": 0.3320576250553131,
+      "learning_rate": 0.0005239545454545454,
+      "loss": 3.4587,
+      "step": 21850
+    },
+    {
+      "epoch": 6.377424427747685,
+      "grad_norm": 0.327897846698761,
+      "learning_rate": 0.0005237797202797202,
+      "loss": 3.484,
+      "step": 21900
+    },
+    {
+      "epoch": 6.391985555361408,
+      "grad_norm": 0.32064688205718994,
+      "learning_rate": 0.000523604895104895,
+      "loss": 3.4837,
+      "step": 21950
+    },
+    {
+      "epoch": 6.406546682975129,
+      "grad_norm": 0.3372121751308441,
+      "learning_rate": 0.0005234300699300698,
+      "loss": 3.4611,
+      "step": 22000
+    },
+    {
+      "epoch": 6.406546682975129,
+      "eval_accuracy": 0.36337061870613885,
+      "eval_loss": 3.607774019241333,
+      "eval_runtime": 176.7678,
+      "eval_samples_per_second": 94.169,
+      "eval_steps_per_second": 5.889,
+      "step": 22000
+    },
+    {
+      "epoch": 6.421107810588852,
+      "grad_norm": 0.3300037384033203,
+      "learning_rate": 0.0005232552447552447,
+      "loss": 3.4701,
+      "step": 22050
+    },
+    {
+      "epoch": 6.435668938202574,
+      "grad_norm": 0.35658490657806396,
+      "learning_rate": 0.0005230804195804195,
+      "loss": 3.4581,
+      "step": 22100
+    },
+    {
+      "epoch": 6.450230065816297,
+      "grad_norm": 0.3384239077568054,
+      "learning_rate": 0.0005229055944055943,
+      "loss": 3.4855,
+      "step": 22150
+    },
+    {
+      "epoch": 6.4647911934300195,
+      "grad_norm": 0.3166094720363617,
+      "learning_rate": 0.0005227307692307691,
+      "loss": 3.4814,
+      "step": 22200
+    },
+    {
+      "epoch": 6.479352321043741,
+      "grad_norm": 0.31494995951652527,
+      "learning_rate": 0.0005225559440559441,
+      "loss": 3.4707,
+      "step": 22250
+    },
+    {
+      "epoch": 6.493913448657464,
+      "grad_norm": 0.33063846826553345,
+      "learning_rate": 0.0005223811188811189,
+      "loss": 3.49,
+      "step": 22300
+    },
+    {
+      "epoch": 6.508474576271187,
+      "grad_norm": 0.3317149579524994,
+      "learning_rate": 0.0005222062937062937,
+      "loss": 3.4876,
+      "step": 22350
+    },
+    {
+      "epoch": 6.523035703884909,
+      "grad_norm": 0.3236587941646576,
+      "learning_rate": 0.0005220314685314686,
+      "loss": 3.476,
+      "step": 22400
+    },
+    {
+      "epoch": 6.5375968314986315,
+      "grad_norm": 0.32496577501296997,
+      "learning_rate": 0.0005218566433566433,
+      "loss": 3.4801,
+      "step": 22450
+    },
+    {
+      "epoch": 6.552157959112353,
+      "grad_norm": 0.3040722906589508,
+      "learning_rate": 0.0005216818181818182,
+      "loss": 3.4714,
+      "step": 22500
+    },
+    {
+      "epoch": 6.566719086726076,
+      "grad_norm": 0.3201970160007477,
+      "learning_rate": 0.000521506993006993,
+      "loss": 3.4741,
+      "step": 22550
+    },
+    {
+      "epoch": 6.581280214339799,
+      "grad_norm": 0.3287389576435089,
+      "learning_rate": 0.0005213321678321678,
+      "loss": 3.4773,
+      "step": 22600
+    },
+    {
+      "epoch": 6.595841341953521,
+      "grad_norm": 0.3368239998817444,
+      "learning_rate": 0.0005211573426573426,
+      "loss": 3.4811,
+      "step": 22650
+    },
+    {
+      "epoch": 6.610402469567243,
+      "grad_norm": 0.341665655374527,
+      "learning_rate": 0.0005209825174825175,
+      "loss": 3.4828,
+      "step": 22700
+    },
+    {
+      "epoch": 6.624963597180965,
+      "grad_norm": 0.32684972882270813,
+      "learning_rate": 0.0005208076923076923,
+      "loss": 3.4814,
+      "step": 22750
+    },
+    {
+      "epoch": 6.639524724794688,
+      "grad_norm": 0.32613569498062134,
+      "learning_rate": 0.0005206328671328671,
+      "loss": 3.4808,
+      "step": 22800
+    },
+    {
+      "epoch": 6.654085852408411,
+      "grad_norm": 0.3358040452003479,
+      "learning_rate": 0.0005204580419580419,
+      "loss": 3.4861,
+      "step": 22850
+    },
+    {
+      "epoch": 6.668646980022133,
+      "grad_norm": 0.3069293797016144,
+      "learning_rate": 0.0005202832167832168,
+      "loss": 3.4817,
+      "step": 22900
+    },
+    {
+      "epoch": 6.683208107635855,
+      "grad_norm": 0.3199792504310608,
+      "learning_rate": 0.0005201083916083916,
+      "loss": 3.4666,
+      "step": 22950
+    },
+    {
+      "epoch": 6.697769235249577,
+      "grad_norm": 0.33008405566215515,
+      "learning_rate": 0.0005199335664335664,
+      "loss": 3.5017,
+      "step": 23000
+    },
+    {
+      "epoch": 6.697769235249577,
+      "eval_accuracy": 0.3638195903037183,
+      "eval_loss": 3.6003408432006836,
+      "eval_runtime": 176.8026,
+      "eval_samples_per_second": 94.15,
+      "eval_steps_per_second": 5.888,
+      "step": 23000
+    },
+    {
+      "epoch": 6.7123303628633,
+      "grad_norm": 0.3358388841152191,
+      "learning_rate": 0.0005197587412587413,
+      "loss": 3.4842,
+      "step": 23050
+    },
+    {
+      "epoch": 6.726891490477023,
+      "grad_norm": 0.32142195105552673,
+      "learning_rate": 0.0005195839160839161,
+      "loss": 3.4794,
+      "step": 23100
+    },
+    {
+      "epoch": 6.741452618090745,
+      "grad_norm": 0.3318839371204376,
+      "learning_rate": 0.0005194090909090909,
+      "loss": 3.4883,
+      "step": 23150
+    },
+    {
+      "epoch": 6.756013745704467,
+      "grad_norm": 0.33538463711738586,
+      "learning_rate": 0.0005192342657342657,
+      "loss": 3.4851,
+      "step": 23200
+    },
+    {
+      "epoch": 6.77057487331819,
+      "grad_norm": 0.31580328941345215,
+      "learning_rate": 0.0005190594405594405,
+      "loss": 3.482,
+      "step": 23250
+    },
+    {
+      "epoch": 6.785136000931912,
+      "grad_norm": 0.3154962658882141,
+      "learning_rate": 0.0005188846153846153,
+      "loss": 3.5067,
+      "step": 23300
+    },
+    {
+      "epoch": 6.799697128545635,
+      "grad_norm": 0.32202768325805664,
+      "learning_rate": 0.0005187097902097902,
+      "loss": 3.4836,
+      "step": 23350
+    },
+    {
+      "epoch": 6.814258256159357,
+      "grad_norm": 0.3195551335811615,
+      "learning_rate": 0.000518534965034965,
+      "loss": 3.4966,
+      "step": 23400
+    },
+    {
+      "epoch": 6.828819383773079,
+      "grad_norm": 0.33164918422698975,
+      "learning_rate": 0.0005183601398601398,
+      "loss": 3.4705,
+      "step": 23450
+    },
+    {
+      "epoch": 6.843380511386802,
+      "grad_norm": 0.32325154542922974,
+      "learning_rate": 0.0005181853146853146,
+      "loss": 3.487,
+      "step": 23500
+    },
+    {
+      "epoch": 6.857941639000524,
+      "grad_norm": 0.3257133662700653,
+      "learning_rate": 0.0005180104895104895,
+      "loss": 3.4918,
+      "step": 23550
+    },
+    {
+      "epoch": 6.872502766614247,
+      "grad_norm": 0.33086922764778137,
+      "learning_rate": 0.0005178356643356643,
+      "loss": 3.4844,
+      "step": 23600
+    },
+    {
+      "epoch": 6.887063894227969,
+      "grad_norm": 0.3378217816352844,
+      "learning_rate": 0.0005176608391608391,
+      "loss": 3.4906,
+      "step": 23650
+    },
+    {
+      "epoch": 6.901625021841691,
+      "grad_norm": 0.3248620629310608,
+      "learning_rate": 0.000517486013986014,
+      "loss": 3.4827,
+      "step": 23700
+    },
+    {
+      "epoch": 6.916186149455414,
+      "grad_norm": 0.3082542419433594,
+      "learning_rate": 0.0005173111888111888,
+      "loss": 3.4889,
+      "step": 23750
+    },
+    {
+      "epoch": 6.930747277069136,
+      "grad_norm": 0.326917439699173,
+      "learning_rate": 0.0005171363636363636,
+      "loss": 3.4922,
+      "step": 23800
+    },
+    {
+      "epoch": 6.945308404682859,
+      "grad_norm": 0.32005560398101807,
+      "learning_rate": 0.0005169615384615384,
+      "loss": 3.5049,
+      "step": 23850
+    },
+    {
+      "epoch": 6.959869532296581,
+      "grad_norm": 0.3334341049194336,
+      "learning_rate": 0.0005167867132867133,
+      "loss": 3.492,
+      "step": 23900
+    },
+    {
+      "epoch": 6.974430659910303,
+      "grad_norm": 0.32381048798561096,
+      "learning_rate": 0.000516611888111888,
+      "loss": 3.4998,
+      "step": 23950
+    },
+    {
+      "epoch": 6.988991787524026,
+      "grad_norm": 0.34249672293663025,
+      "learning_rate": 0.0005164370629370629,
+      "loss": 3.4861,
+      "step": 24000
+    },
+    {
+      "epoch": 6.988991787524026,
+      "eval_accuracy": 0.36504106579438345,
+      "eval_loss": 3.589481830596924,
+      "eval_runtime": 176.7844,
+      "eval_samples_per_second": 94.16,
+      "eval_steps_per_second": 5.889,
+      "step": 24000
+    },
+    {
+      "epoch": 7.003494670627293,
+      "grad_norm": 0.34342801570892334,
+      "learning_rate": 0.0005162622377622377,
+      "loss": 3.4622,
+      "step": 24050
+    },
+    {
+      "epoch": 7.018055798241016,
+      "grad_norm": 0.3223758935928345,
+      "learning_rate": 0.0005160874125874125,
+      "loss": 3.3769,
+      "step": 24100
+    },
+    {
+      "epoch": 7.032616925854738,
+      "grad_norm": 0.33334001898765564,
+      "learning_rate": 0.0005159125874125873,
+      "loss": 3.3793,
+      "step": 24150
+    },
+    {
+      "epoch": 7.0471780534684605,
+      "grad_norm": 0.3260599374771118,
+      "learning_rate": 0.0005157377622377622,
+      "loss": 3.3868,
+      "step": 24200
+    },
+    {
+      "epoch": 7.061739181082183,
+      "grad_norm": 0.32783156633377075,
+      "learning_rate": 0.000515562937062937,
+      "loss": 3.3913,
+      "step": 24250
+    },
+    {
+      "epoch": 7.076300308695905,
+      "grad_norm": 0.33441364765167236,
+      "learning_rate": 0.0005153881118881118,
+      "loss": 3.398,
+      "step": 24300
+    },
+    {
+      "epoch": 7.090861436309628,
+      "grad_norm": 0.31968820095062256,
+      "learning_rate": 0.0005152132867132867,
+      "loss": 3.4004,
+      "step": 24350
+    },
+    {
+      "epoch": 7.105422563923351,
+      "grad_norm": 0.3459835648536682,
+      "learning_rate": 0.0005150384615384615,
+      "loss": 3.3953,
+      "step": 24400
+    },
+    {
+      "epoch": 7.1199836915370724,
+      "grad_norm": 0.3527763783931732,
+      "learning_rate": 0.0005148636363636363,
+      "loss": 3.3997,
+      "step": 24450
+    },
+    {
+      "epoch": 7.134544819150795,
+      "grad_norm": 0.33494487404823303,
+      "learning_rate": 0.0005146888111888111,
+      "loss": 3.4178,
+      "step": 24500
+    },
+    {
+      "epoch": 7.149105946764517,
+      "grad_norm": 0.3221123218536377,
+      "learning_rate": 0.000514513986013986,
+      "loss": 3.4155,
+      "step": 24550
+    },
+    {
+      "epoch": 7.16366707437824,
+      "grad_norm": 0.34533846378326416,
+      "learning_rate": 0.0005143391608391608,
+      "loss": 3.4131,
+      "step": 24600
+    },
+    {
+      "epoch": 7.1782282019919625,
+      "grad_norm": 0.34984448552131653,
+      "learning_rate": 0.0005141643356643356,
+      "loss": 3.4036,
+      "step": 24650
+    },
+    {
+      "epoch": 7.192789329605684,
+      "grad_norm": 0.32715174555778503,
+      "learning_rate": 0.0005139895104895104,
+      "loss": 3.4013,
+      "step": 24700
+    },
+    {
+      "epoch": 7.207350457219407,
+      "grad_norm": 0.3404192626476288,
+      "learning_rate": 0.0005138146853146852,
+      "loss": 3.4106,
+      "step": 24750
+    },
+    {
+      "epoch": 7.22191158483313,
+      "grad_norm": 0.31030765175819397,
+      "learning_rate": 0.00051363986013986,
+      "loss": 3.4243,
+      "step": 24800
+    },
+    {
+      "epoch": 7.236472712446852,
+      "grad_norm": 0.3313232958316803,
+      "learning_rate": 0.0005134650349650349,
+      "loss": 3.422,
+      "step": 24850
+    },
+    {
+      "epoch": 7.2510338400605745,
+      "grad_norm": 0.31506332755088806,
+      "learning_rate": 0.0005132902097902097,
+      "loss": 3.4182,
+      "step": 24900
+    },
+    {
+      "epoch": 7.265594967674296,
+      "grad_norm": 0.32670408487319946,
+      "learning_rate": 0.0005131153846153845,
+      "loss": 3.4169,
+      "step": 24950
+    },
+    {
+      "epoch": 7.280156095288019,
+      "grad_norm": 0.3303210437297821,
+      "learning_rate": 0.0005129405594405594,
+      "loss": 3.4213,
+      "step": 25000
+    },
+    {
+      "epoch": 7.280156095288019,
+      "eval_accuracy": 0.3645839823768949,
+      "eval_loss": 3.5995564460754395,
+      "eval_runtime": 177.1172,
+      "eval_samples_per_second": 93.983,
+      "eval_steps_per_second": 5.877,
+      "step": 25000
+    },
+    {
+      "epoch": 7.294717222901742,
+      "grad_norm": 0.3087610900402069,
+      "learning_rate": 0.0005127657342657342,
+      "loss": 3.4324,
+      "step": 25050
+    },
+    {
+      "epoch": 7.309278350515464,
+      "grad_norm": 0.3354821801185608,
+      "learning_rate": 0.000512590909090909,
+      "loss": 3.4259,
+      "step": 25100
+    },
+    {
+      "epoch": 7.3238394781291865,
+      "grad_norm": 0.32773345708847046,
+      "learning_rate": 0.0005124160839160838,
+      "loss": 3.4342,
+      "step": 25150
+    },
+    {
+      "epoch": 7.338400605742908,
+      "grad_norm": 0.3321012854576111,
+      "learning_rate": 0.0005122412587412588,
+      "loss": 3.4347,
+      "step": 25200
+    },
+    {
+      "epoch": 7.352961733356631,
+      "grad_norm": 0.33389997482299805,
+      "learning_rate": 0.0005120664335664336,
+      "loss": 3.4243,
+      "step": 25250
+    },
+    {
+      "epoch": 7.367522860970354,
+      "grad_norm": 0.34130993485450745,
+      "learning_rate": 0.0005118916083916084,
+      "loss": 3.4286,
+      "step": 25300
+    },
+    {
+      "epoch": 7.382083988584076,
+      "grad_norm": 0.32038894295692444,
+      "learning_rate": 0.0005117167832167832,
+      "loss": 3.4303,
+      "step": 25350
+    },
+    {
+      "epoch": 7.396645116197798,
+      "grad_norm": 0.33677011728286743,
+      "learning_rate": 0.0005115419580419581,
+      "loss": 3.4209,
+      "step": 25400
+    },
+    {
+      "epoch": 7.411206243811521,
+      "grad_norm": 0.32208898663520813,
+      "learning_rate": 0.0005113671328671328,
+      "loss": 3.4312,
+      "step": 25450
+    },
+    {
+      "epoch": 7.425767371425243,
+      "grad_norm": 0.3141185939311981,
+      "learning_rate": 0.0005111923076923077,
+      "loss": 3.4408,
+      "step": 25500
+    },
+    {
+      "epoch": 7.440328499038966,
+      "grad_norm": 0.33646243810653687,
+      "learning_rate": 0.0005110174825174825,
+      "loss": 3.4376,
+      "step": 25550
+    },
+    {
+      "epoch": 7.454889626652688,
+      "grad_norm": 0.34247833490371704,
+      "learning_rate": 0.0005108426573426573,
+      "loss": 3.4332,
+      "step": 25600
+    },
+    {
+      "epoch": 7.46945075426641,
+      "grad_norm": 0.31767964363098145,
+      "learning_rate": 0.0005106678321678321,
+      "loss": 3.4316,
+      "step": 25650
+    },
+    {
+      "epoch": 7.484011881880133,
+      "grad_norm": 0.3620831072330475,
+      "learning_rate": 0.000510493006993007,
+      "loss": 3.4366,
+      "step": 25700
+    },
+    {
+      "epoch": 7.498573009493855,
+      "grad_norm": 0.34545835852622986,
+      "learning_rate": 0.0005103181818181818,
+      "loss": 3.4369,
+      "step": 25750
+    },
+    {
+      "epoch": 7.513134137107578,
+      "grad_norm": 0.35401225090026855,
+      "learning_rate": 0.0005101433566433566,
+      "loss": 3.4461,
+      "step": 25800
+    },
+    {
+      "epoch": 7.5276952647213005,
+      "grad_norm": 0.31585973501205444,
+      "learning_rate": 0.0005099685314685315,
+      "loss": 3.4509,
+      "step": 25850
+    },
+    {
+      "epoch": 7.542256392335022,
+      "grad_norm": 0.32004138827323914,
+      "learning_rate": 0.0005097937062937063,
+      "loss": 3.4571,
+      "step": 25900
+    },
+    {
+      "epoch": 7.556817519948745,
+      "grad_norm": 0.34505805373191833,
+      "learning_rate": 0.0005096188811188811,
+      "loss": 3.4596,
+      "step": 25950
+    },
+    {
+      "epoch": 7.571378647562467,
+      "grad_norm": 0.3367854356765747,
+      "learning_rate": 0.0005094440559440559,
+      "loss": 3.4574,
+      "step": 26000
+    },
+    {
+      "epoch": 7.571378647562467,
+      "eval_accuracy": 0.36489834478902566,
+      "eval_loss": 3.5899016857147217,
+      "eval_runtime": 177.0415,
+      "eval_samples_per_second": 94.023,
+      "eval_steps_per_second": 5.88,
+      "step": 26000
+    },
+    {
+      "epoch": 7.58593977517619,
+      "grad_norm": 0.31563594937324524,
+      "learning_rate": 0.0005092692307692308,
+      "loss": 3.4589,
+      "step": 26050
+    },
+    {
+      "epoch": 7.600500902789912,
+      "grad_norm": 0.3145526945590973,
+      "learning_rate": 0.0005090944055944056,
+      "loss": 3.4487,
+      "step": 26100
+    },
+    {
+      "epoch": 7.615062030403634,
+      "grad_norm": 0.3387201130390167,
+      "learning_rate": 0.0005089195804195804,
+      "loss": 3.4584,
+      "step": 26150
+    },
+    {
+      "epoch": 7.629623158017357,
+      "grad_norm": 0.3117615282535553,
+      "learning_rate": 0.0005087447552447552,
+      "loss": 3.4548,
+      "step": 26200
+    },
+    {
+      "epoch": 7.644184285631079,
+      "grad_norm": 0.3377816379070282,
+      "learning_rate": 0.00050856993006993,
+      "loss": 3.456,
+      "step": 26250
+    },
+    {
+      "epoch": 7.658745413244802,
+      "grad_norm": 0.36369189620018005,
+      "learning_rate": 0.0005083951048951048,
+      "loss": 3.4473,
+      "step": 26300
+    },
+    {
+      "epoch": 7.673306540858524,
+      "grad_norm": 0.3553776741027832,
+      "learning_rate": 0.0005082202797202797,
+      "loss": 3.4433,
+      "step": 26350
+    },
+    {
+      "epoch": 7.687867668472246,
+      "grad_norm": 0.33874639868736267,
+      "learning_rate": 0.0005080454545454545,
+      "loss": 3.4423,
+      "step": 26400
+    },
+    {
+      "epoch": 7.702428796085969,
+      "grad_norm": 0.33977967500686646,
+      "learning_rate": 0.0005078706293706293,
+      "loss": 3.4613,
+      "step": 26450
+    },
+    {
+      "epoch": 7.716989923699691,
+      "grad_norm": 0.3333614468574524,
+      "learning_rate": 0.0005076958041958042,
+      "loss": 3.4496,
+      "step": 26500
+    },
+    {
+      "epoch": 7.731551051313414,
+      "grad_norm": 0.3278277814388275,
+      "learning_rate": 0.000507520979020979,
+      "loss": 3.4582,
+      "step": 26550
+    },
+    {
+      "epoch": 7.746112178927136,
+      "grad_norm": 0.33077800273895264,
+      "learning_rate": 0.0005073461538461538,
+      "loss": 3.4525,
+      "step": 26600
+    },
+    {
+      "epoch": 7.760673306540858,
+      "grad_norm": 0.3300637900829315,
+      "learning_rate": 0.0005071713286713286,
+      "loss": 3.4449,
+      "step": 26650
+    },
+    {
+      "epoch": 7.775234434154581,
+      "grad_norm": 0.3329484760761261,
+      "learning_rate": 0.0005069965034965035,
+      "loss": 3.4613,
+      "step": 26700
+    },
+    {
+      "epoch": 7.789795561768304,
+      "grad_norm": 0.35463809967041016,
+      "learning_rate": 0.0005068216783216783,
+      "loss": 3.4637,
+      "step": 26750
+    },
+    {
+      "epoch": 7.8043566893820255,
+      "grad_norm": 0.3434412181377411,
+      "learning_rate": 0.0005066468531468531,
+      "loss": 3.4636,
+      "step": 26800
+    },
+    {
+      "epoch": 7.818917816995748,
+      "grad_norm": 0.33126378059387207,
+      "learning_rate": 0.0005064720279720279,
+      "loss": 3.4617,
+      "step": 26850
+    },
+    {
+      "epoch": 7.833478944609471,
+      "grad_norm": 0.3325732350349426,
+      "learning_rate": 0.0005062972027972028,
+      "loss": 3.4551,
+      "step": 26900
+    },
+    {
+      "epoch": 7.848040072223193,
+      "grad_norm": 0.3155645430088043,
+      "learning_rate": 0.0005061223776223775,
+      "loss": 3.4577,
+      "step": 26950
+    },
+    {
+      "epoch": 7.862601199836916,
+      "grad_norm": 0.3395119607448578,
+      "learning_rate": 0.0005059475524475524,
+      "loss": 3.4488,
+      "step": 27000
+    },
+    {
+      "epoch": 7.862601199836916,
+      "eval_accuracy": 0.36595053012506545,
+      "eval_loss": 3.581655263900757,
+      "eval_runtime": 176.4968,
+      "eval_samples_per_second": 94.313,
+      "eval_steps_per_second": 5.898,
+      "step": 27000
+    },
+    {
+      "epoch": 7.8771623274506375,
+      "grad_norm": 0.3462318480014801,
+      "learning_rate": 0.0005057727272727272,
+      "loss": 3.4536,
+      "step": 27050
+    },
+    {
+      "epoch": 7.89172345506436,
+      "grad_norm": 0.34466618299484253,
+      "learning_rate": 0.000505597902097902,
+      "loss": 3.45,
+      "step": 27100
+    },
+    {
+      "epoch": 7.906284582678083,
+      "grad_norm": 0.3201284110546112,
+      "learning_rate": 0.0005054230769230769,
+      "loss": 3.4584,
+      "step": 27150
+    },
+    {
+      "epoch": 7.920845710291805,
+      "grad_norm": 0.32837069034576416,
+      "learning_rate": 0.0005052482517482517,
+      "loss": 3.4479,
+      "step": 27200
+    },
+    {
+      "epoch": 7.935406837905528,
+      "grad_norm": 0.3236668109893799,
+      "learning_rate": 0.0005050734265734265,
+      "loss": 3.4531,
+      "step": 27250
+    },
+    {
+      "epoch": 7.9499679655192494,
+      "grad_norm": 0.34047573804855347,
+      "learning_rate": 0.0005048986013986013,
+      "loss": 3.4721,
+      "step": 27300
+    },
+    {
+      "epoch": 7.964529093132972,
+      "grad_norm": 0.3239133954048157,
+      "learning_rate": 0.0005047237762237762,
+      "loss": 3.4515,
+      "step": 27350
+    },
+    {
+      "epoch": 7.979090220746695,
+      "grad_norm": 0.32450243830680847,
+      "learning_rate": 0.000504548951048951,
+      "loss": 3.4584,
+      "step": 27400
+    },
+    {
+      "epoch": 7.993651348360417,
+      "grad_norm": 0.3248816132545471,
+      "learning_rate": 0.0005043741258741258,
+      "loss": 3.4567,
+      "step": 27450
+    },
+    {
+      "epoch": 8.008154231463685,
+      "grad_norm": 0.32652345299720764,
+      "learning_rate": 0.0005041993006993006,
+      "loss": 3.401,
+      "step": 27500
+    },
+    {
+      "epoch": 8.022715359077408,
+      "grad_norm": 0.3587798774242401,
+      "learning_rate": 0.0005040244755244755,
+      "loss": 3.3408,
+      "step": 27550
+    },
+    {
+      "epoch": 8.037276486691129,
+      "grad_norm": 0.3377417027950287,
+      "learning_rate": 0.0005038496503496503,
+      "loss": 3.3456,
+      "step": 27600
+    },
+    {
+      "epoch": 8.051837614304851,
+      "grad_norm": 0.3346528708934784,
+      "learning_rate": 0.0005036748251748251,
+      "loss": 3.3501,
+      "step": 27650
+    },
+    {
+      "epoch": 8.066398741918574,
+      "grad_norm": 0.35164663195610046,
+      "learning_rate": 0.0005034999999999999,
+      "loss": 3.3577,
+      "step": 27700
+    },
+    {
+      "epoch": 8.080959869532297,
+      "grad_norm": 0.3102585971355438,
+      "learning_rate": 0.0005033251748251747,
+      "loss": 3.375,
+      "step": 27750
+    },
+    {
+      "epoch": 8.09552099714602,
+      "grad_norm": 0.3394184708595276,
+      "learning_rate": 0.0005031503496503496,
+      "loss": 3.3755,
+      "step": 27800
+    },
+    {
+      "epoch": 8.11008212475974,
+      "grad_norm": 0.32756221294403076,
+      "learning_rate": 0.0005029755244755244,
+      "loss": 3.3746,
+      "step": 27850
+    },
+    {
+      "epoch": 8.124643252373463,
+      "grad_norm": 0.3159530758857727,
+      "learning_rate": 0.0005028006993006992,
+      "loss": 3.3724,
+      "step": 27900
+    },
+    {
+      "epoch": 8.139204379987186,
+      "grad_norm": 0.33860981464385986,
+      "learning_rate": 0.000502625874125874,
+      "loss": 3.3789,
+      "step": 27950
+    },
+    {
+      "epoch": 8.153765507600909,
+      "grad_norm": 0.34003502130508423,
+      "learning_rate": 0.000502451048951049,
+      "loss": 3.3653,
+      "step": 28000
+    },
+    {
+      "epoch": 8.153765507600909,
+      "eval_accuracy": 0.36557009752758785,
+      "eval_loss": 3.589177370071411,
+      "eval_runtime": 176.3386,
+      "eval_samples_per_second": 94.398,
+      "eval_steps_per_second": 5.903,
+      "step": 28000
+    },
+    {
+      "epoch": 8.168326635214632,
+      "grad_norm": 0.351193904876709,
+      "learning_rate": 0.0005022762237762237,
+      "loss": 3.3757,
+      "step": 28050
+    },
+    {
+      "epoch": 8.182887762828354,
+      "grad_norm": 0.3602054715156555,
+      "learning_rate": 0.0005021013986013985,
+      "loss": 3.3774,
+      "step": 28100
+    },
+    {
+      "epoch": 8.197448890442075,
+      "grad_norm": 0.31822383403778076,
+      "learning_rate": 0.0005019265734265733,
+      "loss": 3.3891,
+      "step": 28150
+    },
+    {
+      "epoch": 8.212010018055798,
+      "grad_norm": 0.3160878121852875,
+      "learning_rate": 0.0005017517482517483,
+      "loss": 3.3887,
+      "step": 28200
+    },
+    {
+      "epoch": 8.22657114566952,
+      "grad_norm": 0.3426283895969391,
+      "learning_rate": 0.0005015769230769231,
+      "loss": 3.3857,
+      "step": 28250
+    },
+    {
+      "epoch": 8.241132273283243,
+      "grad_norm": 0.3402179777622223,
+      "learning_rate": 0.0005014020979020979,
+      "loss": 3.3947,
+      "step": 28300
+    },
+    {
+      "epoch": 8.255693400896966,
+      "grad_norm": 0.3560255169868469,
+      "learning_rate": 0.0005012272727272727,
+      "loss": 3.3902,
+      "step": 28350
+    },
+    {
+      "epoch": 8.270254528510687,
+      "grad_norm": 0.3526296317577362,
+      "learning_rate": 0.0005010524475524476,
+      "loss": 3.392,
+      "step": 28400
+    },
+    {
+      "epoch": 8.28481565612441,
+      "grad_norm": 0.3429584503173828,
+      "learning_rate": 0.0005008776223776223,
+      "loss": 3.3907,
+      "step": 28450
+    },
+    {
+      "epoch": 8.299376783738133,
+      "grad_norm": 0.3741019070148468,
+      "learning_rate": 0.0005007027972027972,
+      "loss": 3.3867,
+      "step": 28500
+    },
+    {
+      "epoch": 8.313937911351855,
+      "grad_norm": 0.3399593234062195,
+      "learning_rate": 0.000500527972027972,
+      "loss": 3.4004,
+      "step": 28550
+    },
+    {
+      "epoch": 8.328499038965578,
+      "grad_norm": 0.3139590322971344,
+      "learning_rate": 0.0005003531468531468,
+      "loss": 3.4095,
+      "step": 28600
+    },
+    {
+      "epoch": 8.3430601665793,
+      "grad_norm": 0.3386857807636261,
+      "learning_rate": 0.0005001783216783217,
+      "loss": 3.4005,
+      "step": 28650
+    },
+    {
+      "epoch": 8.357621294193022,
+      "grad_norm": 0.34544748067855835,
+      "learning_rate": 0.0005000034965034965,
+      "loss": 3.4143,
+      "step": 28700
+    },
+    {
+      "epoch": 8.372182421806745,
+      "grad_norm": 0.3371305763721466,
+      "learning_rate": 0.0004998286713286713,
+      "loss": 3.4018,
+      "step": 28750
+    },
+    {
+      "epoch": 8.386743549420467,
+      "grad_norm": 0.3295310437679291,
+      "learning_rate": 0.0004996538461538461,
+      "loss": 3.3987,
+      "step": 28800
+    },
+    {
+      "epoch": 8.40130467703419,
+      "grad_norm": 0.3291172385215759,
+      "learning_rate": 0.000499479020979021,
+      "loss": 3.4058,
+      "step": 28850
+    },
+    {
+      "epoch": 8.415865804647911,
+      "grad_norm": 0.3469097912311554,
+      "learning_rate": 0.0004993041958041958,
+      "loss": 3.4172,
+      "step": 28900
+    },
+    {
+      "epoch": 8.430426932261634,
+      "grad_norm": 0.36806628108024597,
+      "learning_rate": 0.0004991293706293706,
+      "loss": 3.4096,
+      "step": 28950
+    },
+    {
+      "epoch": 8.444988059875357,
+      "grad_norm": 0.3716620206832886,
+      "learning_rate": 0.0004989545454545454,
+      "loss": 3.4157,
+      "step": 29000
+    },
+    {
+      "epoch": 8.444988059875357,
+      "eval_accuracy": 0.36656561768687107,
+      "eval_loss": 3.5810399055480957,
+      "eval_runtime": 176.3981,
+      "eval_samples_per_second": 94.366,
+      "eval_steps_per_second": 5.901,
+      "step": 29000
+    },
+    {
+      "epoch": 8.45954918748908,
+      "grad_norm": 0.35446473956108093,
+      "learning_rate": 0.0004987797202797203,
+      "loss": 3.4093,
+      "step": 29050
+    },
+    {
+      "epoch": 8.474110315102802,
+      "grad_norm": 0.3686777353286743,
+      "learning_rate": 0.0004986048951048951,
+      "loss": 3.4094,
+      "step": 29100
+    },
+    {
+      "epoch": 8.488671442716523,
+      "grad_norm": 0.35613036155700684,
+      "learning_rate": 0.0004984300699300699,
+      "loss": 3.3991,
+      "step": 29150
+    },
+    {
+      "epoch": 8.503232570330246,
+      "grad_norm": 0.32252880930900574,
+      "learning_rate": 0.0004982552447552448,
+      "loss": 3.4118,
+      "step": 29200
+    },
+    {
+      "epoch": 8.517793697943969,
+      "grad_norm": 0.3352912962436676,
+      "learning_rate": 0.0004980804195804195,
+      "loss": 3.4036,
+      "step": 29250
+    },
+    {
+      "epoch": 8.532354825557691,
+      "grad_norm": 0.31796717643737793,
+      "learning_rate": 0.0004979055944055944,
+      "loss": 3.4199,
+      "step": 29300
+    },
+    {
+      "epoch": 8.546915953171414,
+      "grad_norm": 0.32738158106803894,
+      "learning_rate": 0.0004977307692307692,
+      "loss": 3.4043,
+      "step": 29350
+    },
+    {
+      "epoch": 8.561477080785137,
+      "grad_norm": 0.33792996406555176,
+      "learning_rate": 0.000497555944055944,
+      "loss": 3.419,
+      "step": 29400
+    },
+    {
+      "epoch": 8.576038208398858,
+      "grad_norm": 0.3542904555797577,
+      "learning_rate": 0.0004973811188811188,
+      "loss": 3.4204,
+      "step": 29450
+    },
+    {
+      "epoch": 8.59059933601258,
+      "grad_norm": 0.3942735195159912,
+      "learning_rate": 0.0004972062937062937,
+      "loss": 3.4245,
+      "step": 29500
+    },
+    {
+      "epoch": 8.605160463626303,
+      "grad_norm": 0.31929633021354675,
+      "learning_rate": 0.0004970314685314685,
+      "loss": 3.4133,
+      "step": 29550
+    },
+    {
+      "epoch": 8.619721591240026,
+      "grad_norm": 0.3423312306404114,
+      "learning_rate": 0.0004968566433566433,
+      "loss": 3.4228,
+      "step": 29600
+    },
+    {
+      "epoch": 8.634282718853749,
+      "grad_norm": 0.3417450189590454,
+      "learning_rate": 0.0004966818181818181,
+      "loss": 3.4134,
+      "step": 29650
+    },
+    {
+      "epoch": 8.64884384646747,
+      "grad_norm": 0.3385900855064392,
+      "learning_rate": 0.000496506993006993,
+      "loss": 3.4247,
+      "step": 29700
+    },
+    {
+      "epoch": 8.663404974081192,
+      "grad_norm": 0.32018157839775085,
+      "learning_rate": 0.0004963321678321678,
+      "loss": 3.3997,
+      "step": 29750
+    },
+    {
+      "epoch": 8.677966101694915,
+      "grad_norm": 0.34629449248313904,
+      "learning_rate": 0.0004961573426573426,
+      "loss": 3.4207,
+      "step": 29800
+    },
+    {
+      "epoch": 8.692527229308638,
+      "grad_norm": 0.32567837834358215,
+      "learning_rate": 0.0004959825174825175,
+      "loss": 3.4374,
+      "step": 29850
+    },
+    {
+      "epoch": 8.70708835692236,
+      "grad_norm": 0.334450900554657,
+      "learning_rate": 0.0004958076923076923,
+      "loss": 3.4314,
+      "step": 29900
+    },
+    {
+      "epoch": 8.721649484536082,
+      "grad_norm": 0.3887094557285309,
+      "learning_rate": 0.0004956328671328671,
+      "loss": 3.4253,
+      "step": 29950
+    },
+    {
+      "epoch": 8.736210612149804,
+      "grad_norm": 0.3402259051799774,
+      "learning_rate": 0.0004954580419580419,
+      "loss": 3.4305,
+      "step": 30000
+    },
+    {
+      "epoch": 8.736210612149804,
+      "eval_accuracy": 0.36701117996883653,
+      "eval_loss": 3.574934244155884,
+      "eval_runtime": 176.9658,
+      "eval_samples_per_second": 94.063,
+      "eval_steps_per_second": 5.882,
+      "step": 30000
+    },
+    {
+      "epoch": 8.750771739763527,
+      "grad_norm": 0.3436213731765747,
+      "learning_rate": 0.0004952832167832167,
+      "loss": 3.4177,
+      "step": 30050
+    },
+    {
+      "epoch": 8.76533286737725,
+      "grad_norm": 0.328987717628479,
+      "learning_rate": 0.0004951083916083915,
+      "loss": 3.4336,
+      "step": 30100
+    },
+    {
+      "epoch": 8.779893994990973,
+      "grad_norm": 0.3385300934314728,
+      "learning_rate": 0.0004949335664335664,
+      "loss": 3.4331,
+      "step": 30150
+    },
+    {
+      "epoch": 8.794455122604695,
+      "grad_norm": 0.33109623193740845,
+      "learning_rate": 0.0004947587412587412,
+      "loss": 3.4107,
+      "step": 30200
+    },
+    {
+      "epoch": 8.809016250218416,
+      "grad_norm": 0.31443682312965393,
+      "learning_rate": 0.000494583916083916,
+      "loss": 3.43,
+      "step": 30250
+    },
+    {
+      "epoch": 8.82357737783214,
+      "grad_norm": 0.29897189140319824,
+      "learning_rate": 0.0004944090909090908,
+      "loss": 3.444,
+      "step": 30300
+    },
+    {
+      "epoch": 8.838138505445862,
+      "grad_norm": 0.34077444672584534,
+      "learning_rate": 0.0004942342657342657,
+      "loss": 3.4399,
+      "step": 30350
+    },
+    {
+      "epoch": 8.852699633059585,
+      "grad_norm": 0.3268566131591797,
+      "learning_rate": 0.0004940594405594405,
+      "loss": 3.4268,
+      "step": 30400
+    },
+    {
+      "epoch": 8.867260760673307,
+      "grad_norm": 0.32913246750831604,
+      "learning_rate": 0.0004938846153846153,
+      "loss": 3.433,
+      "step": 30450
+    },
+    {
+      "epoch": 8.881821888287028,
+      "grad_norm": 0.31822359561920166,
+      "learning_rate": 0.0004937097902097901,
+      "loss": 3.4374,
+      "step": 30500
+    },
+    {
+      "epoch": 8.896383015900751,
+      "grad_norm": 0.34571900963783264,
+      "learning_rate": 0.000493534965034965,
+      "loss": 3.4361,
+      "step": 30550
+    },
+    {
+      "epoch": 8.910944143514474,
+      "grad_norm": 0.3010501563549042,
+      "learning_rate": 0.0004933601398601398,
+      "loss": 3.4355,
+      "step": 30600
+    },
+    {
+      "epoch": 8.925505271128197,
+      "grad_norm": 0.3567883372306824,
+      "learning_rate": 0.0004931853146853146,
+      "loss": 3.4244,
+      "step": 30650
+    },
+    {
+      "epoch": 8.94006639874192,
+      "grad_norm": 0.32925522327423096,
+      "learning_rate": 0.0004930104895104895,
+      "loss": 3.4203,
+      "step": 30700
+    },
+    {
+      "epoch": 8.95462752635564,
+      "grad_norm": 0.33393311500549316,
+      "learning_rate": 0.0004928356643356642,
+      "loss": 3.4295,
+      "step": 30750
+    },
+    {
+      "epoch": 8.969188653969363,
+      "grad_norm": 0.34621018171310425,
+      "learning_rate": 0.0004926608391608391,
+      "loss": 3.437,
+      "step": 30800
+    },
+    {
+      "epoch": 8.983749781583086,
+      "grad_norm": 0.359050989151001,
+      "learning_rate": 0.0004924860139860139,
+      "loss": 3.4432,
+      "step": 30850
+    },
+    {
+      "epoch": 8.998310909196809,
+      "grad_norm": 0.329822838306427,
+      "learning_rate": 0.0004923111888111887,
+      "loss": 3.4229,
+      "step": 30900
+    },
+    {
+      "epoch": 9.012813792300076,
+      "grad_norm": 0.34674400091171265,
+      "learning_rate": 0.0004921363636363635,
+      "loss": 3.3361,
+      "step": 30950
+    },
+    {
+      "epoch": 9.027374919913798,
+      "grad_norm": 0.33987441658973694,
+      "learning_rate": 0.0004919615384615384,
+      "loss": 3.3188,
+      "step": 31000
+    },
+    {
+      "epoch": 9.027374919913798,
+      "eval_accuracy": 0.3674608569420602,
+      "eval_loss": 3.5798559188842773,
+      "eval_runtime": 176.6125,
+      "eval_samples_per_second": 94.252,
+      "eval_steps_per_second": 5.894,
+      "step": 31000
+    },
+    {
+      "epoch": 9.041936047527521,
+      "grad_norm": 0.33782774209976196,
+      "learning_rate": 0.0004917867132867132,
+      "loss": 3.327,
+      "step": 31050
+    },
+    {
+      "epoch": 9.056497175141242,
+      "grad_norm": 0.3608669340610504,
+      "learning_rate": 0.000491611888111888,
+      "loss": 3.3257,
+      "step": 31100
+    },
+    {
+      "epoch": 9.071058302754965,
+      "grad_norm": 0.33250337839126587,
+      "learning_rate": 0.0004914370629370628,
+      "loss": 3.3302,
+      "step": 31150
+    },
+    {
+      "epoch": 9.085619430368688,
+      "grad_norm": 0.380667120218277,
+      "learning_rate": 0.0004912622377622378,
+      "loss": 3.3367,
+      "step": 31200
+    },
+    {
+      "epoch": 9.10018055798241,
+      "grad_norm": 0.3177809715270996,
+      "learning_rate": 0.0004910874125874126,
+      "loss": 3.341,
+      "step": 31250
+    },
+    {
+      "epoch": 9.114741685596133,
+      "grad_norm": 0.3426327109336853,
+      "learning_rate": 0.0004909125874125874,
+      "loss": 3.3328,
+      "step": 31300
+    },
+    {
+      "epoch": 9.129302813209854,
+      "grad_norm": 0.3812158703804016,
+      "learning_rate": 0.0004907377622377623,
+      "loss": 3.3401,
+      "step": 31350
+    },
+    {
+      "epoch": 9.143863940823577,
+      "grad_norm": 0.3903525471687317,
+      "learning_rate": 0.0004905629370629371,
+      "loss": 3.358,
+      "step": 31400
+    },
+    {
+      "epoch": 9.1584250684373,
+      "grad_norm": 0.3273092210292816,
+      "learning_rate": 0.0004903881118881119,
+      "loss": 3.3457,
+      "step": 31450
+    },
+    {
+      "epoch": 9.172986196051022,
+      "grad_norm": 0.34901750087738037,
+      "learning_rate": 0.0004902132867132867,
+      "loss": 3.3489,
+      "step": 31500
+    },
+    {
+      "epoch": 9.187547323664745,
+      "grad_norm": 0.3518262803554535,
+      "learning_rate": 0.0004900384615384615,
+      "loss": 3.3524,
+      "step": 31550
+    },
+    {
+      "epoch": 9.202108451278466,
+      "grad_norm": 0.34695637226104736,
+      "learning_rate": 0.0004898636363636363,
+      "loss": 3.3472,
+      "step": 31600
+    },
+    {
+      "epoch": 9.216669578892189,
+      "grad_norm": 0.330387145280838,
+      "learning_rate": 0.0004896888111888112,
+      "loss": 3.3538,
+      "step": 31650
+    },
+    {
+      "epoch": 9.231230706505912,
+      "grad_norm": 0.36280760169029236,
+      "learning_rate": 0.000489513986013986,
+      "loss": 3.3641,
+      "step": 31700
+    },
+    {
+      "epoch": 9.245791834119634,
+      "grad_norm": 0.3654526472091675,
+      "learning_rate": 0.0004893391608391608,
+      "loss": 3.3713,
+      "step": 31750
+    },
+    {
+      "epoch": 9.260352961733357,
+      "grad_norm": 0.3399411737918854,
+      "learning_rate": 0.0004891643356643356,
+      "loss": 3.3678,
+      "step": 31800
+    },
+    {
+      "epoch": 9.27491408934708,
+      "grad_norm": 0.3387416899204254,
+      "learning_rate": 0.0004889895104895105,
+      "loss": 3.3738,
+      "step": 31850
+    },
+    {
+      "epoch": 9.2894752169608,
+      "grad_norm": 0.3442247807979584,
+      "learning_rate": 0.0004888146853146853,
+      "loss": 3.3624,
+      "step": 31900
+    },
+    {
+      "epoch": 9.304036344574524,
+      "grad_norm": 0.3599652647972107,
+      "learning_rate": 0.0004886398601398601,
+      "loss": 3.3695,
+      "step": 31950
+    },
+    {
+      "epoch": 9.318597472188246,
+      "grad_norm": 0.33344852924346924,
+      "learning_rate": 0.000488465034965035,
+      "loss": 3.367,
+      "step": 32000
+    },
+    {
+      "epoch": 9.318597472188246,
+      "eval_accuracy": 0.3671916385711629,
+      "eval_loss": 3.579378604888916,
+      "eval_runtime": 176.8517,
+      "eval_samples_per_second": 94.124,
+      "eval_steps_per_second": 5.886,
+      "step": 32000
+    },
+    {
+      "epoch": 9.333158599801969,
+      "grad_norm": 0.36125627160072327,
+      "learning_rate": 0.0004882902097902098,
+      "loss": 3.3745,
+      "step": 32050
+    },
+    {
+      "epoch": 9.347719727415692,
+      "grad_norm": 0.3284973204135895,
+      "learning_rate": 0.0004881153846153846,
+      "loss": 3.3784,
+      "step": 32100
+    },
+    {
+      "epoch": 9.362280855029413,
+      "grad_norm": 0.3623260259628296,
+      "learning_rate": 0.0004879405594405594,
+      "loss": 3.3884,
+      "step": 32150
+    },
+    {
+      "epoch": 9.376841982643136,
+      "grad_norm": 0.35829871892929077,
+      "learning_rate": 0.00048776573426573424,
+      "loss": 3.3828,
+      "step": 32200
+    },
+    {
+      "epoch": 9.391403110256858,
+      "grad_norm": 0.35492923855781555,
+      "learning_rate": 0.00048759090909090904,
+      "loss": 3.389,
+      "step": 32250
+    },
+    {
+      "epoch": 9.405964237870581,
+      "grad_norm": 0.38198089599609375,
+      "learning_rate": 0.0004874160839160839,
+      "loss": 3.3864,
+      "step": 32300
+    },
+    {
+      "epoch": 9.420525365484304,
+      "grad_norm": 0.3594827651977539,
+      "learning_rate": 0.0004872412587412587,
+      "loss": 3.3896,
+      "step": 32350
+    },
+    {
+      "epoch": 9.435086493098025,
+      "grad_norm": 0.3426271975040436,
+      "learning_rate": 0.00048706643356643354,
+      "loss": 3.3825,
+      "step": 32400
+    },
+    {
+      "epoch": 9.449647620711747,
+      "grad_norm": 0.3374452590942383,
+      "learning_rate": 0.00048689160839160834,
+      "loss": 3.3773,
+      "step": 32450
+    },
+    {
+      "epoch": 9.46420874832547,
+      "grad_norm": 0.3500727415084839,
+      "learning_rate": 0.0004867167832167832,
+      "loss": 3.3978,
+      "step": 32500
+    },
+    {
+      "epoch": 9.478769875939193,
+      "grad_norm": 0.3206511437892914,
+      "learning_rate": 0.00048654195804195794,
+      "loss": 3.3935,
+      "step": 32550
+    },
+    {
+      "epoch": 9.493331003552916,
+      "grad_norm": 0.3213861882686615,
+      "learning_rate": 0.00048636713286713285,
+      "loss": 3.3972,
+      "step": 32600
+    },
+    {
+      "epoch": 9.507892131166638,
+      "grad_norm": 0.3542587459087372,
+      "learning_rate": 0.0004861923076923077,
+      "loss": 3.3891,
+      "step": 32650
+    },
+    {
+      "epoch": 9.52245325878036,
+      "grad_norm": 0.35178741812705994,
+      "learning_rate": 0.00048601748251748245,
+      "loss": 3.3902,
+      "step": 32700
+    },
+    {
+      "epoch": 9.537014386394082,
+      "grad_norm": 0.3213842511177063,
+      "learning_rate": 0.0004858426573426573,
+      "loss": 3.3894,
+      "step": 32750
+    },
+    {
+      "epoch": 9.551575514007805,
+      "grad_norm": 0.3286217451095581,
+      "learning_rate": 0.0004856678321678321,
+      "loss": 3.3926,
+      "step": 32800
+    },
+    {
+      "epoch": 9.566136641621528,
+      "grad_norm": 0.3548128008842468,
+      "learning_rate": 0.00048549300699300696,
+      "loss": 3.3883,
+      "step": 32850
+    },
+    {
+      "epoch": 9.58069776923525,
+      "grad_norm": 0.3568916618824005,
+      "learning_rate": 0.00048531818181818176,
+      "loss": 3.4017,
+      "step": 32900
+    },
+    {
+      "epoch": 9.595258896848971,
+      "grad_norm": 0.3281558156013489,
+      "learning_rate": 0.0004851433566433566,
+      "loss": 3.4038,
+      "step": 32950
+    },
+    {
+      "epoch": 9.609820024462694,
+      "grad_norm": 0.35022494196891785,
+      "learning_rate": 0.0004849685314685314,
+      "loss": 3.4041,
+      "step": 33000
+    },
+    {
+      "epoch": 9.609820024462694,
+      "eval_accuracy": 0.3675565529044665,
+      "eval_loss": 3.573331594467163,
+      "eval_runtime": 177.4614,
+      "eval_samples_per_second": 93.801,
+      "eval_steps_per_second": 5.866,
+      "step": 33000
+    },
+    {
+      "epoch": 9.624381152076417,
+      "grad_norm": 0.34615230560302734,
+      "learning_rate": 0.00048479370629370627,
+      "loss": 3.3877,
+      "step": 33050
+    },
+    {
+      "epoch": 9.63894227969014,
+      "grad_norm": 0.3371587097644806,
+      "learning_rate": 0.00048461888111888106,
+      "loss": 3.3997,
+      "step": 33100
+    },
+    {
+      "epoch": 9.653503407303862,
+      "grad_norm": 0.32450369000434875,
+      "learning_rate": 0.0004844440559440559,
+      "loss": 3.3985,
+      "step": 33150
+    },
+    {
+      "epoch": 9.668064534917583,
+      "grad_norm": 0.3822547495365143,
+      "learning_rate": 0.0004842692307692307,
+      "loss": 3.396,
+      "step": 33200
+    },
+    {
+      "epoch": 9.682625662531306,
+      "grad_norm": 0.3404769003391266,
+      "learning_rate": 0.00048409440559440557,
+      "loss": 3.4074,
+      "step": 33250
+    },
+    {
+      "epoch": 9.697186790145029,
+      "grad_norm": 0.35298749804496765,
+      "learning_rate": 0.0004839195804195803,
+      "loss": 3.4048,
+      "step": 33300
+    },
+    {
+      "epoch": 9.711747917758752,
+      "grad_norm": 0.346872478723526,
+      "learning_rate": 0.0004837447552447552,
+      "loss": 3.3976,
+      "step": 33350
+    },
+    {
+      "epoch": 9.726309045372474,
+      "grad_norm": 0.32681095600128174,
+      "learning_rate": 0.0004835699300699301,
+      "loss": 3.4035,
+      "step": 33400
+    },
+    {
+      "epoch": 9.740870172986195,
+      "grad_norm": 0.3425065279006958,
+      "learning_rate": 0.0004833951048951048,
+      "loss": 3.3915,
+      "step": 33450
+    },
+    {
+      "epoch": 9.755431300599918,
+      "grad_norm": 0.33771848678588867,
+      "learning_rate": 0.0004832202797202797,
+      "loss": 3.3985,
+      "step": 33500
+    },
+    {
+      "epoch": 9.76999242821364,
+      "grad_norm": 0.3263910412788391,
+      "learning_rate": 0.0004830454545454545,
+      "loss": 3.4014,
+      "step": 33550
+    },
+    {
+      "epoch": 9.784553555827364,
+      "grad_norm": 0.332059770822525,
+      "learning_rate": 0.00048287062937062933,
+      "loss": 3.3994,
+      "step": 33600
+    },
+    {
+      "epoch": 9.799114683441086,
+      "grad_norm": 0.35489657521247864,
+      "learning_rate": 0.00048269580419580413,
+      "loss": 3.4078,
+      "step": 33650
+    },
+    {
+      "epoch": 9.813675811054807,
+      "grad_norm": 0.35381078720092773,
+      "learning_rate": 0.000482520979020979,
+      "loss": 3.4005,
+      "step": 33700
+    },
+    {
+      "epoch": 9.82823693866853,
+      "grad_norm": 0.324863463640213,
+      "learning_rate": 0.0004823461538461538,
+      "loss": 3.3952,
+      "step": 33750
+    },
+    {
+      "epoch": 9.842798066282253,
+      "grad_norm": 0.3365383446216583,
+      "learning_rate": 0.00048217132867132864,
+      "loss": 3.4049,
+      "step": 33800
+    },
+    {
+      "epoch": 9.857359193895975,
+      "grad_norm": 0.34103065729141235,
+      "learning_rate": 0.00048199650349650344,
+      "loss": 3.4089,
+      "step": 33850
+    },
+    {
+      "epoch": 9.871920321509698,
+      "grad_norm": 0.3343670070171356,
+      "learning_rate": 0.0004818216783216783,
+      "loss": 3.4227,
+      "step": 33900
+    },
+    {
+      "epoch": 9.88648144912342,
+      "grad_norm": 0.32671719789505005,
+      "learning_rate": 0.0004816468531468531,
+      "loss": 3.4102,
+      "step": 33950
+    },
+    {
+      "epoch": 9.901042576737142,
+      "grad_norm": 0.36884456872940063,
+      "learning_rate": 0.00048147202797202795,
+      "loss": 3.4055,
+      "step": 34000
+    },
+    {
+      "epoch": 9.901042576737142,
+      "eval_accuracy": 0.368497524013926,
+      "eval_loss": 3.564042329788208,
+      "eval_runtime": 176.835,
+      "eval_samples_per_second": 94.133,
+      "eval_steps_per_second": 5.887,
+      "step": 34000
+    },
+    {
+      "epoch": 9.915603704350865,
+      "grad_norm": 0.3321670591831207,
+      "learning_rate": 0.0004812972027972028,
+      "loss": 3.3949,
+      "step": 34050
+    },
+    {
+      "epoch": 9.930164831964587,
+      "grad_norm": 0.36186736822128296,
+      "learning_rate": 0.0004811223776223776,
+      "loss": 3.4091,
+      "step": 34100
+    },
+    {
+      "epoch": 9.94472595957831,
+      "grad_norm": 0.3428935408592224,
+      "learning_rate": 0.00048094755244755245,
+      "loss": 3.4074,
+      "step": 34150
+    },
+    {
+      "epoch": 9.959287087192033,
+      "grad_norm": 0.35069432854652405,
+      "learning_rate": 0.0004807727272727272,
+      "loss": 3.3979,
+      "step": 34200
+    },
+    {
+      "epoch": 9.973848214805754,
+      "grad_norm": 0.32886582612991333,
+      "learning_rate": 0.00048059790209790205,
+      "loss": 3.4107,
+      "step": 34250
+    },
+    {
+      "epoch": 9.988409342419477,
+      "grad_norm": 0.3682115972042084,
+      "learning_rate": 0.00048042307692307685,
+      "loss": 3.4094,
+      "step": 34300
+    },
+    {
+      "epoch": 10.002912225522744,
+      "grad_norm": 0.35744708776474,
+      "learning_rate": 0.0004802482517482517,
+      "loss": 3.3896,
+      "step": 34350
+    },
+    {
+      "epoch": 10.017473353136467,
+      "grad_norm": 0.35240817070007324,
+      "learning_rate": 0.0004800734265734265,
+      "loss": 3.3044,
+      "step": 34400
+    },
+    {
+      "epoch": 10.03203448075019,
+      "grad_norm": 0.34491005539894104,
+      "learning_rate": 0.00047989860139860136,
+      "loss": 3.2982,
+      "step": 34450
+    },
+    {
+      "epoch": 10.046595608363912,
+      "grad_norm": 0.33637934923171997,
+      "learning_rate": 0.00047972377622377616,
+      "loss": 3.3106,
+      "step": 34500
+    },
+    {
+      "epoch": 10.061156735977635,
+      "grad_norm": 0.3516424894332886,
+      "learning_rate": 0.000479548951048951,
+      "loss": 3.3193,
+      "step": 34550
+    },
+    {
+      "epoch": 10.075717863591356,
+      "grad_norm": 0.3707565665245056,
+      "learning_rate": 0.0004793741258741258,
+      "loss": 3.3098,
+      "step": 34600
+    },
+    {
+      "epoch": 10.090278991205079,
+      "grad_norm": 0.33687910437583923,
+      "learning_rate": 0.00047919930069930067,
+      "loss": 3.3132,
+      "step": 34650
+    },
+    {
+      "epoch": 10.104840118818801,
+      "grad_norm": 0.353137731552124,
+      "learning_rate": 0.0004790244755244755,
+      "loss": 3.3231,
+      "step": 34700
+    },
+    {
+      "epoch": 10.119401246432524,
+      "grad_norm": 0.36714377999305725,
+      "learning_rate": 0.0004788496503496503,
+      "loss": 3.324,
+      "step": 34750
+    },
+    {
+      "epoch": 10.133962374046247,
+      "grad_norm": 0.32502633333206177,
+      "learning_rate": 0.0004786748251748252,
+      "loss": 3.3239,
+      "step": 34800
+    },
+    {
+      "epoch": 10.148523501659968,
+      "grad_norm": 0.33693206310272217,
+      "learning_rate": 0.0004785,
+      "loss": 3.3299,
+      "step": 34850
+    },
+    {
+      "epoch": 10.16308462927369,
+      "grad_norm": 0.3602798581123352,
+      "learning_rate": 0.00047832517482517483,
+      "loss": 3.3361,
+      "step": 34900
+    },
+    {
+      "epoch": 10.177645756887413,
+      "grad_norm": 0.3324842154979706,
+      "learning_rate": 0.0004781503496503496,
+      "loss": 3.3397,
+      "step": 34950
+    },
+    {
+      "epoch": 10.192206884501136,
+      "grad_norm": 0.338860422372818,
+      "learning_rate": 0.00047797552447552443,
+      "loss": 3.3366,
+      "step": 35000
+    },
+    {
+      "epoch": 10.192206884501136,
+      "eval_accuracy": 0.3681085093461097,
+      "eval_loss": 3.572119951248169,
+      "eval_runtime": 176.7411,
+      "eval_samples_per_second": 94.183,
+      "eval_steps_per_second": 5.89,
+      "step": 35000
+    },
+    {
+      "epoch": 10.206768012114859,
+      "grad_norm": 0.37433600425720215,
+      "learning_rate": 0.00047780069930069923,
+      "loss": 3.3329,
+      "step": 35050
+    },
+    {
+      "epoch": 10.221329139728581,
+      "grad_norm": 0.3410218060016632,
+      "learning_rate": 0.0004776258741258741,
+      "loss": 3.3312,
+      "step": 35100
+    },
+    {
+      "epoch": 10.235890267342302,
+      "grad_norm": 0.33754661679267883,
+      "learning_rate": 0.0004774510489510489,
+      "loss": 3.3364,
+      "step": 35150
+    },
+    {
+      "epoch": 10.250451394956025,
+      "grad_norm": 0.3520354628562927,
+      "learning_rate": 0.00047727622377622374,
+      "loss": 3.3463,
+      "step": 35200
+    },
+    {
+      "epoch": 10.265012522569748,
+      "grad_norm": 0.3703749179840088,
+      "learning_rate": 0.00047710139860139854,
+      "loss": 3.352,
+      "step": 35250
+    },
+    {
+      "epoch": 10.27957365018347,
+      "grad_norm": 0.3496933877468109,
+      "learning_rate": 0.0004769265734265734,
+      "loss": 3.3469,
+      "step": 35300
+    },
+    {
+      "epoch": 10.294134777797193,
+      "grad_norm": 0.3573146164417267,
+      "learning_rate": 0.0004767517482517482,
+      "loss": 3.3455,
+      "step": 35350
+    },
+    {
+      "epoch": 10.308695905410914,
+      "grad_norm": 0.34615665674209595,
+      "learning_rate": 0.00047657692307692304,
+      "loss": 3.3578,
+      "step": 35400
+    },
+    {
+      "epoch": 10.323257033024637,
+      "grad_norm": 0.3539220690727234,
+      "learning_rate": 0.0004764020979020979,
+      "loss": 3.3548,
+      "step": 35450
+    },
+    {
+      "epoch": 10.33781816063836,
+      "grad_norm": 0.3664166033267975,
+      "learning_rate": 0.0004762272727272727,
+      "loss": 3.3533,
+      "step": 35500
+    },
+    {
+      "epoch": 10.352379288252083,
+      "grad_norm": 0.35653188824653625,
+      "learning_rate": 0.00047605244755244755,
+      "loss": 3.3557,
+      "step": 35550
+    },
+    {
+      "epoch": 10.366940415865805,
+      "grad_norm": 0.3385823965072632,
+      "learning_rate": 0.00047587762237762235,
+      "loss": 3.3517,
+      "step": 35600
+    },
+    {
+      "epoch": 10.381501543479526,
+      "grad_norm": 0.36101484298706055,
+      "learning_rate": 0.0004757027972027972,
+      "loss": 3.3607,
+      "step": 35650
+    },
+    {
+      "epoch": 10.396062671093249,
+      "grad_norm": 0.34660804271698,
+      "learning_rate": 0.00047552797202797195,
+      "loss": 3.3548,
+      "step": 35700
+    },
+    {
+      "epoch": 10.410623798706972,
+      "grad_norm": 0.3321472704410553,
+      "learning_rate": 0.0004753531468531468,
+      "loss": 3.3513,
+      "step": 35750
+    },
+    {
+      "epoch": 10.425184926320695,
+      "grad_norm": 0.3524532616138458,
+      "learning_rate": 0.0004751783216783216,
+      "loss": 3.3542,
+      "step": 35800
+    },
+    {
+      "epoch": 10.439746053934417,
+      "grad_norm": 0.33926790952682495,
+      "learning_rate": 0.00047500349650349646,
+      "loss": 3.3685,
+      "step": 35850
+    },
+    {
+      "epoch": 10.454307181548138,
+      "grad_norm": 0.3707771599292755,
+      "learning_rate": 0.00047482867132867126,
+      "loss": 3.3754,
+      "step": 35900
+    },
+    {
+      "epoch": 10.468868309161861,
+      "grad_norm": 0.36390265822410583,
+      "learning_rate": 0.0004746538461538461,
+      "loss": 3.3671,
+      "step": 35950
+    },
+    {
+      "epoch": 10.483429436775584,
+      "grad_norm": 0.32835954427719116,
+      "learning_rate": 0.0004744790209790209,
+      "loss": 3.3638,
+      "step": 36000
+    },
+    {
+      "epoch": 10.483429436775584,
+      "eval_accuracy": 0.36857488020958123,
+      "eval_loss": 3.5680227279663086,
+      "eval_runtime": 176.8143,
+      "eval_samples_per_second": 94.144,
+      "eval_steps_per_second": 5.888,
+      "step": 36000
+    },
+    {
+      "epoch": 10.497990564389307,
+      "grad_norm": 0.36081641912460327,
+      "learning_rate": 0.00047430419580419576,
+      "loss": 3.3636,
+      "step": 36050
+    },
+    {
+      "epoch": 10.51255169200303,
+      "grad_norm": 0.3701050579547882,
+      "learning_rate": 0.0004741293706293706,
+      "loss": 3.3462,
+      "step": 36100
+    },
+    {
+      "epoch": 10.52711281961675,
+      "grad_norm": 0.3523276150226593,
+      "learning_rate": 0.0004739545454545454,
+      "loss": 3.3713,
+      "step": 36150
+    },
+    {
+      "epoch": 10.541673947230473,
+      "grad_norm": 0.34822702407836914,
+      "learning_rate": 0.00047377972027972027,
+      "loss": 3.3808,
+      "step": 36200
+    },
+    {
+      "epoch": 10.556235074844196,
+      "grad_norm": 0.34755954146385193,
+      "learning_rate": 0.00047360489510489507,
+      "loss": 3.3718,
+      "step": 36250
+    },
+    {
+      "epoch": 10.570796202457919,
+      "grad_norm": 0.3505362868309021,
+      "learning_rate": 0.0004734300699300699,
+      "loss": 3.3684,
+      "step": 36300
+    },
+    {
+      "epoch": 10.585357330071641,
+      "grad_norm": 0.36556077003479004,
+      "learning_rate": 0.0004732552447552447,
+      "loss": 3.3585,
+      "step": 36350
+    },
+    {
+      "epoch": 10.599918457685362,
+      "grad_norm": 0.36185914278030396,
+      "learning_rate": 0.0004730804195804196,
+      "loss": 3.3745,
+      "step": 36400
+    },
+    {
+      "epoch": 10.614479585299085,
+      "grad_norm": 0.3472917973995209,
+      "learning_rate": 0.0004729055944055943,
+      "loss": 3.369,
+      "step": 36450
+    },
+    {
+      "epoch": 10.629040712912808,
+      "grad_norm": 0.3542565405368805,
+      "learning_rate": 0.0004727307692307692,
+      "loss": 3.377,
+      "step": 36500
+    },
+    {
+      "epoch": 10.64360184052653,
+      "grad_norm": 0.3553348779678345,
+      "learning_rate": 0.000472555944055944,
+      "loss": 3.3614,
+      "step": 36550
+    },
+    {
+      "epoch": 10.658162968140253,
+      "grad_norm": 0.3704865872859955,
+      "learning_rate": 0.00047238111888111883,
+      "loss": 3.3763,
+      "step": 36600
+    },
+    {
+      "epoch": 10.672724095753976,
+      "grad_norm": 0.3559943437576294,
+      "learning_rate": 0.00047220629370629363,
+      "loss": 3.3715,
+      "step": 36650
+    },
+    {
+      "epoch": 10.687285223367697,
+      "grad_norm": 0.34045952558517456,
+      "learning_rate": 0.0004720314685314685,
+      "loss": 3.3823,
+      "step": 36700
+    },
+    {
+      "epoch": 10.70184635098142,
+      "grad_norm": 0.3242470622062683,
+      "learning_rate": 0.0004718566433566433,
+      "loss": 3.3757,
+      "step": 36750
+    },
+    {
+      "epoch": 10.716407478595142,
+      "grad_norm": 0.35755789279937744,
+      "learning_rate": 0.00047168181818181814,
+      "loss": 3.3811,
+      "step": 36800
+    },
+    {
+      "epoch": 10.730968606208865,
+      "grad_norm": 0.37564322352409363,
+      "learning_rate": 0.000471506993006993,
+      "loss": 3.3723,
+      "step": 36850
+    },
+    {
+      "epoch": 10.745529733822588,
+      "grad_norm": 0.3541555404663086,
+      "learning_rate": 0.0004713321678321678,
+      "loss": 3.373,
+      "step": 36900
+    },
+    {
+      "epoch": 10.760090861436309,
+      "grad_norm": 0.3599989414215088,
+      "learning_rate": 0.00047115734265734265,
+      "loss": 3.3864,
+      "step": 36950
+    },
+    {
+      "epoch": 10.774651989050032,
+      "grad_norm": 0.35538411140441895,
+      "learning_rate": 0.00047098251748251745,
+      "loss": 3.3768,
+      "step": 37000
+    },
+    {
+      "epoch": 10.774651989050032,
+      "eval_accuracy": 0.3687768527690579,
+      "eval_loss": 3.5605952739715576,
+      "eval_runtime": 176.8123,
+      "eval_samples_per_second": 94.145,
+      "eval_steps_per_second": 5.888,
+      "step": 37000
+    },
+    {
+      "epoch": 10.789213116663754,
+      "grad_norm": 0.32753708958625793,
+      "learning_rate": 0.0004708076923076923,
+      "loss": 3.3872,
+      "step": 37050
+    },
+    {
+      "epoch": 10.803774244277477,
+      "grad_norm": 0.3197953402996063,
+      "learning_rate": 0.0004706328671328671,
+      "loss": 3.3832,
+      "step": 37100
+    },
+    {
+      "epoch": 10.8183353718912,
+      "grad_norm": 0.33160483837127686,
+      "learning_rate": 0.00047045804195804195,
+      "loss": 3.375,
+      "step": 37150
+    },
+    {
+      "epoch": 10.83289649950492,
+      "grad_norm": 0.36038827896118164,
+      "learning_rate": 0.0004702832167832167,
+      "loss": 3.3957,
+      "step": 37200
+    },
+    {
+      "epoch": 10.847457627118644,
+      "grad_norm": 0.33877766132354736,
+      "learning_rate": 0.00047010839160839155,
+      "loss": 3.3873,
+      "step": 37250
+    },
+    {
+      "epoch": 10.862018754732366,
+      "grad_norm": 0.34722471237182617,
+      "learning_rate": 0.00046993356643356635,
+      "loss": 3.3901,
+      "step": 37300
+    },
+    {
+      "epoch": 10.876579882346089,
+      "grad_norm": 0.3494495153427124,
+      "learning_rate": 0.0004697587412587412,
+      "loss": 3.3954,
+      "step": 37350
+    },
+    {
+      "epoch": 10.891141009959812,
+      "grad_norm": 0.34382206201553345,
+      "learning_rate": 0.000469583916083916,
+      "loss": 3.3816,
+      "step": 37400
+    },
+    {
+      "epoch": 10.905702137573535,
+      "grad_norm": 0.33641430735588074,
+      "learning_rate": 0.00046940909090909086,
+      "loss": 3.3772,
+      "step": 37450
+    },
+    {
+      "epoch": 10.920263265187256,
+      "grad_norm": 0.3528263568878174,
+      "learning_rate": 0.0004692342657342657,
+      "loss": 3.3876,
+      "step": 37500
+    },
+    {
+      "epoch": 10.934824392800978,
+      "grad_norm": 0.34849458932876587,
+      "learning_rate": 0.0004690594405594405,
+      "loss": 3.3914,
+      "step": 37550
+    },
+    {
+      "epoch": 10.949385520414701,
+      "grad_norm": 0.34705743193626404,
+      "learning_rate": 0.00046888461538461537,
+      "loss": 3.3793,
+      "step": 37600
+    },
+    {
+      "epoch": 10.963946648028424,
+      "grad_norm": 0.3272669017314911,
+      "learning_rate": 0.00046870979020979017,
+      "loss": 3.3788,
+      "step": 37650
+    },
+    {
+      "epoch": 10.978507775642147,
+      "grad_norm": 0.33112722635269165,
+      "learning_rate": 0.000468534965034965,
+      "loss": 3.3837,
+      "step": 37700
+    },
+    {
+      "epoch": 10.993068903255867,
+      "grad_norm": 0.3313654363155365,
+      "learning_rate": 0.0004683601398601398,
+      "loss": 3.3966,
+      "step": 37750
+    },
+    {
+      "epoch": 11.007571786359136,
+      "grad_norm": 0.3786119222640991,
+      "learning_rate": 0.0004681853146853147,
+      "loss": 3.3267,
+      "step": 37800
+    },
+    {
+      "epoch": 11.022132913972857,
+      "grad_norm": 0.35445159673690796,
+      "learning_rate": 0.0004680104895104895,
+      "loss": 3.2648,
+      "step": 37850
+    },
+    {
+      "epoch": 11.03669404158658,
+      "grad_norm": 0.3338647782802582,
+      "learning_rate": 0.00046783566433566433,
+      "loss": 3.2768,
+      "step": 37900
+    },
+    {
+      "epoch": 11.051255169200303,
+      "grad_norm": 0.33461859822273254,
+      "learning_rate": 0.0004676608391608391,
+      "loss": 3.2856,
+      "step": 37950
+    },
+    {
+      "epoch": 11.065816296814026,
+      "grad_norm": 0.3517346978187561,
+      "learning_rate": 0.00046748601398601393,
+      "loss": 3.2775,
+      "step": 38000
+    },
+    {
+      "epoch": 11.065816296814026,
+      "eval_accuracy": 0.36888348205395044,
+      "eval_loss": 3.570584297180176,
+      "eval_runtime": 176.7964,
+      "eval_samples_per_second": 94.153,
+      "eval_steps_per_second": 5.888,
+      "step": 38000
+    },
+    {
+      "epoch": 11.080377424427748,
+      "grad_norm": 0.3435273766517639,
+      "learning_rate": 0.00046731118881118873,
+      "loss": 3.2915,
+      "step": 38050
+    },
+    {
+      "epoch": 11.09493855204147,
+      "grad_norm": 0.3325134813785553,
+      "learning_rate": 0.0004671363636363636,
+      "loss": 3.2978,
+      "step": 38100
+    },
+    {
+      "epoch": 11.109499679655192,
+      "grad_norm": 0.36426037549972534,
+      "learning_rate": 0.00046696153846153844,
+      "loss": 3.3033,
+      "step": 38150
+    },
+    {
+      "epoch": 11.124060807268915,
+      "grad_norm": 0.327764093875885,
+      "learning_rate": 0.00046678671328671324,
+      "loss": 3.3043,
+      "step": 38200
+    },
+    {
+      "epoch": 11.138621934882638,
+      "grad_norm": 0.3818574845790863,
+      "learning_rate": 0.0004666118881118881,
+      "loss": 3.3042,
+      "step": 38250
+    },
+    {
+      "epoch": 11.15318306249636,
+      "grad_norm": 0.38499313592910767,
+      "learning_rate": 0.0004664370629370629,
+      "loss": 3.2968,
+      "step": 38300
+    },
+    {
+      "epoch": 11.167744190110081,
+      "grad_norm": 0.3457299768924713,
+      "learning_rate": 0.00046626223776223774,
+      "loss": 3.299,
+      "step": 38350
+    },
+    {
+      "epoch": 11.182305317723804,
+      "grad_norm": 0.37761256098747253,
+      "learning_rate": 0.00046608741258741254,
+      "loss": 3.3019,
+      "step": 38400
+    },
+    {
+      "epoch": 11.196866445337527,
+      "grad_norm": 0.35469600558280945,
+      "learning_rate": 0.0004659125874125874,
+      "loss": 3.3116,
+      "step": 38450
+    },
+    {
+      "epoch": 11.21142757295125,
+      "grad_norm": 0.3749145269393921,
+      "learning_rate": 0.0004657377622377622,
+      "loss": 3.3128,
+      "step": 38500
+    },
+    {
+      "epoch": 11.225988700564972,
+      "grad_norm": 0.38890284299850464,
+      "learning_rate": 0.00046556293706293705,
+      "loss": 3.332,
+      "step": 38550
+    },
+    {
+      "epoch": 11.240549828178693,
+      "grad_norm": 0.3799166679382324,
+      "learning_rate": 0.00046538811188811185,
+      "loss": 3.3111,
+      "step": 38600
+    },
+    {
+      "epoch": 11.255110955792416,
+      "grad_norm": 0.3511936366558075,
+      "learning_rate": 0.0004652132867132867,
+      "loss": 3.3251,
+      "step": 38650
+    },
+    {
+      "epoch": 11.269672083406139,
+      "grad_norm": 0.37755730748176575,
+      "learning_rate": 0.00046503846153846145,
+      "loss": 3.3257,
+      "step": 38700
+    },
+    {
+      "epoch": 11.284233211019862,
+      "grad_norm": 0.34114590287208557,
+      "learning_rate": 0.0004648636363636363,
+      "loss": 3.3212,
+      "step": 38750
+    },
+    {
+      "epoch": 11.298794338633584,
+      "grad_norm": 0.3562406897544861,
+      "learning_rate": 0.0004646888111888111,
+      "loss": 3.3259,
+      "step": 38800
+    },
+    {
+      "epoch": 11.313355466247307,
+      "grad_norm": 0.3757557272911072,
+      "learning_rate": 0.00046451398601398596,
+      "loss": 3.3395,
+      "step": 38850
+    },
+    {
+      "epoch": 11.327916593861028,
+      "grad_norm": 0.3693144619464874,
+      "learning_rate": 0.0004643391608391608,
+      "loss": 3.3458,
+      "step": 38900
+    },
+    {
+      "epoch": 11.34247772147475,
+      "grad_norm": 0.3550606071949005,
+      "learning_rate": 0.0004641643356643356,
+      "loss": 3.336,
+      "step": 38950
+    },
+    {
+      "epoch": 11.357038849088473,
+      "grad_norm": 0.36543989181518555,
+      "learning_rate": 0.00046398951048951046,
+      "loss": 3.331,
+      "step": 39000
+    },
+    {
+      "epoch": 11.357038849088473,
+      "eval_accuracy": 0.3695512376638617,
+      "eval_loss": 3.567185878753662,
+      "eval_runtime": 176.8379,
+      "eval_samples_per_second": 94.131,
+      "eval_steps_per_second": 5.887,
+      "step": 39000
+    },
+    {
+      "epoch": 11.371599976702196,
+      "grad_norm": 0.36690041422843933,
+      "learning_rate": 0.00046381468531468526,
+      "loss": 3.3383,
+      "step": 39050
+    },
+    {
+      "epoch": 11.386161104315919,
+      "grad_norm": 0.3439081013202667,
+      "learning_rate": 0.0004636398601398601,
+      "loss": 3.3272,
+      "step": 39100
+    },
+    {
+      "epoch": 11.40072223192964,
+      "grad_norm": 0.3694179654121399,
+      "learning_rate": 0.0004634650349650349,
+      "loss": 3.337,
+      "step": 39150
+    },
+    {
+      "epoch": 11.415283359543363,
+      "grad_norm": 0.36797547340393066,
+      "learning_rate": 0.00046329020979020977,
+      "loss": 3.3276,
+      "step": 39200
+    },
+    {
+      "epoch": 11.429844487157085,
+      "grad_norm": 0.40091055631637573,
+      "learning_rate": 0.00046311538461538457,
+      "loss": 3.3355,
+      "step": 39250
+    },
+    {
+      "epoch": 11.444405614770808,
+      "grad_norm": 0.34869131445884705,
+      "learning_rate": 0.0004629405594405594,
+      "loss": 3.3352,
+      "step": 39300
+    },
+    {
+      "epoch": 11.458966742384531,
+      "grad_norm": 0.35575979948043823,
+      "learning_rate": 0.0004627657342657342,
+      "loss": 3.3462,
+      "step": 39350
+    },
+    {
+      "epoch": 11.473527869998252,
+      "grad_norm": 0.3693273365497589,
+      "learning_rate": 0.0004625909090909091,
+      "loss": 3.3398,
+      "step": 39400
+    },
+    {
+      "epoch": 11.488088997611975,
+      "grad_norm": 0.3422088623046875,
+      "learning_rate": 0.0004624160839160838,
+      "loss": 3.3537,
+      "step": 39450
+    },
+    {
+      "epoch": 11.502650125225697,
+      "grad_norm": 0.35034725069999695,
+      "learning_rate": 0.0004622412587412587,
+      "loss": 3.3536,
+      "step": 39500
+    },
+    {
+      "epoch": 11.51721125283942,
+      "grad_norm": 0.3721543550491333,
+      "learning_rate": 0.00046206643356643353,
+      "loss": 3.3611,
+      "step": 39550
+    },
+    {
+      "epoch": 11.531772380453143,
+      "grad_norm": 0.3786129653453827,
+      "learning_rate": 0.00046189160839160833,
+      "loss": 3.3612,
+      "step": 39600
+    },
+    {
+      "epoch": 11.546333508066864,
+      "grad_norm": 0.37621766328811646,
+      "learning_rate": 0.0004617167832167832,
+      "loss": 3.3513,
+      "step": 39650
+    },
+    {
+      "epoch": 11.560894635680587,
+      "grad_norm": 0.33818793296813965,
+      "learning_rate": 0.000461541958041958,
+      "loss": 3.3585,
+      "step": 39700
+    },
+    {
+      "epoch": 11.57545576329431,
+      "grad_norm": 0.3617473840713501,
+      "learning_rate": 0.00046136713286713284,
+      "loss": 3.3416,
+      "step": 39750
+    },
+    {
+      "epoch": 11.590016890908032,
+      "grad_norm": 0.36892908811569214,
+      "learning_rate": 0.00046119230769230764,
+      "loss": 3.3564,
+      "step": 39800
+    },
+    {
+      "epoch": 11.604578018521755,
+      "grad_norm": 0.3816312849521637,
+      "learning_rate": 0.0004610174825174825,
+      "loss": 3.3536,
+      "step": 39850
+    },
+    {
+      "epoch": 11.619139146135478,
+      "grad_norm": 0.3597758114337921,
+      "learning_rate": 0.0004608426573426573,
+      "loss": 3.3576,
+      "step": 39900
+    },
+    {
+      "epoch": 11.633700273749199,
+      "grad_norm": 0.3639199733734131,
+      "learning_rate": 0.00046066783216783215,
+      "loss": 3.354,
+      "step": 39950
+    },
+    {
+      "epoch": 11.648261401362921,
+      "grad_norm": 0.3267276883125305,
+      "learning_rate": 0.00046049300699300695,
+      "loss": 3.3565,
+      "step": 40000
+    },
+    {
+      "epoch": 11.648261401362921,
+      "eval_accuracy": 0.36973110845315116,
+      "eval_loss": 3.5581958293914795,
+      "eval_runtime": 176.7361,
+      "eval_samples_per_second": 94.186,
+      "eval_steps_per_second": 5.89,
+      "step": 40000
+    },
+    {
+      "epoch": 11.662822528976644,
+      "grad_norm": 0.34614065289497375,
+      "learning_rate": 0.0004603181818181818,
+      "loss": 3.3599,
+      "step": 40050
+    },
+    {
+      "epoch": 11.677383656590367,
+      "grad_norm": 0.3481432795524597,
+      "learning_rate": 0.0004601433566433566,
+      "loss": 3.3579,
+      "step": 40100
+    },
+    {
+      "epoch": 11.69194478420409,
+      "grad_norm": 0.36765721440315247,
+      "learning_rate": 0.00045996853146853145,
+      "loss": 3.3666,
+      "step": 40150
+    },
+    {
+      "epoch": 11.70650591181781,
+      "grad_norm": 0.34848764538764954,
+      "learning_rate": 0.0004597937062937062,
+      "loss": 3.3521,
+      "step": 40200
+    },
+    {
+      "epoch": 11.721067039431533,
+      "grad_norm": 0.33765318989753723,
+      "learning_rate": 0.00045961888111888105,
+      "loss": 3.3765,
+      "step": 40250
+    },
+    {
+      "epoch": 11.735628167045256,
+      "grad_norm": 0.35295817255973816,
+      "learning_rate": 0.0004594440559440559,
+      "loss": 3.3707,
+      "step": 40300
+    },
+    {
+      "epoch": 11.750189294658979,
+      "grad_norm": 0.3578189015388489,
+      "learning_rate": 0.0004592692307692307,
+      "loss": 3.3524,
+      "step": 40350
+    },
+    {
+      "epoch": 11.764750422272702,
+      "grad_norm": 0.36456865072250366,
+      "learning_rate": 0.00045909440559440556,
+      "loss": 3.3704,
+      "step": 40400
+    },
+    {
+      "epoch": 11.779311549886422,
+      "grad_norm": 0.3516547381877899,
+      "learning_rate": 0.00045891958041958036,
+      "loss": 3.3665,
+      "step": 40450
+    },
+    {
+      "epoch": 11.793872677500145,
+      "grad_norm": 0.3550742268562317,
+      "learning_rate": 0.0004587447552447552,
+      "loss": 3.3572,
+      "step": 40500
+    },
+    {
+      "epoch": 11.808433805113868,
+      "grad_norm": 0.34383824467658997,
+      "learning_rate": 0.00045856993006993,
+      "loss": 3.3619,
+      "step": 40550
+    },
+    {
+      "epoch": 11.82299493272759,
+      "grad_norm": 0.3489612638950348,
+      "learning_rate": 0.00045839510489510487,
+      "loss": 3.3644,
+      "step": 40600
+    },
+    {
+      "epoch": 11.837556060341313,
+      "grad_norm": 0.3638046085834503,
+      "learning_rate": 0.00045822027972027967,
+      "loss": 3.3625,
+      "step": 40650
+    },
+    {
+      "epoch": 11.852117187955034,
+      "grad_norm": 0.350432425737381,
+      "learning_rate": 0.0004580454545454545,
+      "loss": 3.3741,
+      "step": 40700
+    },
+    {
+      "epoch": 11.866678315568757,
+      "grad_norm": 0.34485307335853577,
+      "learning_rate": 0.0004578706293706293,
+      "loss": 3.3554,
+      "step": 40750
+    },
+    {
+      "epoch": 11.88123944318248,
+      "grad_norm": 0.35631051659584045,
+      "learning_rate": 0.0004576958041958042,
+      "loss": 3.3613,
+      "step": 40800
+    },
+    {
+      "epoch": 11.895800570796203,
+      "grad_norm": 0.3506964147090912,
+      "learning_rate": 0.000457520979020979,
+      "loss": 3.3655,
+      "step": 40850
+    },
+    {
+      "epoch": 11.910361698409925,
+      "grad_norm": 0.35819366574287415,
+      "learning_rate": 0.00045734615384615383,
+      "loss": 3.3665,
+      "step": 40900
+    },
+    {
+      "epoch": 11.924922826023646,
+      "grad_norm": 0.35843780636787415,
+      "learning_rate": 0.0004571713286713287,
+      "loss": 3.3722,
+      "step": 40950
+    },
+    {
+      "epoch": 11.93948395363737,
+      "grad_norm": 0.3654157519340515,
+      "learning_rate": 0.00045699650349650343,
+      "loss": 3.3535,
+      "step": 41000
+    },
+    {
+      "epoch": 11.93948395363737,
+      "eval_accuracy": 0.37033843688286977,
+      "eval_loss": 3.5503153800964355,
+      "eval_runtime": 176.6657,
+      "eval_samples_per_second": 94.223,
+      "eval_steps_per_second": 5.892,
+      "step": 41000
+    },
+    {
+      "epoch": 11.954045081251092,
+      "grad_norm": 0.34321489930152893,
+      "learning_rate": 0.0004568216783216783,
+      "loss": 3.3734,
+      "step": 41050
+    },
+    {
+      "epoch": 11.968606208864815,
+      "grad_norm": 0.34735623002052307,
+      "learning_rate": 0.0004566468531468531,
+      "loss": 3.3769,
+      "step": 41100
+    },
+    {
+      "epoch": 11.983167336478537,
+      "grad_norm": 0.33347296714782715,
+      "learning_rate": 0.00045647202797202794,
+      "loss": 3.3724,
+      "step": 41150
+    },
+    {
+      "epoch": 11.99772846409226,
+      "grad_norm": 0.35811206698417664,
+      "learning_rate": 0.00045629720279720274,
+      "loss": 3.3783,
+      "step": 41200
+    },
+    {
+      "epoch": 12.012231347195527,
+      "grad_norm": 0.35955357551574707,
+      "learning_rate": 0.0004561223776223776,
+      "loss": 3.2728,
+      "step": 41250
+    },
+    {
+      "epoch": 12.02679247480925,
+      "grad_norm": 0.3704414963722229,
+      "learning_rate": 0.0004559475524475524,
+      "loss": 3.2625,
+      "step": 41300
+    },
+    {
+      "epoch": 12.041353602422971,
+      "grad_norm": 0.3613720238208771,
+      "learning_rate": 0.00045577272727272724,
+      "loss": 3.2562,
+      "step": 41350
+    },
+    {
+      "epoch": 12.055914730036694,
+      "grad_norm": 0.34454989433288574,
+      "learning_rate": 0.00045559790209790204,
+      "loss": 3.2825,
+      "step": 41400
+    },
+    {
+      "epoch": 12.070475857650417,
+      "grad_norm": 0.3962979018688202,
+      "learning_rate": 0.0004554230769230769,
+      "loss": 3.2717,
+      "step": 41450
+    },
+    {
+      "epoch": 12.08503698526414,
+      "grad_norm": 0.39520618319511414,
+      "learning_rate": 0.0004552482517482517,
+      "loss": 3.2835,
+      "step": 41500
+    },
+    {
+      "epoch": 12.099598112877862,
+      "grad_norm": 0.34733277559280396,
+      "learning_rate": 0.00045507342657342655,
+      "loss": 3.2717,
+      "step": 41550
+    },
+    {
+      "epoch": 12.114159240491583,
+      "grad_norm": 0.3350951671600342,
+      "learning_rate": 0.00045489860139860135,
+      "loss": 3.2852,
+      "step": 41600
+    },
+    {
+      "epoch": 12.128720368105306,
+      "grad_norm": 0.3716607391834259,
+      "learning_rate": 0.0004547237762237762,
+      "loss": 3.2812,
+      "step": 41650
+    },
+    {
+      "epoch": 12.143281495719028,
+      "grad_norm": 0.3612712025642395,
+      "learning_rate": 0.00045454895104895106,
+      "loss": 3.2897,
+      "step": 41700
+    },
+    {
+      "epoch": 12.157842623332751,
+      "grad_norm": 0.36439794301986694,
+      "learning_rate": 0.0004543741258741258,
+      "loss": 3.289,
+      "step": 41750
+    },
+    {
+      "epoch": 12.172403750946474,
+      "grad_norm": 0.3498424291610718,
+      "learning_rate": 0.00045419930069930066,
+      "loss": 3.2888,
+      "step": 41800
+    },
+    {
+      "epoch": 12.186964878560195,
+      "grad_norm": 0.34217751026153564,
+      "learning_rate": 0.00045402447552447546,
+      "loss": 3.2825,
+      "step": 41850
+    },
+    {
+      "epoch": 12.201526006173918,
+      "grad_norm": 0.35701295733451843,
+      "learning_rate": 0.0004538496503496503,
+      "loss": 3.2988,
+      "step": 41900
+    },
+    {
+      "epoch": 12.21608713378764,
+      "grad_norm": 0.4013568162918091,
+      "learning_rate": 0.0004536748251748251,
+      "loss": 3.2988,
+      "step": 41950
+    },
+    {
+      "epoch": 12.230648261401363,
+      "grad_norm": 0.35291993618011475,
+      "learning_rate": 0.00045349999999999996,
+      "loss": 3.3005,
+      "step": 42000
+    },
+    {
+      "epoch": 12.230648261401363,
+      "eval_accuracy": 0.370066279446788,
+      "eval_loss": 3.563735246658325,
+      "eval_runtime": 176.7458,
+      "eval_samples_per_second": 94.18,
+      "eval_steps_per_second": 5.89,
+      "step": 42000
+    },
+    {
+      "epoch": 12.245209389015086,
+      "grad_norm": 0.3605182468891144,
+      "learning_rate": 0.00045332517482517476,
+      "loss": 3.3089,
+      "step": 42050
+    },
+    {
+      "epoch": 12.259770516628807,
+      "grad_norm": 0.36443638801574707,
+      "learning_rate": 0.0004531503496503496,
+      "loss": 3.3091,
+      "step": 42100
+    },
+    {
+      "epoch": 12.27433164424253,
+      "grad_norm": 0.3468221127986908,
+      "learning_rate": 0.0004529755244755244,
+      "loss": 3.309,
+      "step": 42150
+    },
+    {
+      "epoch": 12.288892771856252,
+      "grad_norm": 0.3665069341659546,
+      "learning_rate": 0.00045280069930069927,
+      "loss": 3.3124,
+      "step": 42200
+    },
+    {
+      "epoch": 12.303453899469975,
+      "grad_norm": 0.35970741510391235,
+      "learning_rate": 0.00045262587412587407,
+      "loss": 3.3093,
+      "step": 42250
+    },
+    {
+      "epoch": 12.318015027083698,
+      "grad_norm": 0.35841190814971924,
+      "learning_rate": 0.0004524510489510489,
+      "loss": 3.3132,
+      "step": 42300
+    },
+    {
+      "epoch": 12.33257615469742,
+      "grad_norm": 0.37371912598609924,
+      "learning_rate": 0.0004522762237762238,
+      "loss": 3.3182,
+      "step": 42350
+    },
+    {
+      "epoch": 12.347137282311142,
+      "grad_norm": 0.37245097756385803,
+      "learning_rate": 0.0004521013986013986,
+      "loss": 3.3059,
+      "step": 42400
+    },
+    {
+      "epoch": 12.361698409924864,
+      "grad_norm": 0.3761231303215027,
+      "learning_rate": 0.00045192657342657343,
+      "loss": 3.3175,
+      "step": 42450
+    },
+    {
+      "epoch": 12.376259537538587,
+      "grad_norm": 0.3797931969165802,
+      "learning_rate": 0.0004517517482517482,
+      "loss": 3.3263,
+      "step": 42500
+    },
+    {
+      "epoch": 12.39082066515231,
+      "grad_norm": 0.37721502780914307,
+      "learning_rate": 0.00045157692307692303,
+      "loss": 3.3078,
+      "step": 42550
+    },
+    {
+      "epoch": 12.405381792766033,
+      "grad_norm": 0.3475891351699829,
+      "learning_rate": 0.00045140209790209783,
+      "loss": 3.332,
+      "step": 42600
+    },
+    {
+      "epoch": 12.419942920379754,
+      "grad_norm": 0.3628634214401245,
+      "learning_rate": 0.0004512272727272727,
+      "loss": 3.3225,
+      "step": 42650
+    },
+    {
+      "epoch": 12.434504047993476,
+      "grad_norm": 0.3399854302406311,
+      "learning_rate": 0.0004510524475524475,
+      "loss": 3.3231,
+      "step": 42700
+    },
+    {
+      "epoch": 12.449065175607199,
+      "grad_norm": 0.36382219195365906,
+      "learning_rate": 0.00045087762237762234,
+      "loss": 3.3166,
+      "step": 42750
+    },
+    {
+      "epoch": 12.463626303220922,
+      "grad_norm": 0.3841249942779541,
+      "learning_rate": 0.00045070279720279714,
+      "loss": 3.3256,
+      "step": 42800
+    },
+    {
+      "epoch": 12.478187430834645,
+      "grad_norm": 0.37852030992507935,
+      "learning_rate": 0.000450527972027972,
+      "loss": 3.3214,
+      "step": 42850
+    },
+    {
+      "epoch": 12.492748558448366,
+      "grad_norm": 0.3684305250644684,
+      "learning_rate": 0.0004503531468531468,
+      "loss": 3.3269,
+      "step": 42900
+    },
+    {
+      "epoch": 12.507309686062088,
+      "grad_norm": 0.36599355936050415,
+      "learning_rate": 0.00045017832167832165,
+      "loss": 3.3291,
+      "step": 42950
+    },
+    {
+      "epoch": 12.521870813675811,
+      "grad_norm": 0.35258835554122925,
+      "learning_rate": 0.0004500034965034965,
+      "loss": 3.3365,
+      "step": 43000
+    },
+    {
+      "epoch": 12.521870813675811,
+      "eval_accuracy": 0.3700295999132858,
+      "eval_loss": 3.5569796562194824,
+      "eval_runtime": 176.7498,
+      "eval_samples_per_second": 94.178,
+      "eval_steps_per_second": 5.89,
+      "step": 43000
+    },
+    {
+      "epoch": 12.536431941289534,
+      "grad_norm": 0.39876219630241394,
+      "learning_rate": 0.0004498286713286713,
+      "loss": 3.3302,
+      "step": 43050
+    },
+    {
+      "epoch": 12.550993068903256,
+      "grad_norm": 0.3408920168876648,
+      "learning_rate": 0.00044965384615384615,
+      "loss": 3.3342,
+      "step": 43100
+    },
+    {
+      "epoch": 12.565554196516977,
+      "grad_norm": 0.3744554817676544,
+      "learning_rate": 0.00044947902097902095,
+      "loss": 3.3331,
+      "step": 43150
+    },
+    {
+      "epoch": 12.5801153241307,
+      "grad_norm": 0.3452394902706146,
+      "learning_rate": 0.0004493041958041958,
+      "loss": 3.3371,
+      "step": 43200
+    },
+    {
+      "epoch": 12.594676451744423,
+      "grad_norm": 0.3524150550365448,
+      "learning_rate": 0.00044912937062937055,
+      "loss": 3.3321,
+      "step": 43250
+    },
+    {
+      "epoch": 12.609237579358146,
+      "grad_norm": 0.37291622161865234,
+      "learning_rate": 0.0004489545454545454,
+      "loss": 3.3219,
+      "step": 43300
+    },
+    {
+      "epoch": 12.623798706971868,
+      "grad_norm": 0.34290096163749695,
+      "learning_rate": 0.0004487797202797202,
+      "loss": 3.3307,
+      "step": 43350
+    },
+    {
+      "epoch": 12.63835983458559,
+      "grad_norm": 0.36508339643478394,
+      "learning_rate": 0.00044860489510489506,
+      "loss": 3.3332,
+      "step": 43400
+    },
+    {
+      "epoch": 12.652920962199312,
+      "grad_norm": 0.4050483703613281,
+      "learning_rate": 0.00044843006993006986,
+      "loss": 3.3433,
+      "step": 43450
+    },
+    {
+      "epoch": 12.667482089813035,
+      "grad_norm": 0.34905606508255005,
+      "learning_rate": 0.0004482552447552447,
+      "loss": 3.3518,
+      "step": 43500
+    },
+    {
+      "epoch": 12.682043217426758,
+      "grad_norm": 0.34155765175819397,
+      "learning_rate": 0.0004480804195804195,
+      "loss": 3.3444,
+      "step": 43550
+    },
+    {
+      "epoch": 12.69660434504048,
+      "grad_norm": 0.3530701994895935,
+      "learning_rate": 0.00044790559440559437,
+      "loss": 3.3392,
+      "step": 43600
+    },
+    {
+      "epoch": 12.711165472654203,
+      "grad_norm": 0.4012036621570587,
+      "learning_rate": 0.00044773076923076917,
+      "loss": 3.3461,
+      "step": 43650
+    },
+    {
+      "epoch": 12.725726600267924,
+      "grad_norm": 0.3666689097881317,
+      "learning_rate": 0.000447555944055944,
+      "loss": 3.3431,
+      "step": 43700
+    },
+    {
+      "epoch": 12.740287727881647,
+      "grad_norm": 0.377628892660141,
+      "learning_rate": 0.0004473811188811189,
+      "loss": 3.3503,
+      "step": 43750
+    },
+    {
+      "epoch": 12.75484885549537,
+      "grad_norm": 0.3553931713104248,
+      "learning_rate": 0.0004472062937062937,
+      "loss": 3.3467,
+      "step": 43800
+    },
+    {
+      "epoch": 12.769409983109092,
+      "grad_norm": 0.3805387020111084,
+      "learning_rate": 0.00044703146853146853,
+      "loss": 3.3458,
+      "step": 43850
+    },
+    {
+      "epoch": 12.783971110722815,
+      "grad_norm": 0.39872968196868896,
+      "learning_rate": 0.00044685664335664333,
+      "loss": 3.347,
+      "step": 43900
+    },
+    {
+      "epoch": 12.798532238336536,
+      "grad_norm": 0.3444846272468567,
+      "learning_rate": 0.0004466818181818182,
+      "loss": 3.3526,
+      "step": 43950
+    },
+    {
+      "epoch": 12.813093365950259,
+      "grad_norm": 0.3671543598175049,
+      "learning_rate": 0.00044650699300699293,
+      "loss": 3.3409,
+      "step": 44000
+    },
+    {
+      "epoch": 12.813093365950259,
+      "eval_accuracy": 0.37057367966023463,
+      "eval_loss": 3.548551559448242,
+      "eval_runtime": 176.7365,
+      "eval_samples_per_second": 94.185,
+      "eval_steps_per_second": 5.89,
+      "step": 44000
+    },
+    {
+      "epoch": 12.827654493563982,
+      "grad_norm": 0.35598888993263245,
+      "learning_rate": 0.0004463321678321678,
+      "loss": 3.3533,
+      "step": 44050
+    },
+    {
+      "epoch": 12.842215621177704,
+      "grad_norm": 0.34838467836380005,
+      "learning_rate": 0.0004461573426573426,
+      "loss": 3.3469,
+      "step": 44100
+    },
+    {
+      "epoch": 12.856776748791427,
+      "grad_norm": 0.36640146374702454,
+      "learning_rate": 0.00044598251748251744,
+      "loss": 3.3389,
+      "step": 44150
+    },
+    {
+      "epoch": 12.871337876405148,
+      "grad_norm": 0.33382320404052734,
+      "learning_rate": 0.00044580769230769224,
+      "loss": 3.3511,
+      "step": 44200
+    },
+    {
+      "epoch": 12.88589900401887,
+      "grad_norm": 0.341463565826416,
+      "learning_rate": 0.0004456328671328671,
+      "loss": 3.3517,
+      "step": 44250
+    },
+    {
+      "epoch": 12.900460131632594,
+      "grad_norm": 0.37801268696784973,
+      "learning_rate": 0.0004454580419580419,
+      "loss": 3.3596,
+      "step": 44300
+    },
+    {
+      "epoch": 12.915021259246316,
+      "grad_norm": 0.37234899401664734,
+      "learning_rate": 0.00044528321678321674,
+      "loss": 3.3506,
+      "step": 44350
+    },
+    {
+      "epoch": 12.929582386860039,
+      "grad_norm": 0.38583022356033325,
+      "learning_rate": 0.0004451083916083916,
+      "loss": 3.3452,
+      "step": 44400
+    },
+    {
+      "epoch": 12.944143514473762,
+      "grad_norm": 0.3857607841491699,
+      "learning_rate": 0.0004449335664335664,
+      "loss": 3.3561,
+      "step": 44450
+    },
+    {
+      "epoch": 12.958704642087483,
+      "grad_norm": 0.3415960371494293,
+      "learning_rate": 0.00044475874125874125,
+      "loss": 3.3509,
+      "step": 44500
+    },
+    {
+      "epoch": 12.973265769701205,
+      "grad_norm": 0.35617539286613464,
+      "learning_rate": 0.00044458391608391605,
+      "loss": 3.3557,
+      "step": 44550
+    },
+    {
+      "epoch": 12.987826897314928,
+      "grad_norm": 0.3598819673061371,
+      "learning_rate": 0.0004444090909090909,
+      "loss": 3.3423,
+      "step": 44600
+    },
+    {
+      "epoch": 13.002329780418195,
+      "grad_norm": 0.3762507140636444,
+      "learning_rate": 0.0004442342657342657,
+      "loss": 3.3355,
+      "step": 44650
+    },
+    {
+      "epoch": 13.016890908031918,
+      "grad_norm": 0.36438143253326416,
+      "learning_rate": 0.00044405944055944056,
+      "loss": 3.2297,
+      "step": 44700
+    },
+    {
+      "epoch": 13.031452035645641,
+      "grad_norm": 0.35098105669021606,
+      "learning_rate": 0.0004438846153846153,
+      "loss": 3.2446,
+      "step": 44750
+    },
+    {
+      "epoch": 13.046013163259364,
+      "grad_norm": 0.3709072470664978,
+      "learning_rate": 0.00044370979020979016,
+      "loss": 3.2529,
+      "step": 44800
+    },
+    {
+      "epoch": 13.060574290873085,
+      "grad_norm": 0.3667590916156769,
+      "learning_rate": 0.00044353496503496496,
+      "loss": 3.2429,
+      "step": 44850
+    },
+    {
+      "epoch": 13.075135418486807,
+      "grad_norm": 0.3482988774776459,
+      "learning_rate": 0.0004433601398601398,
+      "loss": 3.2617,
+      "step": 44900
+    },
+    {
+      "epoch": 13.08969654610053,
+      "grad_norm": 0.342864453792572,
+      "learning_rate": 0.0004431853146853146,
+      "loss": 3.2595,
+      "step": 44950
+    },
+    {
+      "epoch": 13.104257673714253,
+      "grad_norm": 0.36293044686317444,
+      "learning_rate": 0.00044301048951048946,
+      "loss": 3.2575,
+      "step": 45000
+    },
+    {
+      "epoch": 13.104257673714253,
+      "eval_accuracy": 0.37003489023061786,
+      "eval_loss": 3.5607776641845703,
+      "eval_runtime": 176.7546,
+      "eval_samples_per_second": 94.176,
+      "eval_steps_per_second": 5.89,
+      "step": 45000
+    },
+    {
+      "epoch": 13.118818801327976,
+      "grad_norm": 0.3598528504371643,
+      "learning_rate": 0.00044283566433566426,
+      "loss": 3.2656,
+      "step": 45050
+    },
+    {
+      "epoch": 13.133379928941697,
+      "grad_norm": 0.387949675321579,
+      "learning_rate": 0.0004426608391608391,
+      "loss": 3.266,
+      "step": 45100
+    },
+    {
+      "epoch": 13.14794105655542,
+      "grad_norm": 0.37262654304504395,
+      "learning_rate": 0.00044248601398601397,
+      "loss": 3.2734,
+      "step": 45150
+    },
+    {
+      "epoch": 13.162502184169142,
+      "grad_norm": 0.38459402322769165,
+      "learning_rate": 0.00044231118881118877,
+      "loss": 3.2757,
+      "step": 45200
+    },
+    {
+      "epoch": 13.177063311782865,
+      "grad_norm": 0.3644809424877167,
+      "learning_rate": 0.0004421363636363636,
+      "loss": 3.2803,
+      "step": 45250
+    },
+    {
+      "epoch": 13.191624439396588,
+      "grad_norm": 0.36597248911857605,
+      "learning_rate": 0.0004419615384615384,
+      "loss": 3.2775,
+      "step": 45300
+    },
+    {
+      "epoch": 13.206185567010309,
+      "grad_norm": 0.3780094385147095,
+      "learning_rate": 0.0004417867132867133,
+      "loss": 3.2709,
+      "step": 45350
+    },
+    {
+      "epoch": 13.220746694624031,
+      "grad_norm": 0.3701019585132599,
+      "learning_rate": 0.0004416118881118881,
+      "loss": 3.2711,
+      "step": 45400
+    },
+    {
+      "epoch": 13.235307822237754,
+      "grad_norm": 0.4011037051677704,
+      "learning_rate": 0.00044143706293706293,
+      "loss": 3.2755,
+      "step": 45450
+    },
+    {
+      "epoch": 13.249868949851477,
+      "grad_norm": 0.39571163058280945,
+      "learning_rate": 0.0004412622377622377,
+      "loss": 3.2932,
+      "step": 45500
+    },
+    {
+      "epoch": 13.2644300774652,
+      "grad_norm": 0.3917701244354248,
+      "learning_rate": 0.00044108741258741253,
+      "loss": 3.2888,
+      "step": 45550
+    },
+    {
+      "epoch": 13.27899120507892,
+      "grad_norm": 0.36588260531425476,
+      "learning_rate": 0.00044091258741258733,
+      "loss": 3.29,
+      "step": 45600
+    },
+    {
+      "epoch": 13.293552332692643,
+      "grad_norm": 0.374526709318161,
+      "learning_rate": 0.0004407377622377622,
+      "loss": 3.286,
+      "step": 45650
+    },
+    {
+      "epoch": 13.308113460306366,
+      "grad_norm": 0.3714640438556671,
+      "learning_rate": 0.000440562937062937,
+      "loss": 3.2921,
+      "step": 45700
+    },
+    {
+      "epoch": 13.322674587920089,
+      "grad_norm": 0.3636552095413208,
+      "learning_rate": 0.00044038811188811184,
+      "loss": 3.2898,
+      "step": 45750
+    },
+    {
+      "epoch": 13.337235715533811,
+      "grad_norm": 0.3829694092273712,
+      "learning_rate": 0.0004402132867132867,
+      "loss": 3.295,
+      "step": 45800
+    },
+    {
+      "epoch": 13.351796843147532,
+      "grad_norm": 0.3656383156776428,
+      "learning_rate": 0.0004400384615384615,
+      "loss": 3.302,
+      "step": 45850
+    },
+    {
+      "epoch": 13.366357970761255,
+      "grad_norm": 0.36748331785202026,
+      "learning_rate": 0.00043986363636363635,
+      "loss": 3.3006,
+      "step": 45900
+    },
+    {
+      "epoch": 13.380919098374978,
+      "grad_norm": 0.4061829745769501,
+      "learning_rate": 0.00043968881118881115,
+      "loss": 3.3072,
+      "step": 45950
+    },
+    {
+      "epoch": 13.3954802259887,
+      "grad_norm": 0.35320812463760376,
+      "learning_rate": 0.000439513986013986,
+      "loss": 3.2947,
+      "step": 46000
+    },
+    {
+      "epoch": 13.3954802259887,
+      "eval_accuracy": 0.3706560910480072,
+      "eval_loss": 3.5595102310180664,
+      "eval_runtime": 176.8796,
+      "eval_samples_per_second": 94.109,
+      "eval_steps_per_second": 5.885,
+      "step": 46000
+    },
+    {
+      "epoch": 13.410041353602423,
+      "grad_norm": 0.37065041065216064,
+      "learning_rate": 0.0004393391608391608,
+      "loss": 3.3009,
+      "step": 46050
+    },
+    {
+      "epoch": 13.424602481216146,
+      "grad_norm": 0.34920239448547363,
+      "learning_rate": 0.00043916433566433565,
+      "loss": 3.3226,
+      "step": 46100
+    },
+    {
+      "epoch": 13.439163608829867,
+      "grad_norm": 0.34855538606643677,
+      "learning_rate": 0.00043898951048951045,
+      "loss": 3.3077,
+      "step": 46150
+    },
+    {
+      "epoch": 13.45372473644359,
+      "grad_norm": 0.393020898103714,
+      "learning_rate": 0.0004388146853146853,
+      "loss": 3.3102,
+      "step": 46200
+    },
+    {
+      "epoch": 13.468285864057313,
+      "grad_norm": 0.3607465922832489,
+      "learning_rate": 0.00043863986013986005,
+      "loss": 3.3301,
+      "step": 46250
+    },
+    {
+      "epoch": 13.482846991671035,
+      "grad_norm": 0.3694727420806885,
+      "learning_rate": 0.0004384650349650349,
+      "loss": 3.3042,
+      "step": 46300
+    },
+    {
+      "epoch": 13.497408119284758,
+      "grad_norm": 0.363496869802475,
+      "learning_rate": 0.0004382902097902097,
+      "loss": 3.3042,
+      "step": 46350
+    },
+    {
+      "epoch": 13.51196924689848,
+      "grad_norm": 0.371970534324646,
+      "learning_rate": 0.00043811538461538456,
+      "loss": 3.3135,
+      "step": 46400
+    },
+    {
+      "epoch": 13.526530374512202,
+      "grad_norm": 0.3483518362045288,
+      "learning_rate": 0.0004379405594405594,
+      "loss": 3.3094,
+      "step": 46450
+    },
+    {
+      "epoch": 13.541091502125925,
+      "grad_norm": 0.3581492304801941,
+      "learning_rate": 0.0004377657342657342,
+      "loss": 3.3157,
+      "step": 46500
+    },
+    {
+      "epoch": 13.555652629739647,
+      "grad_norm": 0.38505807518959045,
+      "learning_rate": 0.00043759090909090907,
+      "loss": 3.3226,
+      "step": 46550
+    },
+    {
+      "epoch": 13.57021375735337,
+      "grad_norm": 0.37216296792030334,
+      "learning_rate": 0.00043741608391608387,
+      "loss": 3.3191,
+      "step": 46600
+    },
+    {
+      "epoch": 13.584774884967091,
+      "grad_norm": 0.3617168366909027,
+      "learning_rate": 0.0004372412587412587,
+      "loss": 3.3292,
+      "step": 46650
+    },
+    {
+      "epoch": 13.599336012580814,
+      "grad_norm": 0.3550806939601898,
+      "learning_rate": 0.0004370664335664335,
+      "loss": 3.3298,
+      "step": 46700
+    },
+    {
+      "epoch": 13.613897140194537,
+      "grad_norm": 0.3403339684009552,
+      "learning_rate": 0.0004368916083916084,
+      "loss": 3.3266,
+      "step": 46750
+    },
+    {
+      "epoch": 13.62845826780826,
+      "grad_norm": 0.3585992753505707,
+      "learning_rate": 0.0004367167832167832,
+      "loss": 3.3226,
+      "step": 46800
+    },
+    {
+      "epoch": 13.643019395421982,
+      "grad_norm": 0.37410402297973633,
+      "learning_rate": 0.00043654195804195803,
+      "loss": 3.3137,
+      "step": 46850
+    },
+    {
+      "epoch": 13.657580523035705,
+      "grad_norm": 0.33496248722076416,
+      "learning_rate": 0.00043636713286713283,
+      "loss": 3.3236,
+      "step": 46900
+    },
+    {
+      "epoch": 13.672141650649426,
+      "grad_norm": 0.3673466145992279,
+      "learning_rate": 0.0004361923076923077,
+      "loss": 3.3212,
+      "step": 46950
+    },
+    {
+      "epoch": 13.686702778263149,
+      "grad_norm": 0.39346593618392944,
+      "learning_rate": 0.00043601748251748243,
+      "loss": 3.316,
+      "step": 47000
+    },
+    {
+      "epoch": 13.686702778263149,
+      "eval_accuracy": 0.3710174785030894,
+      "eval_loss": 3.5519208908081055,
+      "eval_runtime": 176.6516,
+      "eval_samples_per_second": 94.231,
+      "eval_steps_per_second": 5.893,
+      "step": 47000
+    },
+    {
+      "epoch": 13.701263905876871,
+      "grad_norm": 0.3582032024860382,
+      "learning_rate": 0.00043584265734265734,
+      "loss": 3.3268,
+      "step": 47050
+    },
+    {
+      "epoch": 13.715825033490594,
+      "grad_norm": 0.3773166835308075,
+      "learning_rate": 0.0004356678321678321,
+      "loss": 3.3306,
+      "step": 47100
+    },
+    {
+      "epoch": 13.730386161104317,
+      "grad_norm": 0.35744547843933105,
+      "learning_rate": 0.00043549300699300694,
+      "loss": 3.3312,
+      "step": 47150
+    },
+    {
+      "epoch": 13.744947288718038,
+      "grad_norm": 0.3659146726131439,
+      "learning_rate": 0.0004353181818181818,
+      "loss": 3.3377,
+      "step": 47200
+    },
+    {
+      "epoch": 13.75950841633176,
+      "grad_norm": 0.3603513240814209,
+      "learning_rate": 0.0004351433566433566,
+      "loss": 3.3284,
+      "step": 47250
+    },
+    {
+      "epoch": 13.774069543945483,
+      "grad_norm": 0.3560996651649475,
+      "learning_rate": 0.00043496853146853144,
+      "loss": 3.333,
+      "step": 47300
+    },
+    {
+      "epoch": 13.788630671559206,
+      "grad_norm": 0.3501158058643341,
+      "learning_rate": 0.00043479370629370624,
+      "loss": 3.3274,
+      "step": 47350
+    },
+    {
+      "epoch": 13.803191799172929,
+      "grad_norm": 0.38781869411468506,
+      "learning_rate": 0.0004346188811188811,
+      "loss": 3.3277,
+      "step": 47400
+    },
+    {
+      "epoch": 13.81775292678665,
+      "grad_norm": 0.3612052798271179,
+      "learning_rate": 0.0004344440559440559,
+      "loss": 3.3432,
+      "step": 47450
+    },
+    {
+      "epoch": 13.832314054400372,
+      "grad_norm": 0.35562631487846375,
+      "learning_rate": 0.00043426923076923075,
+      "loss": 3.3315,
+      "step": 47500
+    },
+    {
+      "epoch": 13.846875182014095,
+      "grad_norm": 0.36037319898605347,
+      "learning_rate": 0.00043409440559440555,
+      "loss": 3.343,
+      "step": 47550
+    },
+    {
+      "epoch": 13.861436309627818,
+      "grad_norm": 0.3716438114643097,
+      "learning_rate": 0.0004339195804195804,
+      "loss": 3.327,
+      "step": 47600
+    },
+    {
+      "epoch": 13.87599743724154,
+      "grad_norm": 0.3590458929538727,
+      "learning_rate": 0.0004337447552447552,
+      "loss": 3.3399,
+      "step": 47650
+    },
+    {
+      "epoch": 13.890558564855262,
+      "grad_norm": 0.36003828048706055,
+      "learning_rate": 0.00043356993006993006,
+      "loss": 3.3323,
+      "step": 47700
+    },
+    {
+      "epoch": 13.905119692468984,
+      "grad_norm": 0.34905195236206055,
+      "learning_rate": 0.0004333951048951048,
+      "loss": 3.3425,
+      "step": 47750
+    },
+    {
+      "epoch": 13.919680820082707,
+      "grad_norm": 0.36625543236732483,
+      "learning_rate": 0.0004332202797202797,
+      "loss": 3.3323,
+      "step": 47800
+    },
+    {
+      "epoch": 13.93424194769643,
+      "grad_norm": 0.3424186110496521,
+      "learning_rate": 0.00043304545454545456,
+      "loss": 3.3272,
+      "step": 47850
+    },
+    {
+      "epoch": 13.948803075310153,
+      "grad_norm": 0.38451874256134033,
+      "learning_rate": 0.0004328706293706293,
+      "loss": 3.3336,
+      "step": 47900
+    },
+    {
+      "epoch": 13.963364202923874,
+      "grad_norm": 0.3730528950691223,
+      "learning_rate": 0.00043269580419580416,
+      "loss": 3.3345,
+      "step": 47950
+    },
+    {
+      "epoch": 13.977925330537596,
+      "grad_norm": 0.35438069701194763,
+      "learning_rate": 0.00043252097902097896,
+      "loss": 3.3273,
+      "step": 48000
+    },
+    {
+      "epoch": 13.977925330537596,
+      "eval_accuracy": 0.37150407013503006,
+      "eval_loss": 3.540936231613159,
+      "eval_runtime": 176.1535,
+      "eval_samples_per_second": 94.497,
+      "eval_steps_per_second": 5.91,
+      "step": 48000
+    },
+    {
+      "epoch": 13.992486458151319,
+      "grad_norm": 0.4073244631290436,
+      "learning_rate": 0.0004323461538461538,
+      "loss": 3.3316,
+      "step": 48050
+    },
+    {
+      "epoch": 14.006989341254586,
+      "grad_norm": 0.3481259346008301,
+      "learning_rate": 0.0004321713286713286,
+      "loss": 3.2878,
+      "step": 48100
+    },
+    {
+      "epoch": 14.021550468868309,
+      "grad_norm": 0.3760628402233124,
+      "learning_rate": 0.00043199650349650347,
+      "loss": 3.2307,
+      "step": 48150
+    },
+    {
+      "epoch": 14.036111596482032,
+      "grad_norm": 0.3724685609340668,
+      "learning_rate": 0.00043182167832167827,
+      "loss": 3.2366,
+      "step": 48200
+    },
+    {
+      "epoch": 14.050672724095755,
+      "grad_norm": 0.3769129812717438,
+      "learning_rate": 0.0004316468531468531,
+      "loss": 3.2383,
+      "step": 48250
+    },
+    {
+      "epoch": 14.065233851709475,
+      "grad_norm": 0.35858991742134094,
+      "learning_rate": 0.0004314720279720279,
+      "loss": 3.24,
+      "step": 48300
+    },
+    {
+      "epoch": 14.079794979323198,
+      "grad_norm": 0.3559223413467407,
+      "learning_rate": 0.0004312972027972028,
+      "loss": 3.2413,
+      "step": 48350
+    },
+    {
+      "epoch": 14.094356106936921,
+      "grad_norm": 0.3592166006565094,
+      "learning_rate": 0.0004311223776223776,
+      "loss": 3.2387,
+      "step": 48400
+    },
+    {
+      "epoch": 14.108917234550644,
+      "grad_norm": 0.35541239380836487,
+      "learning_rate": 0.00043094755244755243,
+      "loss": 3.2419,
+      "step": 48450
+    },
+    {
+      "epoch": 14.123478362164366,
+      "grad_norm": 0.3605928421020508,
+      "learning_rate": 0.0004307727272727272,
+      "loss": 3.2576,
+      "step": 48500
+    },
+    {
+      "epoch": 14.13803948977809,
+      "grad_norm": 0.3428098261356354,
+      "learning_rate": 0.0004305979020979021,
+      "loss": 3.2561,
+      "step": 48550
+    },
+    {
+      "epoch": 14.15260061739181,
+      "grad_norm": 0.35958048701286316,
+      "learning_rate": 0.00043042307692307694,
+      "loss": 3.2628,
+      "step": 48600
+    },
+    {
+      "epoch": 14.167161745005533,
+      "grad_norm": 0.3569062352180481,
+      "learning_rate": 0.0004302482517482517,
+      "loss": 3.2525,
+      "step": 48650
+    },
+    {
+      "epoch": 14.181722872619256,
+      "grad_norm": 0.37909266352653503,
+      "learning_rate": 0.00043007342657342654,
+      "loss": 3.2621,
+      "step": 48700
+    },
+    {
+      "epoch": 14.196284000232978,
+      "grad_norm": 0.36472129821777344,
+      "learning_rate": 0.00042989860139860134,
+      "loss": 3.2607,
+      "step": 48750
+    },
+    {
+      "epoch": 14.210845127846701,
+      "grad_norm": 0.3571818768978119,
+      "learning_rate": 0.0004297237762237762,
+      "loss": 3.2711,
+      "step": 48800
+    },
+    {
+      "epoch": 14.225406255460422,
+      "grad_norm": 0.3748188316822052,
+      "learning_rate": 0.000429548951048951,
+      "loss": 3.2693,
+      "step": 48850
+    },
+    {
+      "epoch": 14.239967383074145,
+      "grad_norm": 0.35874032974243164,
+      "learning_rate": 0.00042937412587412585,
+      "loss": 3.2678,
+      "step": 48900
+    },
+    {
+      "epoch": 14.254528510687868,
+      "grad_norm": 0.3767135739326477,
+      "learning_rate": 0.00042919930069930065,
+      "loss": 3.2705,
+      "step": 48950
+    },
+    {
+      "epoch": 14.26908963830159,
+      "grad_norm": 0.39121341705322266,
+      "learning_rate": 0.0004290244755244755,
+      "loss": 3.2788,
+      "step": 49000
+    },
+    {
+      "epoch": 14.26908963830159,
+      "eval_accuracy": 0.37083478621122284,
+      "eval_loss": 3.5569944381713867,
+      "eval_runtime": 176.6049,
+      "eval_samples_per_second": 94.256,
+      "eval_steps_per_second": 5.895,
+      "step": 49000
+    },
+    {
+      "epoch": 14.283650765915313,
+      "grad_norm": 0.34727412462234497,
+      "learning_rate": 0.0004288496503496503,
+      "loss": 3.2756,
+      "step": 49050
+    },
+    {
+      "epoch": 14.298211893529034,
+      "grad_norm": 0.38754183053970337,
+      "learning_rate": 0.00042867482517482515,
+      "loss": 3.2804,
+      "step": 49100
+    },
+    {
+      "epoch": 14.312773021142757,
+      "grad_norm": 0.3641073703765869,
+      "learning_rate": 0.00042849999999999995,
+      "loss": 3.2789,
+      "step": 49150
+    },
+    {
+      "epoch": 14.32733414875648,
+      "grad_norm": 0.3624279499053955,
+      "learning_rate": 0.0004283251748251748,
+      "loss": 3.2686,
+      "step": 49200
+    },
+    {
+      "epoch": 14.341895276370202,
+      "grad_norm": 0.3492709696292877,
+      "learning_rate": 0.00042815034965034966,
+      "loss": 3.2804,
+      "step": 49250
+    },
+    {
+      "epoch": 14.356456403983925,
+      "grad_norm": 0.3812512159347534,
+      "learning_rate": 0.00042797552447552446,
+      "loss": 3.2914,
+      "step": 49300
+    },
+    {
+      "epoch": 14.371017531597648,
+      "grad_norm": 0.3958264887332916,
+      "learning_rate": 0.0004278006993006993,
+      "loss": 3.2856,
+      "step": 49350
+    },
+    {
+      "epoch": 14.385578659211369,
+      "grad_norm": 0.42628660798072815,
+      "learning_rate": 0.00042762587412587406,
+      "loss": 3.2918,
+      "step": 49400
+    },
+    {
+      "epoch": 14.400139786825092,
+      "grad_norm": 0.3979325592517853,
+      "learning_rate": 0.0004274510489510489,
+      "loss": 3.2881,
+      "step": 49450
+    },
+    {
+      "epoch": 14.414700914438814,
+      "grad_norm": 0.36341267824172974,
+      "learning_rate": 0.0004272762237762237,
+      "loss": 3.3029,
+      "step": 49500
+    },
+    {
+      "epoch": 14.429262042052537,
+      "grad_norm": 0.41063278913497925,
+      "learning_rate": 0.00042710139860139857,
+      "loss": 3.2871,
+      "step": 49550
+    },
+    {
+      "epoch": 14.44382316966626,
+      "grad_norm": 0.37008652091026306,
+      "learning_rate": 0.00042692657342657337,
+      "loss": 3.2857,
+      "step": 49600
+    },
+    {
+      "epoch": 14.45838429727998,
+      "grad_norm": 0.3727281391620636,
+      "learning_rate": 0.0004267517482517482,
+      "loss": 3.2988,
+      "step": 49650
+    },
+    {
+      "epoch": 14.472945424893704,
+      "grad_norm": 0.390828013420105,
+      "learning_rate": 0.000426576923076923,
+      "loss": 3.3007,
+      "step": 49700
+    },
+    {
+      "epoch": 14.487506552507426,
+      "grad_norm": 0.39125722646713257,
+      "learning_rate": 0.0004264020979020979,
+      "loss": 3.2951,
+      "step": 49750
+    },
+    {
+      "epoch": 14.502067680121149,
+      "grad_norm": 0.3569597899913788,
+      "learning_rate": 0.0004262272727272727,
+      "loss": 3.29,
+      "step": 49800
+    },
+    {
+      "epoch": 14.516628807734872,
+      "grad_norm": 0.3710508644580841,
+      "learning_rate": 0.00042605244755244753,
+      "loss": 3.31,
+      "step": 49850
+    },
+    {
+      "epoch": 14.531189935348593,
+      "grad_norm": 0.3962744474411011,
+      "learning_rate": 0.00042587762237762233,
+      "loss": 3.2959,
+      "step": 49900
+    },
+    {
+      "epoch": 14.545751062962315,
+      "grad_norm": 0.381999671459198,
+      "learning_rate": 0.0004257027972027972,
+      "loss": 3.301,
+      "step": 49950
+    },
+    {
+      "epoch": 14.560312190576038,
+      "grad_norm": 0.3498106300830841,
+      "learning_rate": 0.00042552797202797204,
+      "loss": 3.3039,
+      "step": 50000
+    },
+    {
+      "epoch": 14.560312190576038,
+      "eval_accuracy": 0.3711573780058701,
+      "eval_loss": 3.5517921447753906,
+      "eval_runtime": 176.7467,
+      "eval_samples_per_second": 94.18,
+      "eval_steps_per_second": 5.89,
+      "step": 50000
+    },
+    {
+      "epoch": 14.574873318189761,
+      "grad_norm": 0.3523785173892975,
+      "learning_rate": 0.00042535314685314684,
+      "loss": 3.3028,
+      "step": 50050
+    },
+    {
+      "epoch": 14.589434445803484,
+      "grad_norm": 0.36807239055633545,
+      "learning_rate": 0.0004251783216783217,
+      "loss": 3.2884,
+      "step": 50100
+    },
+    {
+      "epoch": 14.603995573417205,
+      "grad_norm": 0.36314699053764343,
+      "learning_rate": 0.00042500349650349643,
+      "loss": 3.2948,
+      "step": 50150
+    },
+    {
+      "epoch": 14.618556701030927,
+      "grad_norm": 0.36440083384513855,
+      "learning_rate": 0.0004248286713286713,
+      "loss": 3.3073,
+      "step": 50200
+    },
+    {
+      "epoch": 14.63311782864465,
+      "grad_norm": 0.37359005212783813,
+      "learning_rate": 0.0004246538461538461,
+      "loss": 3.2999,
+      "step": 50250
+    },
+    {
+      "epoch": 14.647678956258373,
+      "grad_norm": 0.3689667284488678,
+      "learning_rate": 0.00042447902097902094,
+      "loss": 3.311,
+      "step": 50300
+    },
+    {
+      "epoch": 14.662240083872096,
+      "grad_norm": 0.3932121694087982,
+      "learning_rate": 0.00042430419580419574,
+      "loss": 3.3077,
+      "step": 50350
+    },
+    {
+      "epoch": 14.676801211485817,
+      "grad_norm": 0.3882981538772583,
+      "learning_rate": 0.0004241293706293706,
+      "loss": 3.3027,
+      "step": 50400
+    },
+    {
+      "epoch": 14.69136233909954,
+      "grad_norm": 0.3506374955177307,
+      "learning_rate": 0.0004239545454545454,
+      "loss": 3.3008,
+      "step": 50450
+    },
+    {
+      "epoch": 14.705923466713262,
+      "grad_norm": 0.3615427017211914,
+      "learning_rate": 0.00042377972027972025,
+      "loss": 3.3073,
+      "step": 50500
+    },
+    {
+      "epoch": 14.720484594326985,
+      "grad_norm": 0.4344766139984131,
+      "learning_rate": 0.00042360489510489505,
+      "loss": 3.3079,
+      "step": 50550
+    },
+    {
+      "epoch": 14.735045721940708,
+      "grad_norm": 0.36327511072158813,
+      "learning_rate": 0.0004234300699300699,
+      "loss": 3.3088,
+      "step": 50600
+    },
+    {
+      "epoch": 14.749606849554429,
+      "grad_norm": 0.3784516751766205,
+      "learning_rate": 0.00042325524475524476,
+      "loss": 3.3218,
+      "step": 50650
+    },
+    {
+      "epoch": 14.764167977168151,
+      "grad_norm": 0.3505086302757263,
+      "learning_rate": 0.00042308041958041956,
+      "loss": 3.3191,
+      "step": 50700
+    },
+    {
+      "epoch": 14.778729104781874,
+      "grad_norm": 0.36759206652641296,
+      "learning_rate": 0.0004229055944055944,
+      "loss": 3.3038,
+      "step": 50750
+    },
+    {
+      "epoch": 14.793290232395597,
+      "grad_norm": 0.36110028624534607,
+      "learning_rate": 0.0004227307692307692,
+      "loss": 3.316,
+      "step": 50800
+    },
+    {
+      "epoch": 14.80785136000932,
+      "grad_norm": 0.3693578839302063,
+      "learning_rate": 0.00042255594405594406,
+      "loss": 3.3142,
+      "step": 50850
+    },
+    {
+      "epoch": 14.822412487623042,
+      "grad_norm": 0.39012613892555237,
+      "learning_rate": 0.0004223811188811188,
+      "loss": 3.3252,
+      "step": 50900
+    },
+    {
+      "epoch": 14.836973615236763,
+      "grad_norm": 0.3453630805015564,
+      "learning_rate": 0.00042220629370629366,
+      "loss": 3.3151,
+      "step": 50950
+    },
+    {
+      "epoch": 14.851534742850486,
+      "grad_norm": 0.3628343343734741,
+      "learning_rate": 0.00042203146853146846,
+      "loss": 3.3262,
+      "step": 51000
+    },
+    {
+      "epoch": 14.851534742850486,
+      "eval_accuracy": 0.371780930075407,
+      "eval_loss": 3.543164014816284,
+      "eval_runtime": 192.0146,
+      "eval_samples_per_second": 86.691,
+      "eval_steps_per_second": 5.421,
+      "step": 51000
+    },
+    {
+      "epoch": 14.866095870464209,
+      "grad_norm": 0.3488563597202301,
+      "learning_rate": 0.0004218566433566433,
+      "loss": 3.3265,
+      "step": 51050
+    },
+    {
+      "epoch": 14.880656998077932,
+      "grad_norm": 0.35563579201698303,
+      "learning_rate": 0.0004216818181818181,
+      "loss": 3.319,
+      "step": 51100
+    },
+    {
+      "epoch": 14.895218125691654,
+      "grad_norm": 0.3641704022884369,
+      "learning_rate": 0.00042150699300699297,
+      "loss": 3.3326,
+      "step": 51150
+    },
+    {
+      "epoch": 14.909779253305375,
+      "grad_norm": 0.35692253708839417,
+      "learning_rate": 0.00042133216783216777,
+      "loss": 3.3256,
+      "step": 51200
+    },
+    {
+      "epoch": 14.924340380919098,
+      "grad_norm": 0.35215649008750916,
+      "learning_rate": 0.0004211573426573426,
+      "loss": 3.3076,
+      "step": 51250
+    },
+    {
+      "epoch": 14.93890150853282,
+      "grad_norm": 0.38561251759529114,
+      "learning_rate": 0.0004209825174825175,
+      "loss": 3.3155,
+      "step": 51300
+    },
+    {
+      "epoch": 14.953462636146543,
+      "grad_norm": 0.38138601183891296,
+      "learning_rate": 0.0004208076923076923,
+      "loss": 3.3298,
+      "step": 51350
+    },
+    {
+      "epoch": 14.968023763760266,
+      "grad_norm": 0.33545321226119995,
+      "learning_rate": 0.00042063286713286713,
+      "loss": 3.3351,
+      "step": 51400
+    },
+    {
+      "epoch": 14.982584891373987,
+      "grad_norm": 0.3515348732471466,
+      "learning_rate": 0.00042045804195804193,
+      "loss": 3.3166,
+      "step": 51450
+    },
+    {
+      "epoch": 14.99714601898771,
+      "grad_norm": 0.3943118751049042,
+      "learning_rate": 0.0004202832167832168,
+      "loss": 3.3196,
+      "step": 51500
+    },
+    {
+      "epoch": 15.011648902090977,
+      "grad_norm": 0.38474971055984497,
+      "learning_rate": 0.0004201083916083916,
+      "loss": 3.2397,
+      "step": 51550
+    },
+    {
+      "epoch": 15.0262100297047,
+      "grad_norm": 0.34515100717544556,
+      "learning_rate": 0.00041993356643356644,
+      "loss": 3.2115,
+      "step": 51600
+    },
+    {
+      "epoch": 15.040771157318423,
+      "grad_norm": 0.3761281967163086,
+      "learning_rate": 0.0004197587412587412,
+      "loss": 3.2062,
+      "step": 51650
+    },
+    {
+      "epoch": 15.055332284932145,
+      "grad_norm": 0.37440815567970276,
+      "learning_rate": 0.00041958391608391604,
+      "loss": 3.21,
+      "step": 51700
+    },
+    {
+      "epoch": 15.069893412545868,
+      "grad_norm": 0.372003436088562,
+      "learning_rate": 0.00041940909090909084,
+      "loss": 3.219,
+      "step": 51750
+    },
+    {
+      "epoch": 15.084454540159589,
+      "grad_norm": 0.3606085479259491,
+      "learning_rate": 0.0004192342657342657,
+      "loss": 3.2271,
+      "step": 51800
+    },
+    {
+      "epoch": 15.099015667773312,
+      "grad_norm": 0.3450397551059723,
+      "learning_rate": 0.0004190594405594405,
+      "loss": 3.2275,
+      "step": 51850
+    },
+    {
+      "epoch": 15.113576795387035,
+      "grad_norm": 0.3841650187969208,
+      "learning_rate": 0.00041888461538461535,
+      "loss": 3.2392,
+      "step": 51900
+    },
+    {
+      "epoch": 15.128137923000757,
+      "grad_norm": 0.37573206424713135,
+      "learning_rate": 0.00041870979020979015,
+      "loss": 3.2433,
+      "step": 51950
+    },
+    {
+      "epoch": 15.14269905061448,
+      "grad_norm": 0.3840627372264862,
+      "learning_rate": 0.000418534965034965,
+      "loss": 3.2326,
+      "step": 52000
+    },
+    {
+      "epoch": 15.14269905061448,
+      "eval_accuracy": 0.3707733009675638,
+      "eval_loss": 3.5588457584381104,
+      "eval_runtime": 234.5819,
+      "eval_samples_per_second": 70.96,
+      "eval_steps_per_second": 4.438,
+      "step": 52000
+    },
+    {
+      "epoch": 15.157260178228203,
+      "grad_norm": 0.3871009647846222,
+      "learning_rate": 0.00041836013986013985,
+      "loss": 3.2445,
+      "step": 52050
+    },
+    {
+      "epoch": 15.171821305841924,
+      "grad_norm": 0.38697710633277893,
+      "learning_rate": 0.00041818531468531465,
+      "loss": 3.2426,
+      "step": 52100
+    },
+    {
+      "epoch": 15.186382433455647,
+      "grad_norm": 0.3763573467731476,
+      "learning_rate": 0.0004180104895104895,
+      "loss": 3.2361,
+      "step": 52150
+    },
+    {
+      "epoch": 15.20094356106937,
+      "grad_norm": 0.3644348084926605,
+      "learning_rate": 0.0004178356643356643,
+      "loss": 3.2524,
+      "step": 52200
+    },
+    {
+      "epoch": 15.215504688683092,
+      "grad_norm": 0.3680911064147949,
+      "learning_rate": 0.00041766083916083916,
+      "loss": 3.2562,
+      "step": 52250
+    },
+    {
+      "epoch": 15.230065816296815,
+      "grad_norm": 0.36646798253059387,
+      "learning_rate": 0.00041748601398601396,
+      "loss": 3.253,
+      "step": 52300
+    },
+    {
+      "epoch": 15.244626943910536,
+      "grad_norm": 0.36125648021698,
+      "learning_rate": 0.0004173111888111888,
+      "loss": 3.2513,
+      "step": 52350
+    },
+    {
+      "epoch": 15.259188071524258,
+      "grad_norm": 0.4070035517215729,
+      "learning_rate": 0.00041713636363636356,
+      "loss": 3.2524,
+      "step": 52400
+    },
+    {
+      "epoch": 15.273749199137981,
+      "grad_norm": 0.3583230972290039,
+      "learning_rate": 0.0004169615384615384,
+      "loss": 3.2626,
+      "step": 52450
+    },
+    {
+      "epoch": 15.288310326751704,
+      "grad_norm": 0.35534557700157166,
+      "learning_rate": 0.0004167867132867132,
+      "loss": 3.2654,
+      "step": 52500
+    },
+    {
+      "epoch": 15.302871454365427,
+      "grad_norm": 0.3749006688594818,
+      "learning_rate": 0.00041661188811188807,
+      "loss": 3.2722,
+      "step": 52550
+    },
+    {
+      "epoch": 15.317432581979148,
+      "grad_norm": 0.3583391010761261,
+      "learning_rate": 0.00041643706293706287,
+      "loss": 3.2732,
+      "step": 52600
+    },
+    {
+      "epoch": 15.33199370959287,
+      "grad_norm": 0.38538479804992676,
+      "learning_rate": 0.0004162622377622377,
+      "loss": 3.2584,
+      "step": 52650
+    },
+    {
+      "epoch": 15.346554837206593,
+      "grad_norm": 0.3732128143310547,
+      "learning_rate": 0.0004160874125874126,
+      "loss": 3.2695,
+      "step": 52700
+    },
+    {
+      "epoch": 15.361115964820316,
+      "grad_norm": 0.3843606412410736,
+      "learning_rate": 0.0004159125874125874,
+      "loss": 3.2872,
+      "step": 52750
+    },
+    {
+      "epoch": 15.375677092434039,
+      "grad_norm": 0.3661695122718811,
+      "learning_rate": 0.00041573776223776223,
+      "loss": 3.2795,
+      "step": 52800
+    },
+    {
+      "epoch": 15.39023822004776,
+      "grad_norm": 0.37233254313468933,
+      "learning_rate": 0.00041556293706293703,
+      "loss": 3.2667,
+      "step": 52850
+    },
+    {
+      "epoch": 15.404799347661482,
+      "grad_norm": 0.3598570227622986,
+      "learning_rate": 0.0004153881118881119,
+      "loss": 3.2878,
+      "step": 52900
+    },
+    {
+      "epoch": 15.419360475275205,
+      "grad_norm": 0.4191581606864929,
+      "learning_rate": 0.0004152132867132867,
+      "loss": 3.2742,
+      "step": 52950
+    },
+    {
+      "epoch": 15.433921602888928,
+      "grad_norm": 0.3881734013557434,
+      "learning_rate": 0.00041503846153846154,
+      "loss": 3.2782,
+      "step": 53000
+    },
+    {
+      "epoch": 15.433921602888928,
+      "eval_accuracy": 0.3715597948109276,
+      "eval_loss": 3.553468704223633,
+      "eval_runtime": 176.6089,
+      "eval_samples_per_second": 94.253,
+      "eval_steps_per_second": 5.894,
+      "step": 53000
+    },
+    {
+      "epoch": 15.44848273050265,
+      "grad_norm": 0.3607991337776184,
+      "learning_rate": 0.00041486363636363634,
+      "loss": 3.2807,
+      "step": 53050
+    },
+    {
+      "epoch": 15.463043858116373,
+      "grad_norm": 0.3780747056007385,
+      "learning_rate": 0.0004146888111888112,
+      "loss": 3.2863,
+      "step": 53100
+    },
+    {
+      "epoch": 15.477604985730094,
+      "grad_norm": 0.36878126859664917,
+      "learning_rate": 0.00041451398601398593,
+      "loss": 3.2814,
+      "step": 53150
+    },
+    {
+      "epoch": 15.492166113343817,
+      "grad_norm": 0.3719026744365692,
+      "learning_rate": 0.0004143391608391608,
+      "loss": 3.2728,
+      "step": 53200
+    },
+    {
+      "epoch": 15.50672724095754,
+      "grad_norm": 0.3782918155193329,
+      "learning_rate": 0.0004141643356643356,
+      "loss": 3.2923,
+      "step": 53250
+    },
+    {
+      "epoch": 15.521288368571263,
+      "grad_norm": 0.39025405049324036,
+      "learning_rate": 0.00041398951048951044,
+      "loss": 3.2833,
+      "step": 53300
+    },
+    {
+      "epoch": 15.535849496184985,
+      "grad_norm": 0.3703659474849701,
+      "learning_rate": 0.00041381468531468524,
+      "loss": 3.2985,
+      "step": 53350
+    },
+    {
+      "epoch": 15.550410623798706,
+      "grad_norm": 0.38350167870521545,
+      "learning_rate": 0.0004136398601398601,
+      "loss": 3.2747,
+      "step": 53400
+    },
+    {
+      "epoch": 15.564971751412429,
+      "grad_norm": 0.3728027939796448,
+      "learning_rate": 0.00041346503496503495,
+      "loss": 3.2896,
+      "step": 53450
+    },
+    {
+      "epoch": 15.579532879026152,
+      "grad_norm": 0.3606811761856079,
+      "learning_rate": 0.00041329020979020975,
+      "loss": 3.2925,
+      "step": 53500
+    },
+    {
+      "epoch": 15.594094006639875,
+      "grad_norm": 0.396258682012558,
+      "learning_rate": 0.0004131153846153846,
+      "loss": 3.2913,
+      "step": 53550
+    },
+    {
+      "epoch": 15.608655134253597,
+      "grad_norm": 0.34805792570114136,
+      "learning_rate": 0.0004129405594405594,
+      "loss": 3.2967,
+      "step": 53600
+    },
+    {
+      "epoch": 15.623216261867318,
+      "grad_norm": 0.36827948689460754,
+      "learning_rate": 0.00041276573426573426,
+      "loss": 3.3035,
+      "step": 53650
+    },
+    {
+      "epoch": 15.637777389481041,
+      "grad_norm": 0.3775792717933655,
+      "learning_rate": 0.00041259090909090906,
+      "loss": 3.2914,
+      "step": 53700
+    },
+    {
+      "epoch": 15.652338517094764,
+      "grad_norm": 0.349970281124115,
+      "learning_rate": 0.0004124160839160839,
+      "loss": 3.2794,
+      "step": 53750
+    },
+    {
+      "epoch": 15.666899644708487,
+      "grad_norm": 0.3787984848022461,
+      "learning_rate": 0.0004122412587412587,
+      "loss": 3.3042,
+      "step": 53800
+    },
+    {
+      "epoch": 15.68146077232221,
+      "grad_norm": 0.37652289867401123,
+      "learning_rate": 0.00041206643356643356,
+      "loss": 3.2926,
+      "step": 53850
+    },
+    {
+      "epoch": 15.69602189993593,
+      "grad_norm": 0.3538573980331421,
+      "learning_rate": 0.0004118916083916083,
+      "loss": 3.2842,
+      "step": 53900
+    },
+    {
+      "epoch": 15.710583027549653,
+      "grad_norm": 0.38587185740470886,
+      "learning_rate": 0.00041171678321678316,
+      "loss": 3.2925,
+      "step": 53950
+    },
+    {
+      "epoch": 15.725144155163376,
+      "grad_norm": 0.389236181974411,
+      "learning_rate": 0.00041154195804195796,
+      "loss": 3.3013,
+      "step": 54000
+    },
+    {
+      "epoch": 15.725144155163376,
+      "eval_accuracy": 0.37206578427308573,
+      "eval_loss": 3.5431649684906006,
+      "eval_runtime": 176.9505,
+      "eval_samples_per_second": 94.072,
+      "eval_steps_per_second": 5.883,
+      "step": 54000
+    },
+    {
+      "epoch": 15.739705282777098,
+      "grad_norm": 0.3759095072746277,
+      "learning_rate": 0.0004113671328671328,
+      "loss": 3.2932,
+      "step": 54050
+    },
+    {
+      "epoch": 15.754266410390821,
+      "grad_norm": 0.3525058329105377,
+      "learning_rate": 0.00041119230769230767,
+      "loss": 3.3078,
+      "step": 54100
+    },
+    {
+      "epoch": 15.768827538004544,
+      "grad_norm": 0.3641799986362457,
+      "learning_rate": 0.00041101748251748247,
+      "loss": 3.3085,
+      "step": 54150
+    },
+    {
+      "epoch": 15.783388665618265,
+      "grad_norm": 0.35214030742645264,
+      "learning_rate": 0.0004108426573426573,
+      "loss": 3.304,
+      "step": 54200
+    },
+    {
+      "epoch": 15.797949793231988,
+      "grad_norm": 0.37992826104164124,
+      "learning_rate": 0.0004106678321678321,
+      "loss": 3.3111,
+      "step": 54250
+    },
+    {
+      "epoch": 15.81251092084571,
+      "grad_norm": 0.3650667667388916,
+      "learning_rate": 0.000410493006993007,
+      "loss": 3.302,
+      "step": 54300
+    },
+    {
+      "epoch": 15.827072048459433,
+      "grad_norm": 0.40603387355804443,
+      "learning_rate": 0.0004103181818181818,
+      "loss": 3.3107,
+      "step": 54350
+    },
+    {
+      "epoch": 15.841633176073156,
+      "grad_norm": 0.3534162938594818,
+      "learning_rate": 0.00041014335664335663,
+      "loss": 3.2988,
+      "step": 54400
+    },
+    {
+      "epoch": 15.856194303686877,
+      "grad_norm": 0.3707028925418854,
+      "learning_rate": 0.00040996853146853143,
+      "loss": 3.2963,
+      "step": 54450
+    },
+    {
+      "epoch": 15.8707554313006,
+      "grad_norm": 0.3668697774410248,
+      "learning_rate": 0.0004097937062937063,
+      "loss": 3.2991,
+      "step": 54500
+    },
+    {
+      "epoch": 15.885316558914322,
+      "grad_norm": 0.36460986733436584,
+      "learning_rate": 0.0004096188811188811,
+      "loss": 3.3037,
+      "step": 54550
+    },
+    {
+      "epoch": 15.899877686528045,
+      "grad_norm": 0.37511947751045227,
+      "learning_rate": 0.00040944405594405594,
+      "loss": 3.3046,
+      "step": 54600
+    },
+    {
+      "epoch": 15.914438814141768,
+      "grad_norm": 0.3535802364349365,
+      "learning_rate": 0.0004092692307692307,
+      "loss": 3.3105,
+      "step": 54650
+    },
+    {
+      "epoch": 15.928999941755489,
+      "grad_norm": 0.3434038758277893,
+      "learning_rate": 0.00040909440559440554,
+      "loss": 3.3249,
+      "step": 54700
+    },
+    {
+      "epoch": 15.943561069369212,
+      "grad_norm": 0.3652832806110382,
+      "learning_rate": 0.00040891958041958034,
+      "loss": 3.3041,
+      "step": 54750
+    },
+    {
+      "epoch": 15.958122196982934,
+      "grad_norm": 0.4125228226184845,
+      "learning_rate": 0.0004087447552447552,
+      "loss": 3.2997,
+      "step": 54800
+    },
+    {
+      "epoch": 15.972683324596657,
+      "grad_norm": 0.3930729329586029,
+      "learning_rate": 0.00040856993006993005,
+      "loss": 3.3162,
+      "step": 54850
+    },
+    {
+      "epoch": 15.98724445221038,
+      "grad_norm": 0.3785282373428345,
+      "learning_rate": 0.00040839510489510485,
+      "loss": 3.3036,
+      "step": 54900
+    },
+    {
+      "epoch": 16.001747335313645,
+      "grad_norm": 0.38858112692832947,
+      "learning_rate": 0.0004082202797202797,
+      "loss": 3.3013,
+      "step": 54950
+    },
+    {
+      "epoch": 16.01630846292737,
+      "grad_norm": 0.3704979419708252,
+      "learning_rate": 0.0004080454545454545,
+      "loss": 3.2179,
+      "step": 55000
+    },
+    {
+      "epoch": 16.01630846292737,
+      "eval_accuracy": 0.37145504652775313,
+      "eval_loss": 3.5535995960235596,
+      "eval_runtime": 1031.4847,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 1.009,
+      "step": 55000
+    },
+    {
+      "epoch": 16.03086959054109,
+      "grad_norm": 0.3996082842350006,
+      "learning_rate": 0.00040787062937062935,
+      "loss": 3.1959,
+      "step": 55050
+    },
+    {
+      "epoch": 16.045430718154815,
+      "grad_norm": 0.36863553524017334,
+      "learning_rate": 0.00040769580419580415,
+      "loss": 3.2073,
+      "step": 55100
+    },
+    {
+      "epoch": 16.059991845768536,
+      "grad_norm": 0.3670997619628906,
+      "learning_rate": 0.000407520979020979,
+      "loss": 3.2119,
+      "step": 55150
+    },
+    {
+      "epoch": 16.074552973382257,
+      "grad_norm": 0.36851897835731506,
+      "learning_rate": 0.0004073461538461538,
+      "loss": 3.2101,
+      "step": 55200
+    },
+    {
+      "epoch": 16.08911410099598,
+      "grad_norm": 0.35575056076049805,
+      "learning_rate": 0.00040717132867132866,
+      "loss": 3.221,
+      "step": 55250
+    },
+    {
+      "epoch": 16.103675228609703,
+      "grad_norm": 0.35587427020072937,
+      "learning_rate": 0.00040699650349650346,
+      "loss": 3.2227,
+      "step": 55300
+    },
+    {
+      "epoch": 16.118236356223427,
+      "grad_norm": 0.3824006915092468,
+      "learning_rate": 0.0004068216783216783,
+      "loss": 3.2373,
+      "step": 55350
+    },
+    {
+      "epoch": 16.132797483837148,
+      "grad_norm": 0.36476144194602966,
+      "learning_rate": 0.00040664685314685306,
+      "loss": 3.2184,
+      "step": 55400
+    },
+    {
+      "epoch": 16.14735861145087,
+      "grad_norm": 0.3826502859592438,
+      "learning_rate": 0.0004064720279720279,
+      "loss": 3.223,
+      "step": 55450
+    },
+    {
+      "epoch": 16.161919739064594,
+      "grad_norm": 0.3793790638446808,
+      "learning_rate": 0.00040629720279720277,
+      "loss": 3.2229,
+      "step": 55500
+    },
+    {
+      "epoch": 16.176480866678315,
+      "grad_norm": 0.3965374529361725,
+      "learning_rate": 0.00040612237762237757,
+      "loss": 3.2315,
+      "step": 55550
+    },
+    {
+      "epoch": 16.19104199429204,
+      "grad_norm": 0.40393194556236267,
+      "learning_rate": 0.0004059475524475524,
+      "loss": 3.2372,
+      "step": 55600
+    },
+    {
+      "epoch": 16.20560312190576,
+      "grad_norm": 0.37616032361984253,
+      "learning_rate": 0.0004057727272727272,
+      "loss": 3.2396,
+      "step": 55650
+    },
+    {
+      "epoch": 16.22016424951948,
+      "grad_norm": 0.3601635992527008,
+      "learning_rate": 0.0004055979020979021,
+      "loss": 3.2295,
+      "step": 55700
+    },
+    {
+      "epoch": 16.234725377133206,
+      "grad_norm": 0.35567811131477356,
+      "learning_rate": 0.0004054230769230769,
+      "loss": 3.2405,
+      "step": 55750
+    },
+    {
+      "epoch": 16.249286504746927,
+      "grad_norm": 0.39285311102867126,
+      "learning_rate": 0.00040524825174825173,
+      "loss": 3.2432,
+      "step": 55800
+    },
+    {
+      "epoch": 16.26384763236065,
+      "grad_norm": 0.37538665533065796,
+      "learning_rate": 0.00040507342657342653,
+      "loss": 3.2541,
+      "step": 55850
+    },
+    {
+      "epoch": 16.278408759974372,
+      "grad_norm": 0.3745563328266144,
+      "learning_rate": 0.0004048986013986014,
+      "loss": 3.243,
+      "step": 55900
+    },
+    {
+      "epoch": 16.292969887588093,
+      "grad_norm": 0.37519165873527527,
+      "learning_rate": 0.0004047237762237762,
+      "loss": 3.2425,
+      "step": 55950
+    },
+    {
+      "epoch": 16.307531015201818,
+      "grad_norm": 0.39395275712013245,
+      "learning_rate": 0.00040454895104895104,
+      "loss": 3.2511,
+      "step": 56000
+    },
+    {
+      "epoch": 16.307531015201818,
+      "eval_accuracy": 0.37147256335625256,
+      "eval_loss": 3.5547873973846436,
+      "eval_runtime": 176.8077,
+      "eval_samples_per_second": 94.147,
+      "eval_steps_per_second": 5.888,
+      "step": 56000
+    },
+    {
+      "epoch": 16.32209214281554,
+      "grad_norm": 0.3879060745239258,
+      "learning_rate": 0.00040437412587412583,
+      "loss": 3.2549,
+      "step": 56050
+    },
+    {
+      "epoch": 16.336653270429263,
+      "grad_norm": 0.40138155221939087,
+      "learning_rate": 0.0004041993006993007,
+      "loss": 3.2667,
+      "step": 56100
+    },
+    {
+      "epoch": 16.351214398042984,
+      "grad_norm": 0.37023717164993286,
+      "learning_rate": 0.00040402447552447554,
+      "loss": 3.2522,
+      "step": 56150
+    },
+    {
+      "epoch": 16.36577552565671,
+      "grad_norm": 0.3748226463794708,
+      "learning_rate": 0.0004038496503496503,
+      "loss": 3.2659,
+      "step": 56200
+    },
+    {
+      "epoch": 16.38033665327043,
+      "grad_norm": 0.4036701023578644,
+      "learning_rate": 0.00040367482517482514,
+      "loss": 3.2558,
+      "step": 56250
+    },
+    {
+      "epoch": 16.39489778088415,
+      "grad_norm": 0.3921171724796295,
+      "learning_rate": 0.00040349999999999994,
+      "loss": 3.2668,
+      "step": 56300
+    },
+    {
+      "epoch": 16.409458908497875,
+      "grad_norm": 0.3839941620826721,
+      "learning_rate": 0.0004033251748251748,
+      "loss": 3.2673,
+      "step": 56350
+    },
+    {
+      "epoch": 16.424020036111596,
+      "grad_norm": 0.37228092551231384,
+      "learning_rate": 0.0004031503496503496,
+      "loss": 3.253,
+      "step": 56400
+    },
+    {
+      "epoch": 16.43858116372532,
+      "grad_norm": 0.37965667247772217,
+      "learning_rate": 0.00040297552447552445,
+      "loss": 3.2689,
+      "step": 56450
+    },
+    {
+      "epoch": 16.45314229133904,
+      "grad_norm": 0.37183481454849243,
+      "learning_rate": 0.00040280069930069925,
+      "loss": 3.2604,
+      "step": 56500
+    },
+    {
+      "epoch": 16.467703418952762,
+      "grad_norm": 0.38374224305152893,
+      "learning_rate": 0.0004026258741258741,
+      "loss": 3.2681,
+      "step": 56550
+    },
+    {
+      "epoch": 16.482264546566487,
+      "grad_norm": 0.36288636922836304,
+      "learning_rate": 0.0004024510489510489,
+      "loss": 3.2725,
+      "step": 56600
+    },
+    {
+      "epoch": 16.496825674180208,
+      "grad_norm": 0.37157145142555237,
+      "learning_rate": 0.00040227622377622376,
+      "loss": 3.2709,
+      "step": 56650
+    },
+    {
+      "epoch": 16.511386801793932,
+      "grad_norm": 0.42881858348846436,
+      "learning_rate": 0.00040210139860139856,
+      "loss": 3.2691,
+      "step": 56700
+    },
+    {
+      "epoch": 16.525947929407653,
+      "grad_norm": 0.3978147804737091,
+      "learning_rate": 0.0004019265734265734,
+      "loss": 3.271,
+      "step": 56750
+    },
+    {
+      "epoch": 16.540509057021374,
+      "grad_norm": 0.3714050352573395,
+      "learning_rate": 0.0004017517482517482,
+      "loss": 3.2847,
+      "step": 56800
+    },
+    {
+      "epoch": 16.5550701846351,
+      "grad_norm": 0.385984867811203,
+      "learning_rate": 0.00040157692307692306,
+      "loss": 3.2766,
+      "step": 56850
+    },
+    {
+      "epoch": 16.56963131224882,
+      "grad_norm": 0.3949538469314575,
+      "learning_rate": 0.0004014020979020979,
+      "loss": 3.2751,
+      "step": 56900
+    },
+    {
+      "epoch": 16.584192439862544,
+      "grad_norm": 0.3426177501678467,
+      "learning_rate": 0.00040122727272727266,
+      "loss": 3.2844,
+      "step": 56950
+    },
+    {
+      "epoch": 16.598753567476265,
+      "grad_norm": 0.367607980966568,
+      "learning_rate": 0.0004010524475524475,
+      "loss": 3.2776,
+      "step": 57000
+    },
+    {
+      "epoch": 16.598753567476265,
+      "eval_accuracy": 0.3724155330300375,
+      "eval_loss": 3.545069694519043,
+      "eval_runtime": 176.6764,
+      "eval_samples_per_second": 94.217,
+      "eval_steps_per_second": 5.892,
+      "step": 57000
+    },
+    {
+      "epoch": 16.613314695089986,
+      "grad_norm": 0.3597051799297333,
+      "learning_rate": 0.0004008776223776223,
+      "loss": 3.2801,
+      "step": 57050
+    },
+    {
+      "epoch": 16.62787582270371,
+      "grad_norm": 0.4093649089336395,
+      "learning_rate": 0.00040070279720279717,
+      "loss": 3.2799,
+      "step": 57100
+    },
+    {
+      "epoch": 16.642436950317432,
+      "grad_norm": 0.3761058747768402,
+      "learning_rate": 0.00040052797202797197,
+      "loss": 3.2746,
+      "step": 57150
+    },
+    {
+      "epoch": 16.656998077931156,
+      "grad_norm": 0.3886961340904236,
+      "learning_rate": 0.0004003531468531468,
+      "loss": 3.2749,
+      "step": 57200
+    },
+    {
+      "epoch": 16.671559205544877,
+      "grad_norm": 0.4015483260154724,
+      "learning_rate": 0.0004001783216783216,
+      "loss": 3.2848,
+      "step": 57250
+    },
+    {
+      "epoch": 16.6861203331586,
+      "grad_norm": 0.32575860619544983,
+      "learning_rate": 0.0004000034965034965,
+      "loss": 3.2882,
+      "step": 57300
+    },
+    {
+      "epoch": 16.700681460772323,
+      "grad_norm": 0.3802250027656555,
+      "learning_rate": 0.0003998286713286713,
+      "loss": 3.2752,
+      "step": 57350
+    },
+    {
+      "epoch": 16.715242588386044,
+      "grad_norm": 0.40613871812820435,
+      "learning_rate": 0.00039965384615384613,
+      "loss": 3.282,
+      "step": 57400
+    },
+    {
+      "epoch": 16.72980371599977,
+      "grad_norm": 0.3728645145893097,
+      "learning_rate": 0.00039947902097902093,
+      "loss": 3.2825,
+      "step": 57450
+    },
+    {
+      "epoch": 16.74436484361349,
+      "grad_norm": 0.3911345899105072,
+      "learning_rate": 0.0003993041958041958,
+      "loss": 3.2838,
+      "step": 57500
+    },
+    {
+      "epoch": 16.75892597122721,
+      "grad_norm": 0.3494943380355835,
+      "learning_rate": 0.00039912937062937064,
+      "loss": 3.2856,
+      "step": 57550
+    },
+    {
+      "epoch": 16.773487098840935,
+      "grad_norm": 0.3775020241737366,
+      "learning_rate": 0.00039895454545454544,
+      "loss": 3.2892,
+      "step": 57600
+    },
+    {
+      "epoch": 16.788048226454656,
+      "grad_norm": 0.38465675711631775,
+      "learning_rate": 0.0003987797202797203,
+      "loss": 3.2902,
+      "step": 57650
+    },
+    {
+      "epoch": 16.80260935406838,
+      "grad_norm": 0.35246166586875916,
+      "learning_rate": 0.00039860489510489504,
+      "loss": 3.2878,
+      "step": 57700
+    },
+    {
+      "epoch": 16.8171704816821,
+      "grad_norm": 0.39134299755096436,
+      "learning_rate": 0.0003984300699300699,
+      "loss": 3.2983,
+      "step": 57750
+    },
+    {
+      "epoch": 16.831731609295822,
+      "grad_norm": 0.37794309854507446,
+      "learning_rate": 0.0003982552447552447,
+      "loss": 3.2789,
+      "step": 57800
+    },
+    {
+      "epoch": 16.846292736909547,
+      "grad_norm": 0.3562477231025696,
+      "learning_rate": 0.00039808041958041955,
+      "loss": 3.283,
+      "step": 57850
+    },
+    {
+      "epoch": 16.860853864523268,
+      "grad_norm": 0.3600902855396271,
+      "learning_rate": 0.00039790559440559435,
+      "loss": 3.2844,
+      "step": 57900
+    },
+    {
+      "epoch": 16.875414992136992,
+      "grad_norm": 0.37436723709106445,
+      "learning_rate": 0.0003977307692307692,
+      "loss": 3.2983,
+      "step": 57950
+    },
+    {
+      "epoch": 16.889976119750713,
+      "grad_norm": 0.3761354088783264,
+      "learning_rate": 0.000397555944055944,
+      "loss": 3.3012,
+      "step": 58000
+    },
+    {
+      "epoch": 16.889976119750713,
+      "eval_accuracy": 0.3723873180042666,
+      "eval_loss": 3.537418842315674,
+      "eval_runtime": 176.6887,
+      "eval_samples_per_second": 94.211,
+      "eval_steps_per_second": 5.892,
+      "step": 58000
+    },
+    {
+      "epoch": 16.904537247364434,
+      "grad_norm": 0.38446810841560364,
+      "learning_rate": 0.00039738111888111885,
+      "loss": 3.2982,
+      "step": 58050
+    },
+    {
+      "epoch": 16.91909837497816,
+      "grad_norm": 0.3951273262500763,
+      "learning_rate": 0.00039720629370629365,
+      "loss": 3.2955,
+      "step": 58100
+    },
+    {
+      "epoch": 16.93365950259188,
+      "grad_norm": 0.3898870348930359,
+      "learning_rate": 0.0003970314685314685,
+      "loss": 3.2935,
+      "step": 58150
+    },
+    {
+      "epoch": 16.948220630205604,
+      "grad_norm": 0.42458799481391907,
+      "learning_rate": 0.0003968566433566433,
+      "loss": 3.2932,
+      "step": 58200
+    },
+    {
+      "epoch": 16.962781757819325,
+      "grad_norm": 0.3850706219673157,
+      "learning_rate": 0.00039668181818181816,
+      "loss": 3.2934,
+      "step": 58250
+    },
+    {
+      "epoch": 16.977342885433046,
+      "grad_norm": 0.3601333200931549,
+      "learning_rate": 0.000396506993006993,
+      "loss": 3.2967,
+      "step": 58300
+    },
+    {
+      "epoch": 16.99190401304677,
+      "grad_norm": 0.3920387625694275,
+      "learning_rate": 0.0003963321678321678,
+      "loss": 3.291,
+      "step": 58350
+    },
+    {
+      "epoch": 17.006406896150036,
+      "grad_norm": 0.356235533952713,
+      "learning_rate": 0.00039615734265734267,
+      "loss": 3.2548,
+      "step": 58400
+    },
+    {
+      "epoch": 17.02096802376376,
+      "grad_norm": 0.3619932532310486,
+      "learning_rate": 0.0003959825174825174,
+      "loss": 3.1841,
+      "step": 58450
+    },
+    {
+      "epoch": 17.03552915137748,
+      "grad_norm": 0.36357995867729187,
+      "learning_rate": 0.00039580769230769227,
+      "loss": 3.1947,
+      "step": 58500
+    },
+    {
+      "epoch": 17.050090278991206,
+      "grad_norm": 0.3915276527404785,
+      "learning_rate": 0.00039563286713286707,
+      "loss": 3.197,
+      "step": 58550
+    },
+    {
+      "epoch": 17.064651406604927,
+      "grad_norm": 0.38006600737571716,
+      "learning_rate": 0.0003954580419580419,
+      "loss": 3.2012,
+      "step": 58600
+    },
+    {
+      "epoch": 17.07921253421865,
+      "grad_norm": 0.38852155208587646,
+      "learning_rate": 0.0003952832167832167,
+      "loss": 3.2075,
+      "step": 58650
+    },
+    {
+      "epoch": 17.093773661832373,
+      "grad_norm": 0.39789924025535583,
+      "learning_rate": 0.0003951083916083916,
+      "loss": 3.2123,
+      "step": 58700
+    },
+    {
+      "epoch": 17.108334789446094,
+      "grad_norm": 0.37218016386032104,
+      "learning_rate": 0.0003949335664335664,
+      "loss": 3.2188,
+      "step": 58750
+    },
+    {
+      "epoch": 17.122895917059818,
+      "grad_norm": 0.4078844487667084,
+      "learning_rate": 0.00039475874125874123,
+      "loss": 3.2145,
+      "step": 58800
+    },
+    {
+      "epoch": 17.13745704467354,
+      "grad_norm": 0.36902859807014465,
+      "learning_rate": 0.00039458391608391603,
+      "loss": 3.2187,
+      "step": 58850
+    },
+    {
+      "epoch": 17.152018172287264,
+      "grad_norm": 0.3675275146961212,
+      "learning_rate": 0.0003944090909090909,
+      "loss": 3.214,
+      "step": 58900
+    },
+    {
+      "epoch": 17.166579299900985,
+      "grad_norm": 0.3768998086452484,
+      "learning_rate": 0.00039423426573426573,
+      "loss": 3.2148,
+      "step": 58950
+    },
+    {
+      "epoch": 17.181140427514705,
+      "grad_norm": 0.3707525134086609,
+      "learning_rate": 0.00039405944055944053,
+      "loss": 3.2162,
+      "step": 59000
+    },
+    {
+      "epoch": 17.181140427514705,
+      "eval_accuracy": 0.3720486201324084,
+      "eval_loss": 3.556793451309204,
+      "eval_runtime": 176.7267,
+      "eval_samples_per_second": 94.191,
+      "eval_steps_per_second": 5.89,
+      "step": 59000
+    },
+    {
+      "epoch": 17.19570155512843,
+      "grad_norm": 0.43334779143333435,
+      "learning_rate": 0.0003938846153846154,
+      "loss": 3.2422,
+      "step": 59050
+    },
+    {
+      "epoch": 17.21026268274215,
+      "grad_norm": 0.3813783526420593,
+      "learning_rate": 0.0003937097902097902,
+      "loss": 3.2418,
+      "step": 59100
+    },
+    {
+      "epoch": 17.224823810355876,
+      "grad_norm": 0.3798333704471588,
+      "learning_rate": 0.00039353496503496504,
+      "loss": 3.2242,
+      "step": 59150
+    },
+    {
+      "epoch": 17.239384937969596,
+      "grad_norm": 0.3806883692741394,
+      "learning_rate": 0.0003933601398601398,
+      "loss": 3.2254,
+      "step": 59200
+    },
+    {
+      "epoch": 17.253946065583317,
+      "grad_norm": 0.36410582065582275,
+      "learning_rate": 0.00039318531468531464,
+      "loss": 3.2408,
+      "step": 59250
+    },
+    {
+      "epoch": 17.268507193197042,
+      "grad_norm": 0.35998713970184326,
+      "learning_rate": 0.00039301048951048944,
+      "loss": 3.2307,
+      "step": 59300
+    },
+    {
+      "epoch": 17.283068320810763,
+      "grad_norm": 0.3704741597175598,
+      "learning_rate": 0.0003928356643356643,
+      "loss": 3.2397,
+      "step": 59350
+    },
+    {
+      "epoch": 17.297629448424487,
+      "grad_norm": 0.3956085741519928,
+      "learning_rate": 0.0003926608391608391,
+      "loss": 3.2416,
+      "step": 59400
+    },
+    {
+      "epoch": 17.31219057603821,
+      "grad_norm": 0.42266151309013367,
+      "learning_rate": 0.00039248601398601395,
+      "loss": 3.2274,
+      "step": 59450
+    },
+    {
+      "epoch": 17.32675170365193,
+      "grad_norm": 0.3695532977581024,
+      "learning_rate": 0.00039231118881118875,
+      "loss": 3.2472,
+      "step": 59500
+    },
+    {
+      "epoch": 17.341312831265654,
+      "grad_norm": 0.3581390380859375,
+      "learning_rate": 0.0003921363636363636,
+      "loss": 3.2371,
+      "step": 59550
+    },
+    {
+      "epoch": 17.355873958879375,
+      "grad_norm": 0.3758191466331482,
+      "learning_rate": 0.00039196153846153846,
+      "loss": 3.2587,
+      "step": 59600
+    },
+    {
+      "epoch": 17.3704350864931,
+      "grad_norm": 0.39191579818725586,
+      "learning_rate": 0.00039178671328671326,
+      "loss": 3.248,
+      "step": 59650
+    },
+    {
+      "epoch": 17.38499621410682,
+      "grad_norm": 0.3746451735496521,
+      "learning_rate": 0.0003916118881118881,
+      "loss": 3.2459,
+      "step": 59700
+    },
+    {
+      "epoch": 17.39955734172054,
+      "grad_norm": 0.4051074981689453,
+      "learning_rate": 0.0003914370629370629,
+      "loss": 3.2527,
+      "step": 59750
+    },
+    {
+      "epoch": 17.414118469334266,
+      "grad_norm": 0.37997952103614807,
+      "learning_rate": 0.00039126223776223776,
+      "loss": 3.2562,
+      "step": 59800
+    },
+    {
+      "epoch": 17.428679596947987,
+      "grad_norm": 0.39231473207473755,
+      "learning_rate": 0.00039108741258741256,
+      "loss": 3.2435,
+      "step": 59850
+    },
+    {
+      "epoch": 17.44324072456171,
+      "grad_norm": 0.36344578862190247,
+      "learning_rate": 0.0003909125874125874,
+      "loss": 3.2429,
+      "step": 59900
+    },
+    {
+      "epoch": 17.457801852175432,
+      "grad_norm": 0.3916427195072174,
+      "learning_rate": 0.00039073776223776216,
+      "loss": 3.2645,
+      "step": 59950
+    },
+    {
+      "epoch": 17.472362979789153,
+      "grad_norm": 0.38155242800712585,
+      "learning_rate": 0.000390562937062937,
+      "loss": 3.2594,
+      "step": 60000
+    },
+    {
+      "epoch": 17.472362979789153,
+      "eval_accuracy": 0.37229009372796434,
+      "eval_loss": 3.546949625015259,
+      "eval_runtime": 176.6688,
+      "eval_samples_per_second": 94.222,
+      "eval_steps_per_second": 5.892,
+      "step": 60000
+    },
+    {
+      "epoch": 17.486924107402878,
+      "grad_norm": 0.379336953163147,
+      "learning_rate": 0.0003903881118881118,
+      "loss": 3.2501,
+      "step": 60050
+    },
+    {
+      "epoch": 17.5014852350166,
+      "grad_norm": 0.38774874806404114,
+      "learning_rate": 0.00039021328671328667,
+      "loss": 3.2633,
+      "step": 60100
+    },
+    {
+      "epoch": 17.516046362630323,
+      "grad_norm": 0.38635146617889404,
+      "learning_rate": 0.00039003846153846147,
+      "loss": 3.2788,
+      "step": 60150
+    },
+    {
+      "epoch": 17.530607490244044,
+      "grad_norm": 0.3680366277694702,
+      "learning_rate": 0.0003898636363636363,
+      "loss": 3.2576,
+      "step": 60200
+    },
+    {
+      "epoch": 17.545168617857765,
+      "grad_norm": 0.42120352387428284,
+      "learning_rate": 0.0003896888111888111,
+      "loss": 3.2614,
+      "step": 60250
+    },
+    {
+      "epoch": 17.55972974547149,
+      "grad_norm": 0.3842788636684418,
+      "learning_rate": 0.000389513986013986,
+      "loss": 3.2616,
+      "step": 60300
+    },
+    {
+      "epoch": 17.57429087308521,
+      "grad_norm": 0.41452935338020325,
+      "learning_rate": 0.00038933916083916083,
+      "loss": 3.2673,
+      "step": 60350
+    },
+    {
+      "epoch": 17.588852000698935,
+      "grad_norm": 0.3977508246898651,
+      "learning_rate": 0.00038916433566433563,
+      "loss": 3.2756,
+      "step": 60400
+    },
+    {
+      "epoch": 17.603413128312656,
+      "grad_norm": 0.3657890856266022,
+      "learning_rate": 0.0003889895104895105,
+      "loss": 3.2675,
+      "step": 60450
+    },
+    {
+      "epoch": 17.617974255926377,
+      "grad_norm": 0.37939056754112244,
+      "learning_rate": 0.0003888146853146853,
+      "loss": 3.2713,
+      "step": 60500
+    },
+    {
+      "epoch": 17.6325353835401,
+      "grad_norm": 0.37325090169906616,
+      "learning_rate": 0.00038863986013986014,
+      "loss": 3.2587,
+      "step": 60550
+    },
+    {
+      "epoch": 17.647096511153823,
+      "grad_norm": 0.38156962394714355,
+      "learning_rate": 0.00038846503496503494,
+      "loss": 3.2694,
+      "step": 60600
+    },
+    {
+      "epoch": 17.661657638767547,
+      "grad_norm": 0.3832074999809265,
+      "learning_rate": 0.0003882902097902098,
+      "loss": 3.2743,
+      "step": 60650
+    },
+    {
+      "epoch": 17.676218766381268,
+      "grad_norm": 0.36399969458580017,
+      "learning_rate": 0.00038811538461538454,
+      "loss": 3.2614,
+      "step": 60700
+    },
+    {
+      "epoch": 17.690779893994993,
+      "grad_norm": 0.37998104095458984,
+      "learning_rate": 0.0003879405594405594,
+      "loss": 3.2678,
+      "step": 60750
+    },
+    {
+      "epoch": 17.705341021608714,
+      "grad_norm": 0.3782190978527069,
+      "learning_rate": 0.0003877657342657342,
+      "loss": 3.2588,
+      "step": 60800
+    },
+    {
+      "epoch": 17.719902149222435,
+      "grad_norm": 0.3613925278186798,
+      "learning_rate": 0.00038759090909090905,
+      "loss": 3.272,
+      "step": 60850
+    },
+    {
+      "epoch": 17.73446327683616,
+      "grad_norm": 0.4221477210521698,
+      "learning_rate": 0.00038741608391608384,
+      "loss": 3.2725,
+      "step": 60900
+    },
+    {
+      "epoch": 17.74902440444988,
+      "grad_norm": 0.38485991954803467,
+      "learning_rate": 0.0003872412587412587,
+      "loss": 3.2701,
+      "step": 60950
+    },
+    {
+      "epoch": 17.763585532063605,
+      "grad_norm": 0.4068686068058014,
+      "learning_rate": 0.00038706643356643355,
+      "loss": 3.262,
+      "step": 61000
+    },
+    {
+      "epoch": 17.763585532063605,
+      "eval_accuracy": 0.37286426950240215,
+      "eval_loss": 3.5396716594696045,
+      "eval_runtime": 176.7294,
+      "eval_samples_per_second": 94.189,
+      "eval_steps_per_second": 5.89,
+      "step": 61000
+    },
+    {
+      "epoch": 17.778146659677326,
+      "grad_norm": 0.408058226108551,
+      "learning_rate": 0.00038689160839160835,
+      "loss": 3.2739,
+      "step": 61050
+    },
+    {
+      "epoch": 17.792707787291047,
+      "grad_norm": 0.3897380530834198,
+      "learning_rate": 0.0003867167832167832,
+      "loss": 3.2618,
+      "step": 61100
+    },
+    {
+      "epoch": 17.80726891490477,
+      "grad_norm": 0.37916770577430725,
+      "learning_rate": 0.000386541958041958,
+      "loss": 3.2838,
+      "step": 61150
+    },
+    {
+      "epoch": 17.821830042518492,
+      "grad_norm": 0.3925793468952179,
+      "learning_rate": 0.00038636713286713286,
+      "loss": 3.2773,
+      "step": 61200
+    },
+    {
+      "epoch": 17.836391170132217,
+      "grad_norm": 0.3665933609008789,
+      "learning_rate": 0.00038619230769230766,
+      "loss": 3.2799,
+      "step": 61250
+    },
+    {
+      "epoch": 17.850952297745938,
+      "grad_norm": 0.3553600311279297,
+      "learning_rate": 0.0003860174825174825,
+      "loss": 3.2711,
+      "step": 61300
+    },
+    {
+      "epoch": 17.86551342535966,
+      "grad_norm": 0.392948180437088,
+      "learning_rate": 0.0003858426573426573,
+      "loss": 3.2847,
+      "step": 61350
+    },
+    {
+      "epoch": 17.880074552973383,
+      "grad_norm": 0.3728949725627899,
+      "learning_rate": 0.00038566783216783217,
+      "loss": 3.2846,
+      "step": 61400
+    },
+    {
+      "epoch": 17.894635680587104,
+      "grad_norm": 0.36359310150146484,
+      "learning_rate": 0.0003854930069930069,
+      "loss": 3.2798,
+      "step": 61450
+    },
+    {
+      "epoch": 17.90919680820083,
+      "grad_norm": 0.38679951429367065,
+      "learning_rate": 0.00038531818181818177,
+      "loss": 3.2756,
+      "step": 61500
+    },
+    {
+      "epoch": 17.92375793581455,
+      "grad_norm": 0.3753286898136139,
+      "learning_rate": 0.00038514335664335657,
+      "loss": 3.2761,
+      "step": 61550
+    },
+    {
+      "epoch": 17.93831906342827,
+      "grad_norm": 0.38257867097854614,
+      "learning_rate": 0.0003849685314685314,
+      "loss": 3.2809,
+      "step": 61600
+    },
+    {
+      "epoch": 17.952880191041995,
+      "grad_norm": 0.3879117965698242,
+      "learning_rate": 0.0003847937062937062,
+      "loss": 3.2732,
+      "step": 61650
+    },
+    {
+      "epoch": 17.967441318655716,
+      "grad_norm": 0.3776424825191498,
+      "learning_rate": 0.0003846188811188811,
+      "loss": 3.2809,
+      "step": 61700
+    },
+    {
+      "epoch": 17.98200244626944,
+      "grad_norm": 0.3808050751686096,
+      "learning_rate": 0.00038444405594405593,
+      "loss": 3.2822,
+      "step": 61750
+    },
+    {
+      "epoch": 17.99656357388316,
+      "grad_norm": 0.4322393536567688,
+      "learning_rate": 0.00038426923076923073,
+      "loss": 3.2754,
+      "step": 61800
+    },
+    {
+      "epoch": 18.01106645698643,
+      "grad_norm": 0.3647869825363159,
+      "learning_rate": 0.0003840944055944056,
+      "loss": 3.2042,
+      "step": 61850
+    },
+    {
+      "epoch": 18.02562758460015,
+      "grad_norm": 0.3591197729110718,
+      "learning_rate": 0.0003839195804195804,
+      "loss": 3.1856,
+      "step": 61900
+    },
+    {
+      "epoch": 18.040188712213872,
+      "grad_norm": 0.36585232615470886,
+      "learning_rate": 0.00038374475524475523,
+      "loss": 3.1831,
+      "step": 61950
+    },
+    {
+      "epoch": 18.054749839827597,
+      "grad_norm": 0.37394168972969055,
+      "learning_rate": 0.00038356993006993003,
+      "loss": 3.1853,
+      "step": 62000
+    },
+    {
+      "epoch": 18.054749839827597,
+      "eval_accuracy": 0.37270273847986374,
+      "eval_loss": 3.549687623977661,
+      "eval_runtime": 176.6151,
+      "eval_samples_per_second": 94.25,
+      "eval_steps_per_second": 5.894,
+      "step": 62000
+    },
+    {
+      "epoch": 18.069310967441318,
+      "grad_norm": 0.3535243570804596,
+      "learning_rate": 0.0003833951048951049,
+      "loss": 3.1757,
+      "step": 62050
+    },
+    {
+      "epoch": 18.083872095055042,
+      "grad_norm": 0.4245968163013458,
+      "learning_rate": 0.0003832202797202797,
+      "loss": 3.1926,
+      "step": 62100
+    },
+    {
+      "epoch": 18.098433222668763,
+      "grad_norm": 0.3650285601615906,
+      "learning_rate": 0.00038304545454545454,
+      "loss": 3.1886,
+      "step": 62150
+    },
+    {
+      "epoch": 18.112994350282484,
+      "grad_norm": 0.36255475878715515,
+      "learning_rate": 0.0003828706293706293,
+      "loss": 3.1969,
+      "step": 62200
+    },
+    {
+      "epoch": 18.12755547789621,
+      "grad_norm": 0.4018917381763458,
+      "learning_rate": 0.00038269580419580414,
+      "loss": 3.2017,
+      "step": 62250
+    },
+    {
+      "epoch": 18.14211660550993,
+      "grad_norm": 0.39976486563682556,
+      "learning_rate": 0.00038252097902097894,
+      "loss": 3.202,
+      "step": 62300
+    },
+    {
+      "epoch": 18.156677733123654,
+      "grad_norm": 0.4026753306388855,
+      "learning_rate": 0.0003823461538461538,
+      "loss": 3.2164,
+      "step": 62350
+    },
+    {
+      "epoch": 18.171238860737375,
+      "grad_norm": 0.3889211416244507,
+      "learning_rate": 0.00038217132867132865,
+      "loss": 3.2084,
+      "step": 62400
+    },
+    {
+      "epoch": 18.185799988351096,
+      "grad_norm": 0.3884575664997101,
+      "learning_rate": 0.00038199650349650345,
+      "loss": 3.2041,
+      "step": 62450
+    },
+    {
+      "epoch": 18.20036111596482,
+      "grad_norm": 0.4104594886302948,
+      "learning_rate": 0.0003818216783216783,
+      "loss": 3.2115,
+      "step": 62500
+    },
+    {
+      "epoch": 18.214922243578542,
+      "grad_norm": 0.40058377385139465,
+      "learning_rate": 0.0003816468531468531,
+      "loss": 3.2133,
+      "step": 62550
+    },
+    {
+      "epoch": 18.229483371192266,
+      "grad_norm": 0.38720259070396423,
+      "learning_rate": 0.00038147202797202796,
+      "loss": 3.2087,
+      "step": 62600
+    },
+    {
+      "epoch": 18.244044498805987,
+      "grad_norm": 0.38542911410331726,
+      "learning_rate": 0.00038129720279720276,
+      "loss": 3.2212,
+      "step": 62650
+    },
+    {
+      "epoch": 18.25860562641971,
+      "grad_norm": 0.39218926429748535,
+      "learning_rate": 0.0003811223776223776,
+      "loss": 3.2203,
+      "step": 62700
+    },
+    {
+      "epoch": 18.273166754033433,
+      "grad_norm": 0.40636080503463745,
+      "learning_rate": 0.0003809475524475524,
+      "loss": 3.2261,
+      "step": 62750
+    },
+    {
+      "epoch": 18.287727881647154,
+      "grad_norm": 0.3898778259754181,
+      "learning_rate": 0.00038077272727272726,
+      "loss": 3.2183,
+      "step": 62800
+    },
+    {
+      "epoch": 18.30228900926088,
+      "grad_norm": 0.3944312334060669,
+      "learning_rate": 0.00038059790209790206,
+      "loss": 3.2277,
+      "step": 62850
+    },
+    {
+      "epoch": 18.3168501368746,
+      "grad_norm": 0.38517752289772034,
+      "learning_rate": 0.0003804230769230769,
+      "loss": 3.2321,
+      "step": 62900
+    },
+    {
+      "epoch": 18.33141126448832,
+      "grad_norm": 0.38035276532173157,
+      "learning_rate": 0.00038024825174825166,
+      "loss": 3.2303,
+      "step": 62950
+    },
+    {
+      "epoch": 18.345972392102045,
+      "grad_norm": 0.3749290704727173,
+      "learning_rate": 0.0003800734265734265,
+      "loss": 3.239,
+      "step": 63000
+    },
+    {
+      "epoch": 18.345972392102045,
+      "eval_accuracy": 0.3725924647541425,
+      "eval_loss": 3.548588514328003,
+      "eval_runtime": 176.7112,
+      "eval_samples_per_second": 94.199,
+      "eval_steps_per_second": 5.891,
+      "step": 63000
+    },
+    {
+      "epoch": 18.360533519715766,
+      "grad_norm": 0.36092501878738403,
+      "learning_rate": 0.0003798986013986013,
+      "loss": 3.2431,
+      "step": 63050
+    },
+    {
+      "epoch": 18.37509464732949,
+      "grad_norm": 0.36838671565055847,
+      "learning_rate": 0.00037972377622377617,
+      "loss": 3.2186,
+      "step": 63100
+    },
+    {
+      "epoch": 18.38965577494321,
+      "grad_norm": 0.38395121693611145,
+      "learning_rate": 0.000379548951048951,
+      "loss": 3.2228,
+      "step": 63150
+    },
+    {
+      "epoch": 18.404216902556932,
+      "grad_norm": 0.40639710426330566,
+      "learning_rate": 0.0003793741258741258,
+      "loss": 3.2346,
+      "step": 63200
+    },
+    {
+      "epoch": 18.418778030170657,
+      "grad_norm": 0.3863862156867981,
+      "learning_rate": 0.0003791993006993007,
+      "loss": 3.2347,
+      "step": 63250
+    },
+    {
+      "epoch": 18.433339157784378,
+      "grad_norm": 0.40013137459754944,
+      "learning_rate": 0.0003790244755244755,
+      "loss": 3.2429,
+      "step": 63300
+    },
+    {
+      "epoch": 18.447900285398102,
+      "grad_norm": 0.40367773175239563,
+      "learning_rate": 0.00037884965034965033,
+      "loss": 3.2509,
+      "step": 63350
+    },
+    {
+      "epoch": 18.462461413011823,
+      "grad_norm": 0.38813164830207825,
+      "learning_rate": 0.00037867482517482513,
+      "loss": 3.2466,
+      "step": 63400
+    },
+    {
+      "epoch": 18.477022540625548,
+      "grad_norm": 0.3755754232406616,
+      "learning_rate": 0.0003785,
+      "loss": 3.2389,
+      "step": 63450
+    },
+    {
+      "epoch": 18.49158366823927,
+      "grad_norm": 0.3727249503135681,
+      "learning_rate": 0.0003783251748251748,
+      "loss": 3.2401,
+      "step": 63500
+    },
+    {
+      "epoch": 18.50614479585299,
+      "grad_norm": 0.3614943027496338,
+      "learning_rate": 0.00037815034965034964,
+      "loss": 3.2367,
+      "step": 63550
+    },
+    {
+      "epoch": 18.520705923466714,
+      "grad_norm": 0.3896731436252594,
+      "learning_rate": 0.00037797552447552444,
+      "loss": 3.2412,
+      "step": 63600
+    },
+    {
+      "epoch": 18.535267051080435,
+      "grad_norm": 0.3661154508590698,
+      "learning_rate": 0.0003778006993006993,
+      "loss": 3.2522,
+      "step": 63650
+    },
+    {
+      "epoch": 18.54982817869416,
+      "grad_norm": 0.37164583802223206,
+      "learning_rate": 0.00037762587412587404,
+      "loss": 3.2627,
+      "step": 63700
+    },
+    {
+      "epoch": 18.56438930630788,
+      "grad_norm": 0.4042896628379822,
+      "learning_rate": 0.0003774510489510489,
+      "loss": 3.2595,
+      "step": 63750
+    },
+    {
+      "epoch": 18.5789504339216,
+      "grad_norm": 0.36743268370628357,
+      "learning_rate": 0.0003772762237762238,
+      "loss": 3.245,
+      "step": 63800
+    },
+    {
+      "epoch": 18.593511561535326,
+      "grad_norm": 0.3608546853065491,
+      "learning_rate": 0.00037710139860139854,
+      "loss": 3.2389,
+      "step": 63850
+    },
+    {
+      "epoch": 18.608072689149047,
+      "grad_norm": 0.37642180919647217,
+      "learning_rate": 0.0003769265734265734,
+      "loss": 3.248,
+      "step": 63900
+    },
+    {
+      "epoch": 18.62263381676277,
+      "grad_norm": 0.3853236734867096,
+      "learning_rate": 0.0003767517482517482,
+      "loss": 3.253,
+      "step": 63950
+    },
+    {
+      "epoch": 18.637194944376493,
+      "grad_norm": 0.39252620935440063,
+      "learning_rate": 0.00037657692307692305,
+      "loss": 3.2576,
+      "step": 64000
+    },
+    {
+      "epoch": 18.637194944376493,
+      "eval_accuracy": 0.3727976115040184,
+      "eval_loss": 3.543614149093628,
+      "eval_runtime": 176.7187,
+      "eval_samples_per_second": 94.195,
+      "eval_steps_per_second": 5.891,
+      "step": 64000
+    },
+    {
+      "epoch": 18.651756071990214,
+      "grad_norm": 0.3591403663158417,
+      "learning_rate": 0.00037640209790209785,
+      "loss": 3.2592,
+      "step": 64050
+    },
+    {
+      "epoch": 18.666317199603938,
+      "grad_norm": 0.3907413184642792,
+      "learning_rate": 0.0003762272727272727,
+      "loss": 3.2667,
+      "step": 64100
+    },
+    {
+      "epoch": 18.68087832721766,
+      "grad_norm": 0.37977805733680725,
+      "learning_rate": 0.0003760524475524475,
+      "loss": 3.2631,
+      "step": 64150
+    },
+    {
+      "epoch": 18.695439454831384,
+      "grad_norm": 0.3672906160354614,
+      "learning_rate": 0.00037587762237762236,
+      "loss": 3.2624,
+      "step": 64200
+    },
+    {
+      "epoch": 18.710000582445105,
+      "grad_norm": 0.3918323516845703,
+      "learning_rate": 0.00037570279720279716,
+      "loss": 3.2652,
+      "step": 64250
+    },
+    {
+      "epoch": 18.724561710058826,
+      "grad_norm": 0.3797747790813446,
+      "learning_rate": 0.000375527972027972,
+      "loss": 3.2573,
+      "step": 64300
+    },
+    {
+      "epoch": 18.73912283767255,
+      "grad_norm": 0.3932407796382904,
+      "learning_rate": 0.0003753531468531468,
+      "loss": 3.2666,
+      "step": 64350
+    },
+    {
+      "epoch": 18.75368396528627,
+      "grad_norm": 0.3932470381259918,
+      "learning_rate": 0.00037517832167832167,
+      "loss": 3.261,
+      "step": 64400
+    },
+    {
+      "epoch": 18.768245092899996,
+      "grad_norm": 0.42663368582725525,
+      "learning_rate": 0.0003750034965034965,
+      "loss": 3.2692,
+      "step": 64450
+    },
+    {
+      "epoch": 18.782806220513717,
+      "grad_norm": 0.3719504773616791,
+      "learning_rate": 0.00037482867132867127,
+      "loss": 3.2746,
+      "step": 64500
+    },
+    {
+      "epoch": 18.797367348127437,
+      "grad_norm": 0.3768567442893982,
+      "learning_rate": 0.0003746538461538462,
+      "loss": 3.263,
+      "step": 64550
+    },
+    {
+      "epoch": 18.811928475741162,
+      "grad_norm": 0.38159942626953125,
+      "learning_rate": 0.0003744790209790209,
+      "loss": 3.2707,
+      "step": 64600
+    },
+    {
+      "epoch": 18.826489603354883,
+      "grad_norm": 0.3998315632343292,
+      "learning_rate": 0.0003743041958041958,
+      "loss": 3.2678,
+      "step": 64650
+    },
+    {
+      "epoch": 18.841050730968607,
+      "grad_norm": 0.40491533279418945,
+      "learning_rate": 0.0003741293706293706,
+      "loss": 3.2784,
+      "step": 64700
+    },
+    {
+      "epoch": 18.85561185858233,
+      "grad_norm": 0.39045944809913635,
+      "learning_rate": 0.0003739545454545454,
+      "loss": 3.2642,
+      "step": 64750
+    },
+    {
+      "epoch": 18.87017298619605,
+      "grad_norm": 0.3865518569946289,
+      "learning_rate": 0.0003737797202797202,
+      "loss": 3.278,
+      "step": 64800
+    },
+    {
+      "epoch": 18.884734113809774,
+      "grad_norm": 0.40172427892684937,
+      "learning_rate": 0.0003736048951048951,
+      "loss": 3.2685,
+      "step": 64850
+    },
+    {
+      "epoch": 18.899295241423495,
+      "grad_norm": 0.37912994623184204,
+      "learning_rate": 0.0003734300699300699,
+      "loss": 3.2602,
+      "step": 64900
+    },
+    {
+      "epoch": 18.91385636903722,
+      "grad_norm": 0.36898526549339294,
+      "learning_rate": 0.00037325524475524473,
+      "loss": 3.2762,
+      "step": 64950
+    },
+    {
+      "epoch": 18.92841749665094,
+      "grad_norm": 0.4474957287311554,
+      "learning_rate": 0.00037308041958041953,
+      "loss": 3.2736,
+      "step": 65000
+    },
+    {
+      "epoch": 18.92841749665094,
+      "eval_accuracy": 0.3733603837055405,
+      "eval_loss": 3.5318191051483154,
+      "eval_runtime": 176.8986,
+      "eval_samples_per_second": 94.099,
+      "eval_steps_per_second": 5.885,
+      "step": 65000
+    },
+    {
+      "epoch": 18.94297862426466,
+      "grad_norm": 0.3818124830722809,
+      "learning_rate": 0.0003729055944055944,
+      "loss": 3.2739,
+      "step": 65050
+    },
+    {
+      "epoch": 18.957539751878386,
+      "grad_norm": 0.39567485451698303,
+      "learning_rate": 0.0003727307692307692,
+      "loss": 3.2638,
+      "step": 65100
+    },
+    {
+      "epoch": 18.972100879492107,
+      "grad_norm": 0.3894628882408142,
+      "learning_rate": 0.00037255594405594404,
+      "loss": 3.2719,
+      "step": 65150
+    },
+    {
+      "epoch": 18.98666200710583,
+      "grad_norm": 0.3980902135372162,
+      "learning_rate": 0.0003723811188811189,
+      "loss": 3.2853,
+      "step": 65200
+    },
+    {
+      "epoch": 19.001164890209097,
+      "grad_norm": 0.40676209330558777,
+      "learning_rate": 0.00037220629370629364,
+      "loss": 3.2692,
+      "step": 65250
+    },
+    {
+      "epoch": 19.01572601782282,
+      "grad_norm": 0.43653473258018494,
+      "learning_rate": 0.00037203146853146855,
+      "loss": 3.1594,
+      "step": 65300
+    },
+    {
+      "epoch": 19.030287145436542,
+      "grad_norm": 0.3918614387512207,
+      "learning_rate": 0.0003718566433566433,
+      "loss": 3.1726,
+      "step": 65350
+    },
+    {
+      "epoch": 19.044848273050263,
+      "grad_norm": 0.3767155110836029,
+      "learning_rate": 0.00037168181818181815,
+      "loss": 3.1696,
+      "step": 65400
+    },
+    {
+      "epoch": 19.059409400663988,
+      "grad_norm": 0.3941446840763092,
+      "learning_rate": 0.00037150699300699295,
+      "loss": 3.1736,
+      "step": 65450
+    },
+    {
+      "epoch": 19.07397052827771,
+      "grad_norm": 0.4108849763870239,
+      "learning_rate": 0.0003713321678321678,
+      "loss": 3.1941,
+      "step": 65500
+    },
+    {
+      "epoch": 19.088531655891433,
+      "grad_norm": 0.40548184514045715,
+      "learning_rate": 0.0003711573426573426,
+      "loss": 3.186,
+      "step": 65550
+    },
+    {
+      "epoch": 19.103092783505154,
+      "grad_norm": 0.3613770306110382,
+      "learning_rate": 0.00037098251748251746,
+      "loss": 3.1908,
+      "step": 65600
+    },
+    {
+      "epoch": 19.11765391111888,
+      "grad_norm": 0.379202276468277,
+      "learning_rate": 0.00037080769230769226,
+      "loss": 3.1949,
+      "step": 65650
+    },
+    {
+      "epoch": 19.1322150387326,
+      "grad_norm": 0.39599117636680603,
+      "learning_rate": 0.0003706328671328671,
+      "loss": 3.1999,
+      "step": 65700
+    },
+    {
+      "epoch": 19.14677616634632,
+      "grad_norm": 0.36203324794769287,
+      "learning_rate": 0.0003704580419580419,
+      "loss": 3.2018,
+      "step": 65750
+    },
+    {
+      "epoch": 19.161337293960045,
+      "grad_norm": 0.389349102973938,
+      "learning_rate": 0.00037028321678321676,
+      "loss": 3.1938,
+      "step": 65800
+    },
+    {
+      "epoch": 19.175898421573766,
+      "grad_norm": 0.3745954930782318,
+      "learning_rate": 0.0003701083916083916,
+      "loss": 3.2047,
+      "step": 65850
+    },
+    {
+      "epoch": 19.19045954918749,
+      "grad_norm": 0.41214945912361145,
+      "learning_rate": 0.0003699335664335664,
+      "loss": 3.209,
+      "step": 65900
+    },
+    {
+      "epoch": 19.20502067680121,
+      "grad_norm": 0.411510169506073,
+      "learning_rate": 0.00036975874125874127,
+      "loss": 3.2059,
+      "step": 65950
+    },
+    {
+      "epoch": 19.219581804414933,
+      "grad_norm": 0.3707365393638611,
+      "learning_rate": 0.00036958391608391607,
+      "loss": 3.1832,
+      "step": 66000
+    },
+    {
+      "epoch": 19.219581804414933,
+      "eval_accuracy": 0.3726959198486358,
+      "eval_loss": 3.5537943840026855,
+      "eval_runtime": 176.7606,
+      "eval_samples_per_second": 94.173,
+      "eval_steps_per_second": 5.889,
+      "step": 66000
+    },
+    {
+      "epoch": 19.234142932028657,
+      "grad_norm": 0.44211095571517944,
+      "learning_rate": 0.0003694090909090909,
+      "loss": 3.2018,
+      "step": 66050
+    },
+    {
+      "epoch": 19.248704059642378,
+      "grad_norm": 0.3968811631202698,
+      "learning_rate": 0.00036923426573426567,
+      "loss": 3.2157,
+      "step": 66100
+    },
+    {
+      "epoch": 19.263265187256103,
+      "grad_norm": 0.3941178321838379,
+      "learning_rate": 0.0003690594405594405,
+      "loss": 3.2102,
+      "step": 66150
+    },
+    {
+      "epoch": 19.277826314869824,
+      "grad_norm": 0.36997464299201965,
+      "learning_rate": 0.0003688846153846153,
+      "loss": 3.2282,
+      "step": 66200
+    },
+    {
+      "epoch": 19.292387442483545,
+      "grad_norm": 0.3910622000694275,
+      "learning_rate": 0.0003687097902097902,
+      "loss": 3.2084,
+      "step": 66250
+    },
+    {
+      "epoch": 19.30694857009727,
+      "grad_norm": 0.4019615054130554,
+      "learning_rate": 0.000368534965034965,
+      "loss": 3.2072,
+      "step": 66300
+    },
+    {
+      "epoch": 19.32150969771099,
+      "grad_norm": 0.4168713092803955,
+      "learning_rate": 0.00036836013986013983,
+      "loss": 3.2272,
+      "step": 66350
+    },
+    {
+      "epoch": 19.336070825324715,
+      "grad_norm": 0.3616064190864563,
+      "learning_rate": 0.00036818531468531463,
+      "loss": 3.2187,
+      "step": 66400
+    },
+    {
+      "epoch": 19.350631952938436,
+      "grad_norm": 0.38284698128700256,
+      "learning_rate": 0.0003680104895104895,
+      "loss": 3.2132,
+      "step": 66450
+    },
+    {
+      "epoch": 19.365193080552157,
+      "grad_norm": 0.38134893774986267,
+      "learning_rate": 0.0003678356643356643,
+      "loss": 3.2144,
+      "step": 66500
+    },
+    {
+      "epoch": 19.37975420816588,
+      "grad_norm": 0.37159982323646545,
+      "learning_rate": 0.00036766083916083914,
+      "loss": 3.2269,
+      "step": 66550
+    },
+    {
+      "epoch": 19.394315335779602,
+      "grad_norm": 0.3975840210914612,
+      "learning_rate": 0.000367486013986014,
+      "loss": 3.2155,
+      "step": 66600
+    },
+    {
+      "epoch": 19.408876463393327,
+      "grad_norm": 0.40514418482780457,
+      "learning_rate": 0.0003673111888111888,
+      "loss": 3.2185,
+      "step": 66650
+    },
+    {
+      "epoch": 19.423437591007048,
+      "grad_norm": 0.9633327722549438,
+      "learning_rate": 0.00036713636363636365,
+      "loss": 3.2242,
+      "step": 66700
+    },
+    {
+      "epoch": 19.43799871862077,
+      "grad_norm": 0.38619816303253174,
+      "learning_rate": 0.00036696153846153844,
+      "loss": 3.2197,
+      "step": 66750
+    },
+    {
+      "epoch": 19.452559846234493,
+      "grad_norm": 0.38963261246681213,
+      "learning_rate": 0.0003667867132867133,
+      "loss": 3.2261,
+      "step": 66800
+    },
+    {
+      "epoch": 19.467120973848214,
+      "grad_norm": 0.4200080931186676,
+      "learning_rate": 0.00036661188811188804,
+      "loss": 3.2345,
+      "step": 66850
+    },
+    {
+      "epoch": 19.48168210146194,
+      "grad_norm": 0.3921654522418976,
+      "learning_rate": 0.0003664370629370629,
+      "loss": 3.2386,
+      "step": 66900
+    },
+    {
+      "epoch": 19.49624322907566,
+      "grad_norm": 0.37555941939353943,
+      "learning_rate": 0.0003662622377622377,
+      "loss": 3.2454,
+      "step": 66950
+    },
+    {
+      "epoch": 19.51080435668938,
+      "grad_norm": 0.38313743472099304,
+      "learning_rate": 0.00036608741258741255,
+      "loss": 3.2362,
+      "step": 67000
+    },
+    {
+      "epoch": 19.51080435668938,
+      "eval_accuracy": 0.3727880889328207,
+      "eval_loss": 3.547034740447998,
+      "eval_runtime": 176.9088,
+      "eval_samples_per_second": 94.094,
+      "eval_steps_per_second": 5.884,
+      "step": 67000
+    },
+    {
+      "epoch": 19.525365484303105,
+      "grad_norm": 0.39016804099082947,
+      "learning_rate": 0.00036591258741258735,
+      "loss": 3.2385,
+      "step": 67050
+    },
+    {
+      "epoch": 19.539926611916826,
+      "grad_norm": 0.3868093490600586,
+      "learning_rate": 0.0003657377622377622,
+      "loss": 3.2346,
+      "step": 67100
+    },
+    {
+      "epoch": 19.55448773953055,
+      "grad_norm": 0.3916149139404297,
+      "learning_rate": 0.000365562937062937,
+      "loss": 3.2407,
+      "step": 67150
+    },
+    {
+      "epoch": 19.56904886714427,
+      "grad_norm": 0.41935995221138,
+      "learning_rate": 0.00036538811188811186,
+      "loss": 3.2306,
+      "step": 67200
+    },
+    {
+      "epoch": 19.583609994757992,
+      "grad_norm": 0.37837082147598267,
+      "learning_rate": 0.0003652132867132867,
+      "loss": 3.2337,
+      "step": 67250
+    },
+    {
+      "epoch": 19.598171122371717,
+      "grad_norm": 0.38619598746299744,
+      "learning_rate": 0.0003650384615384615,
+      "loss": 3.2514,
+      "step": 67300
+    },
+    {
+      "epoch": 19.612732249985438,
+      "grad_norm": 0.399539053440094,
+      "learning_rate": 0.00036486363636363637,
+      "loss": 3.2456,
+      "step": 67350
+    },
+    {
+      "epoch": 19.627293377599162,
+      "grad_norm": 0.41010504961013794,
+      "learning_rate": 0.00036468881118881117,
+      "loss": 3.2411,
+      "step": 67400
+    },
+    {
+      "epoch": 19.641854505212883,
+      "grad_norm": 0.40376871824264526,
+      "learning_rate": 0.000364513986013986,
+      "loss": 3.25,
+      "step": 67450
+    },
+    {
+      "epoch": 19.656415632826604,
+      "grad_norm": 0.3830404579639435,
+      "learning_rate": 0.0003643391608391608,
+      "loss": 3.2437,
+      "step": 67500
+    },
+    {
+      "epoch": 19.67097676044033,
+      "grad_norm": 0.4260205030441284,
+      "learning_rate": 0.0003641643356643357,
+      "loss": 3.2468,
+      "step": 67550
+    },
+    {
+      "epoch": 19.68553788805405,
+      "grad_norm": 0.386489599943161,
+      "learning_rate": 0.0003639895104895104,
+      "loss": 3.2339,
+      "step": 67600
+    },
+    {
+      "epoch": 19.700099015667774,
+      "grad_norm": 0.3752810060977936,
+      "learning_rate": 0.0003638146853146853,
+      "loss": 3.2543,
+      "step": 67650
+    },
+    {
+      "epoch": 19.714660143281495,
+      "grad_norm": 0.40031254291534424,
+      "learning_rate": 0.00036363986013986007,
+      "loss": 3.2511,
+      "step": 67700
+    },
+    {
+      "epoch": 19.729221270895216,
+      "grad_norm": 0.38260605931282043,
+      "learning_rate": 0.0003634650349650349,
+      "loss": 3.2462,
+      "step": 67750
+    },
+    {
+      "epoch": 19.74378239850894,
+      "grad_norm": 0.4128180742263794,
+      "learning_rate": 0.0003632902097902097,
+      "loss": 3.2518,
+      "step": 67800
+    },
+    {
+      "epoch": 19.758343526122662,
+      "grad_norm": 0.3811860978603363,
+      "learning_rate": 0.0003631153846153846,
+      "loss": 3.2542,
+      "step": 67850
+    },
+    {
+      "epoch": 19.772904653736386,
+      "grad_norm": 0.3886754512786865,
+      "learning_rate": 0.00036294055944055943,
+      "loss": 3.256,
+      "step": 67900
+    },
+    {
+      "epoch": 19.787465781350107,
+      "grad_norm": 0.41729703545570374,
+      "learning_rate": 0.00036276573426573423,
+      "loss": 3.2541,
+      "step": 67950
+    },
+    {
+      "epoch": 19.802026908963832,
+      "grad_norm": 0.37454211711883545,
+      "learning_rate": 0.0003625909090909091,
+      "loss": 3.2708,
+      "step": 68000
+    },
+    {
+      "epoch": 19.802026908963832,
+      "eval_accuracy": 0.3732008512473275,
+      "eval_loss": 3.5364534854888916,
+      "eval_runtime": 176.7504,
+      "eval_samples_per_second": 94.178,
+      "eval_steps_per_second": 5.89,
+      "step": 68000
+    },
+    {
+      "epoch": 19.816588036577553,
+      "grad_norm": 0.41786983609199524,
+      "learning_rate": 0.0003624160839160839,
+      "loss": 3.2541,
+      "step": 68050
+    },
+    {
+      "epoch": 19.831149164191274,
+      "grad_norm": 0.3999040722846985,
+      "learning_rate": 0.00036224125874125874,
+      "loss": 3.2462,
+      "step": 68100
+    },
+    {
+      "epoch": 19.845710291805,
+      "grad_norm": 0.38314878940582275,
+      "learning_rate": 0.00036206643356643354,
+      "loss": 3.2512,
+      "step": 68150
+    },
+    {
+      "epoch": 19.86027141941872,
+      "grad_norm": 0.3930557370185852,
+      "learning_rate": 0.0003618916083916084,
+      "loss": 3.2551,
+      "step": 68200
+    },
+    {
+      "epoch": 19.874832547032444,
+      "grad_norm": 0.38350459933280945,
+      "learning_rate": 0.0003617167832167832,
+      "loss": 3.2679,
+      "step": 68250
+    },
+    {
+      "epoch": 19.889393674646165,
+      "grad_norm": 0.395722895860672,
+      "learning_rate": 0.00036154195804195805,
+      "loss": 3.2665,
+      "step": 68300
+    },
+    {
+      "epoch": 19.903954802259886,
+      "grad_norm": 0.4258788526058197,
+      "learning_rate": 0.0003613671328671328,
+      "loss": 3.2693,
+      "step": 68350
+    },
+    {
+      "epoch": 19.91851592987361,
+      "grad_norm": 0.408592164516449,
+      "learning_rate": 0.00036119230769230765,
+      "loss": 3.253,
+      "step": 68400
+    },
+    {
+      "epoch": 19.93307705748733,
+      "grad_norm": 0.36436349153518677,
+      "learning_rate": 0.00036101748251748245,
+      "loss": 3.2564,
+      "step": 68450
+    },
+    {
+      "epoch": 19.947638185101056,
+      "grad_norm": 0.36261311173439026,
+      "learning_rate": 0.0003608426573426573,
+      "loss": 3.2638,
+      "step": 68500
+    },
+    {
+      "epoch": 19.962199312714777,
+      "grad_norm": 0.40500375628471375,
+      "learning_rate": 0.0003606678321678321,
+      "loss": 3.2594,
+      "step": 68550
+    },
+    {
+      "epoch": 19.976760440328498,
+      "grad_norm": 0.400728702545166,
+      "learning_rate": 0.00036049300699300696,
+      "loss": 3.2556,
+      "step": 68600
+    },
+    {
+      "epoch": 19.991321567942222,
+      "grad_norm": 0.38320621848106384,
+      "learning_rate": 0.0003603181818181818,
+      "loss": 3.2556,
+      "step": 68650
+    },
+    {
+      "epoch": 20.005824451045488,
+      "grad_norm": 0.387529581785202,
+      "learning_rate": 0.0003601433566433566,
+      "loss": 3.2309,
+      "step": 68700
+    },
+    {
+      "epoch": 20.020385578659212,
+      "grad_norm": 0.37018364667892456,
+      "learning_rate": 0.00035996853146853146,
+      "loss": 3.1433,
+      "step": 68750
+    },
+    {
+      "epoch": 20.034946706272933,
+      "grad_norm": 0.4018804728984833,
+      "learning_rate": 0.00035979370629370626,
+      "loss": 3.1541,
+      "step": 68800
+    },
+    {
+      "epoch": 20.049507833886658,
+      "grad_norm": 0.4254935681819916,
+      "learning_rate": 0.0003596188811188811,
+      "loss": 3.161,
+      "step": 68850
+    },
+    {
+      "epoch": 20.06406896150038,
+      "grad_norm": 0.3899883031845093,
+      "learning_rate": 0.0003594440559440559,
+      "loss": 3.1717,
+      "step": 68900
+    },
+    {
+      "epoch": 20.0786300891141,
+      "grad_norm": 0.41561880707740784,
+      "learning_rate": 0.00035926923076923077,
+      "loss": 3.1755,
+      "step": 68950
+    },
+    {
+      "epoch": 20.093191216727824,
+      "grad_norm": 0.3727256655693054,
+      "learning_rate": 0.00035909440559440557,
+      "loss": 3.1756,
+      "step": 69000
+    },
+    {
+      "epoch": 20.093191216727824,
+      "eval_accuracy": 0.37314242263146025,
+      "eval_loss": 3.547905683517456,
+      "eval_runtime": 176.1962,
+      "eval_samples_per_second": 94.474,
+      "eval_steps_per_second": 5.908,
+      "step": 69000
+    },
+    {
+      "epoch": 20.107752344341545,
+      "grad_norm": 0.41738736629486084,
+      "learning_rate": 0.0003589195804195804,
+      "loss": 3.1825,
+      "step": 69050
+    },
+    {
+      "epoch": 20.12231347195527,
+      "grad_norm": 0.38791269063949585,
+      "learning_rate": 0.00035874475524475517,
+      "loss": 3.1829,
+      "step": 69100
+    },
+    {
+      "epoch": 20.13687459956899,
+      "grad_norm": 0.3935379981994629,
+      "learning_rate": 0.00035856993006993,
+      "loss": 3.1919,
+      "step": 69150
+    },
+    {
+      "epoch": 20.15143572718271,
+      "grad_norm": 0.3939097821712494,
+      "learning_rate": 0.0003583951048951048,
+      "loss": 3.1868,
+      "step": 69200
+    },
+    {
+      "epoch": 20.165996854796436,
+      "grad_norm": 0.41222381591796875,
+      "learning_rate": 0.0003582202797202797,
+      "loss": 3.1935,
+      "step": 69250
+    },
+    {
+      "epoch": 20.180557982410157,
+      "grad_norm": 0.3910202085971832,
+      "learning_rate": 0.00035804545454545453,
+      "loss": 3.189,
+      "step": 69300
+    },
+    {
+      "epoch": 20.19511911002388,
+      "grad_norm": 0.3945898711681366,
+      "learning_rate": 0.00035787062937062933,
+      "loss": 3.1866,
+      "step": 69350
+    },
+    {
+      "epoch": 20.209680237637603,
+      "grad_norm": 0.41844651103019714,
+      "learning_rate": 0.0003576958041958042,
+      "loss": 3.1961,
+      "step": 69400
+    },
+    {
+      "epoch": 20.224241365251324,
+      "grad_norm": 0.39506423473358154,
+      "learning_rate": 0.000357520979020979,
+      "loss": 3.2082,
+      "step": 69450
+    },
+    {
+      "epoch": 20.238802492865048,
+      "grad_norm": 0.4030466079711914,
+      "learning_rate": 0.00035734615384615384,
+      "loss": 3.1928,
+      "step": 69500
+    },
+    {
+      "epoch": 20.25336362047877,
+      "grad_norm": 0.3733743131160736,
+      "learning_rate": 0.00035717132867132864,
+      "loss": 3.1989,
+      "step": 69550
+    },
+    {
+      "epoch": 20.267924748092494,
+      "grad_norm": 0.39910760521888733,
+      "learning_rate": 0.0003569965034965035,
+      "loss": 3.2017,
+      "step": 69600
+    },
+    {
+      "epoch": 20.282485875706215,
+      "grad_norm": 0.4016686975955963,
+      "learning_rate": 0.0003568216783216783,
+      "loss": 3.208,
+      "step": 69650
+    },
+    {
+      "epoch": 20.297047003319935,
+      "grad_norm": 0.4043295383453369,
+      "learning_rate": 0.00035664685314685314,
+      "loss": 3.1946,
+      "step": 69700
+    },
+    {
+      "epoch": 20.31160813093366,
+      "grad_norm": 0.43771249055862427,
+      "learning_rate": 0.00035647202797202794,
+      "loss": 3.2096,
+      "step": 69750
+    },
+    {
+      "epoch": 20.32616925854738,
+      "grad_norm": 0.40544554591178894,
+      "learning_rate": 0.0003562972027972028,
+      "loss": 3.2098,
+      "step": 69800
+    },
+    {
+      "epoch": 20.340730386161106,
+      "grad_norm": 0.3885228931903839,
+      "learning_rate": 0.00035612237762237754,
+      "loss": 3.2138,
+      "step": 69850
+    },
+    {
+      "epoch": 20.355291513774826,
+      "grad_norm": 0.375761479139328,
+      "learning_rate": 0.0003559475524475524,
+      "loss": 3.2018,
+      "step": 69900
+    },
+    {
+      "epoch": 20.369852641388547,
+      "grad_norm": 0.3793325424194336,
+      "learning_rate": 0.0003557727272727272,
+      "loss": 3.2129,
+      "step": 69950
+    },
+    {
+      "epoch": 20.384413769002272,
+      "grad_norm": 0.4034629762172699,
+      "learning_rate": 0.00035559790209790205,
+      "loss": 3.2256,
+      "step": 70000
+    },
+    {
+      "epoch": 20.384413769002272,
+      "eval_accuracy": 0.3726505406821876,
+      "eval_loss": 3.5475194454193115,
+      "eval_runtime": 176.6033,
+      "eval_samples_per_second": 94.256,
+      "eval_steps_per_second": 5.895,
+      "step": 70000
+    },
+    {
+      "epoch": 20.398974896615993,
+      "grad_norm": 0.378439724445343,
+      "learning_rate": 0.0003554230769230769,
+      "loss": 3.2109,
+      "step": 70050
+    },
+    {
+      "epoch": 20.413536024229717,
+      "grad_norm": 0.3883298933506012,
+      "learning_rate": 0.0003552482517482517,
+      "loss": 3.2145,
+      "step": 70100
+    },
+    {
+      "epoch": 20.42809715184344,
+      "grad_norm": 0.37452587485313416,
+      "learning_rate": 0.00035507342657342656,
+      "loss": 3.2067,
+      "step": 70150
+    },
+    {
+      "epoch": 20.442658279457163,
+      "grad_norm": 0.408031165599823,
+      "learning_rate": 0.00035489860139860136,
+      "loss": 3.2018,
+      "step": 70200
+    },
+    {
+      "epoch": 20.457219407070884,
+      "grad_norm": 0.39902517199516296,
+      "learning_rate": 0.0003547237762237762,
+      "loss": 3.213,
+      "step": 70250
+    },
+    {
+      "epoch": 20.471780534684605,
+      "grad_norm": 0.41096633672714233,
+      "learning_rate": 0.000354548951048951,
+      "loss": 3.224,
+      "step": 70300
+    },
+    {
+      "epoch": 20.48634166229833,
+      "grad_norm": 0.3833194673061371,
+      "learning_rate": 0.00035437412587412587,
+      "loss": 3.2183,
+      "step": 70350
+    },
+    {
+      "epoch": 20.50090278991205,
+      "grad_norm": 0.43597906827926636,
+      "learning_rate": 0.00035419930069930067,
+      "loss": 3.2267,
+      "step": 70400
+    },
+    {
+      "epoch": 20.51546391752577,
+      "grad_norm": 0.38731446862220764,
+      "learning_rate": 0.0003540244755244755,
+      "loss": 3.2153,
+      "step": 70450
+    },
+    {
+      "epoch": 20.530025045139496,
+      "grad_norm": 0.40491819381713867,
+      "learning_rate": 0.0003538496503496503,
+      "loss": 3.231,
+      "step": 70500
+    },
+    {
+      "epoch": 20.544586172753217,
+      "grad_norm": 0.394074946641922,
+      "learning_rate": 0.0003536748251748252,
+      "loss": 3.2364,
+      "step": 70550
+    },
+    {
+      "epoch": 20.55914730036694,
+      "grad_norm": 0.38757994771003723,
+      "learning_rate": 0.0003534999999999999,
+      "loss": 3.2268,
+      "step": 70600
+    },
+    {
+      "epoch": 20.573708427980662,
+      "grad_norm": 0.4309883415699005,
+      "learning_rate": 0.00035332517482517477,
+      "loss": 3.233,
+      "step": 70650
+    },
+    {
+      "epoch": 20.588269555594387,
+      "grad_norm": 0.3867761194705963,
+      "learning_rate": 0.0003531503496503496,
+      "loss": 3.2245,
+      "step": 70700
+    },
+    {
+      "epoch": 20.602830683208108,
+      "grad_norm": 0.40435120463371277,
+      "learning_rate": 0.0003529755244755244,
+      "loss": 3.2288,
+      "step": 70750
+    },
+    {
+      "epoch": 20.61739181082183,
+      "grad_norm": 0.3826408088207245,
+      "learning_rate": 0.0003528006993006993,
+      "loss": 3.231,
+      "step": 70800
+    },
+    {
+      "epoch": 20.631952938435553,
+      "grad_norm": 0.3879588842391968,
+      "learning_rate": 0.0003526258741258741,
+      "loss": 3.2358,
+      "step": 70850
+    },
+    {
+      "epoch": 20.646514066049274,
+      "grad_norm": 0.3931640386581421,
+      "learning_rate": 0.00035245104895104893,
+      "loss": 3.2322,
+      "step": 70900
+    },
+    {
+      "epoch": 20.661075193663,
+      "grad_norm": 0.42434677481651306,
+      "learning_rate": 0.00035227622377622373,
+      "loss": 3.2263,
+      "step": 70950
+    },
+    {
+      "epoch": 20.67563632127672,
+      "grad_norm": 0.4020628333091736,
+      "learning_rate": 0.0003521013986013986,
+      "loss": 3.2426,
+      "step": 71000
+    },
+    {
+      "epoch": 20.67563632127672,
+      "eval_accuracy": 0.37326292430402347,
+      "eval_loss": 3.540471315383911,
+      "eval_runtime": 176.6008,
+      "eval_samples_per_second": 94.258,
+      "eval_steps_per_second": 5.895,
+      "step": 71000
+    },
+    {
+      "epoch": 20.69019744889044,
+      "grad_norm": 0.3879493474960327,
+      "learning_rate": 0.0003519265734265734,
+      "loss": 3.2328,
+      "step": 71050
+    },
+    {
+      "epoch": 20.704758576504165,
+      "grad_norm": 0.41016706824302673,
+      "learning_rate": 0.00035175174825174824,
+      "loss": 3.2518,
+      "step": 71100
+    },
+    {
+      "epoch": 20.719319704117886,
+      "grad_norm": 0.41510647535324097,
+      "learning_rate": 0.00035157692307692304,
+      "loss": 3.2364,
+      "step": 71150
+    },
+    {
+      "epoch": 20.73388083173161,
+      "grad_norm": 0.37887150049209595,
+      "learning_rate": 0.0003514020979020979,
+      "loss": 3.2328,
+      "step": 71200
+    },
+    {
+      "epoch": 20.74844195934533,
+      "grad_norm": 0.40134233236312866,
+      "learning_rate": 0.0003512272727272727,
+      "loss": 3.2413,
+      "step": 71250
+    },
+    {
+      "epoch": 20.763003086959053,
+      "grad_norm": 0.3549582064151764,
+      "learning_rate": 0.00035105244755244755,
+      "loss": 3.2469,
+      "step": 71300
+    },
+    {
+      "epoch": 20.777564214572777,
+      "grad_norm": 0.40901145339012146,
+      "learning_rate": 0.0003508776223776223,
+      "loss": 3.2349,
+      "step": 71350
+    },
+    {
+      "epoch": 20.792125342186498,
+      "grad_norm": 0.3721853494644165,
+      "learning_rate": 0.00035070279720279715,
+      "loss": 3.2393,
+      "step": 71400
+    },
+    {
+      "epoch": 20.806686469800223,
+      "grad_norm": 0.3537469208240509,
+      "learning_rate": 0.000350527972027972,
+      "loss": 3.246,
+      "step": 71450
+    },
+    {
+      "epoch": 20.821247597413944,
+      "grad_norm": 0.3981519341468811,
+      "learning_rate": 0.0003503531468531468,
+      "loss": 3.2462,
+      "step": 71500
+    },
+    {
+      "epoch": 20.835808725027665,
+      "grad_norm": 0.3936062455177307,
+      "learning_rate": 0.00035017832167832166,
+      "loss": 3.2562,
+      "step": 71550
+    },
+    {
+      "epoch": 20.85036985264139,
+      "grad_norm": 0.4099869430065155,
+      "learning_rate": 0.00035000349650349645,
+      "loss": 3.2435,
+      "step": 71600
+    },
+    {
+      "epoch": 20.86493098025511,
+      "grad_norm": 0.40036559104919434,
+      "learning_rate": 0.0003498286713286713,
+      "loss": 3.2392,
+      "step": 71650
+    },
+    {
+      "epoch": 20.879492107868835,
+      "grad_norm": 0.3792988955974579,
+      "learning_rate": 0.0003496538461538461,
+      "loss": 3.245,
+      "step": 71700
+    },
+    {
+      "epoch": 20.894053235482556,
+      "grad_norm": 0.3893389403820038,
+      "learning_rate": 0.00034947902097902096,
+      "loss": 3.258,
+      "step": 71750
+    },
+    {
+      "epoch": 20.908614363096277,
+      "grad_norm": 0.40422552824020386,
+      "learning_rate": 0.00034930419580419576,
+      "loss": 3.2478,
+      "step": 71800
+    },
+    {
+      "epoch": 20.92317549071,
+      "grad_norm": 0.36332860589027405,
+      "learning_rate": 0.0003491293706293706,
+      "loss": 3.2579,
+      "step": 71850
+    },
+    {
+      "epoch": 20.937736618323722,
+      "grad_norm": 0.41438668966293335,
+      "learning_rate": 0.0003489545454545454,
+      "loss": 3.2507,
+      "step": 71900
+    },
+    {
+      "epoch": 20.952297745937447,
+      "grad_norm": 0.3990573287010193,
+      "learning_rate": 0.00034877972027972027,
+      "loss": 3.2588,
+      "step": 71950
+    },
+    {
+      "epoch": 20.966858873551168,
+      "grad_norm": 0.40813231468200684,
+      "learning_rate": 0.00034860489510489507,
+      "loss": 3.2544,
+      "step": 72000
+    },
+    {
+      "epoch": 20.966858873551168,
+      "eval_accuracy": 0.3740354281971092,
+      "eval_loss": 3.5308823585510254,
+      "eval_runtime": 176.4619,
+      "eval_samples_per_second": 94.332,
+      "eval_steps_per_second": 5.899,
+      "step": 72000
+    },
+    {
+      "epoch": 20.98142000116489,
+      "grad_norm": 0.40066397190093994,
+      "learning_rate": 0.0003484300699300699,
+      "loss": 3.246,
+      "step": 72050
+    },
+    {
+      "epoch": 20.995981128778613,
+      "grad_norm": 0.36475440859794617,
+      "learning_rate": 0.0003482552447552448,
+      "loss": 3.2523,
+      "step": 72100
+    },
+    {
+      "epoch": 21.01048401188188,
+      "grad_norm": 0.40291792154312134,
+      "learning_rate": 0.0003480804195804195,
+      "loss": 3.1823,
+      "step": 72150
+    },
+    {
+      "epoch": 21.025045139495603,
+      "grad_norm": 0.39930352568626404,
+      "learning_rate": 0.0003479055944055944,
+      "loss": 3.1448,
+      "step": 72200
+    },
+    {
+      "epoch": 21.039606267109324,
+      "grad_norm": 0.4272351861000061,
+      "learning_rate": 0.0003477307692307692,
+      "loss": 3.1528,
+      "step": 72250
+    },
+    {
+      "epoch": 21.05416739472305,
+      "grad_norm": 0.37879079580307007,
+      "learning_rate": 0.00034755594405594403,
+      "loss": 3.1626,
+      "step": 72300
+    },
+    {
+      "epoch": 21.06872852233677,
+      "grad_norm": 0.37789633870124817,
+      "learning_rate": 0.00034738111888111883,
+      "loss": 3.1653,
+      "step": 72350
+    },
+    {
+      "epoch": 21.08328964995049,
+      "grad_norm": 0.3773847818374634,
+      "learning_rate": 0.0003472062937062937,
+      "loss": 3.1581,
+      "step": 72400
+    },
+    {
+      "epoch": 21.097850777564215,
+      "grad_norm": 0.39947396516799927,
+      "learning_rate": 0.0003470314685314685,
+      "loss": 3.1668,
+      "step": 72450
+    },
+    {
+      "epoch": 21.112411905177936,
+      "grad_norm": 0.42217087745666504,
+      "learning_rate": 0.00034685664335664334,
+      "loss": 3.1779,
+      "step": 72500
+    },
+    {
+      "epoch": 21.12697303279166,
+      "grad_norm": 0.40650278329849243,
+      "learning_rate": 0.00034668181818181814,
+      "loss": 3.161,
+      "step": 72550
+    },
+    {
+      "epoch": 21.14153416040538,
+      "grad_norm": 0.4245632588863373,
+      "learning_rate": 0.000346506993006993,
+      "loss": 3.1737,
+      "step": 72600
+    },
+    {
+      "epoch": 21.156095288019102,
+      "grad_norm": 0.42671096324920654,
+      "learning_rate": 0.0003463321678321678,
+      "loss": 3.1808,
+      "step": 72650
+    },
+    {
+      "epoch": 21.170656415632827,
+      "grad_norm": 0.38275977969169617,
+      "learning_rate": 0.00034615734265734264,
+      "loss": 3.179,
+      "step": 72700
+    },
+    {
+      "epoch": 21.185217543246548,
+      "grad_norm": 0.37691304087638855,
+      "learning_rate": 0.0003459825174825175,
+      "loss": 3.1855,
+      "step": 72750
+    },
+    {
+      "epoch": 21.199778670860272,
+      "grad_norm": 0.4288446009159088,
+      "learning_rate": 0.0003458076923076923,
+      "loss": 3.1851,
+      "step": 72800
+    },
+    {
+      "epoch": 21.214339798473993,
+      "grad_norm": 0.39425280690193176,
+      "learning_rate": 0.00034563286713286715,
+      "loss": 3.1851,
+      "step": 72850
+    },
+    {
+      "epoch": 21.228900926087718,
+      "grad_norm": 0.4057941138744354,
+      "learning_rate": 0.0003454580419580419,
+      "loss": 3.1758,
+      "step": 72900
+    },
+    {
+      "epoch": 21.24346205370144,
+      "grad_norm": 0.43333491683006287,
+      "learning_rate": 0.00034528321678321675,
+      "loss": 3.1818,
+      "step": 72950
+    },
+    {
+      "epoch": 21.25802318131516,
+      "grad_norm": 0.3756488859653473,
+      "learning_rate": 0.00034510839160839155,
+      "loss": 3.1981,
+      "step": 73000
+    },
+    {
+      "epoch": 21.25802318131516,
+      "eval_accuracy": 0.3729420959484869,
+      "eval_loss": 3.5505897998809814,
+      "eval_runtime": 179.1,
+      "eval_samples_per_second": 92.942,
+      "eval_steps_per_second": 5.812,
+      "step": 73000
+    },
+    {
+      "epoch": 21.272584308928884,
+      "grad_norm": 0.38150060176849365,
+      "learning_rate": 0.0003449335664335664,
+      "loss": 3.1967,
+      "step": 73050
+    },
+    {
+      "epoch": 21.287145436542605,
+      "grad_norm": 0.38942277431488037,
+      "learning_rate": 0.0003447587412587412,
+      "loss": 3.197,
+      "step": 73100
+    },
+    {
+      "epoch": 21.30170656415633,
+      "grad_norm": 0.40391889214515686,
+      "learning_rate": 0.00034458391608391606,
+      "loss": 3.1933,
+      "step": 73150
+    },
+    {
+      "epoch": 21.31626769177005,
+      "grad_norm": 0.4259825646877289,
+      "learning_rate": 0.00034440909090909086,
+      "loss": 3.1901,
+      "step": 73200
+    },
+    {
+      "epoch": 21.330828819383772,
+      "grad_norm": 0.4131545126438141,
+      "learning_rate": 0.0003442342657342657,
+      "loss": 3.1913,
+      "step": 73250
+    },
+    {
+      "epoch": 21.345389946997496,
+      "grad_norm": 0.40731021761894226,
+      "learning_rate": 0.0003440594405594405,
+      "loss": 3.2106,
+      "step": 73300
+    },
+    {
+      "epoch": 21.359951074611217,
+      "grad_norm": 0.39349207282066345,
+      "learning_rate": 0.00034388461538461537,
+      "loss": 3.2047,
+      "step": 73350
+    },
+    {
+      "epoch": 21.374512202224942,
+      "grad_norm": 0.3948974907398224,
+      "learning_rate": 0.00034370979020979017,
+      "loss": 3.2043,
+      "step": 73400
+    },
+    {
+      "epoch": 21.389073329838663,
+      "grad_norm": 0.3808380961418152,
+      "learning_rate": 0.000343534965034965,
+      "loss": 3.2066,
+      "step": 73450
+    },
+    {
+      "epoch": 21.403634457452384,
+      "grad_norm": 0.3924180865287781,
+      "learning_rate": 0.0003433601398601399,
+      "loss": 3.2087,
+      "step": 73500
+    },
+    {
+      "epoch": 21.41819558506611,
+      "grad_norm": 0.38892677426338196,
+      "learning_rate": 0.0003431853146853147,
+      "loss": 3.201,
+      "step": 73550
+    },
+    {
+      "epoch": 21.43275671267983,
+      "grad_norm": 0.38498401641845703,
+      "learning_rate": 0.0003430104895104895,
+      "loss": 3.2173,
+      "step": 73600
+    },
+    {
+      "epoch": 21.447317840293554,
+      "grad_norm": 0.4188741445541382,
+      "learning_rate": 0.00034283566433566427,
+      "loss": 3.2155,
+      "step": 73650
+    },
+    {
+      "epoch": 21.461878967907275,
+      "grad_norm": 0.37732937932014465,
+      "learning_rate": 0.0003426608391608391,
+      "loss": 3.2243,
+      "step": 73700
+    },
+    {
+      "epoch": 21.476440095520996,
+      "grad_norm": 0.41169071197509766,
+      "learning_rate": 0.0003424860139860139,
+      "loss": 3.2149,
+      "step": 73750
+    },
+    {
+      "epoch": 21.49100122313472,
+      "grad_norm": 0.39298930764198303,
+      "learning_rate": 0.0003423111888111888,
+      "loss": 3.201,
+      "step": 73800
+    },
+    {
+      "epoch": 21.50556235074844,
+      "grad_norm": 0.426394522190094,
+      "learning_rate": 0.0003421363636363636,
+      "loss": 3.2131,
+      "step": 73850
+    },
+    {
+      "epoch": 21.520123478362166,
+      "grad_norm": 0.4090413451194763,
+      "learning_rate": 0.00034196153846153843,
+      "loss": 3.2134,
+      "step": 73900
+    },
+    {
+      "epoch": 21.534684605975887,
+      "grad_norm": 0.41145437955856323,
+      "learning_rate": 0.00034178671328671323,
+      "loss": 3.207,
+      "step": 73950
+    },
+    {
+      "epoch": 21.549245733589608,
+      "grad_norm": 0.4169312119483948,
+      "learning_rate": 0.0003416118881118881,
+      "loss": 3.2153,
+      "step": 74000
+    },
+    {
+      "epoch": 21.549245733589608,
+      "eval_accuracy": 0.37356447239194995,
+      "eval_loss": 3.5442752838134766,
+      "eval_runtime": 178.9213,
+      "eval_samples_per_second": 93.035,
+      "eval_steps_per_second": 5.818,
+      "step": 74000
+    },
+    {
+      "epoch": 21.563806861203332,
+      "grad_norm": 0.40608638525009155,
+      "learning_rate": 0.0003414370629370629,
+      "loss": 3.213,
+      "step": 74050
+    },
+    {
+      "epoch": 21.578367988817053,
+      "grad_norm": 0.397708922624588,
+      "learning_rate": 0.00034126223776223774,
+      "loss": 3.2146,
+      "step": 74100
+    },
+    {
+      "epoch": 21.592929116430778,
+      "grad_norm": 0.3892466723918915,
+      "learning_rate": 0.0003410874125874126,
+      "loss": 3.2289,
+      "step": 74150
+    },
+    {
+      "epoch": 21.6074902440445,
+      "grad_norm": 0.39655137062072754,
+      "learning_rate": 0.0003409125874125874,
+      "loss": 3.2232,
+      "step": 74200
+    },
+    {
+      "epoch": 21.62205137165822,
+      "grad_norm": 0.3863903284072876,
+      "learning_rate": 0.00034073776223776225,
+      "loss": 3.2263,
+      "step": 74250
+    },
+    {
+      "epoch": 21.636612499271944,
+      "grad_norm": 0.3805066645145416,
+      "learning_rate": 0.00034056293706293705,
+      "loss": 3.2317,
+      "step": 74300
+    },
+    {
+      "epoch": 21.651173626885665,
+      "grad_norm": 0.3918662965297699,
+      "learning_rate": 0.0003403881118881119,
+      "loss": 3.2376,
+      "step": 74350
+    },
+    {
+      "epoch": 21.66573475449939,
+      "grad_norm": 0.40353602170944214,
+      "learning_rate": 0.00034021328671328665,
+      "loss": 3.2326,
+      "step": 74400
+    },
+    {
+      "epoch": 21.68029588211311,
+      "grad_norm": 0.3726404309272766,
+      "learning_rate": 0.0003400384615384615,
+      "loss": 3.2354,
+      "step": 74450
+    },
+    {
+      "epoch": 21.69485700972683,
+      "grad_norm": 0.38688957691192627,
+      "learning_rate": 0.0003398636363636363,
+      "loss": 3.2209,
+      "step": 74500
+    },
+    {
+      "epoch": 21.709418137340556,
+      "grad_norm": 0.3922900855541229,
+      "learning_rate": 0.00033968881118881115,
+      "loss": 3.2291,
+      "step": 74550
+    },
+    {
+      "epoch": 21.723979264954277,
+      "grad_norm": 0.39975613355636597,
+      "learning_rate": 0.00033951398601398595,
+      "loss": 3.2215,
+      "step": 74600
+    },
+    {
+      "epoch": 21.738540392568,
+      "grad_norm": 0.4230974614620209,
+      "learning_rate": 0.0003393391608391608,
+      "loss": 3.2271,
+      "step": 74650
+    },
+    {
+      "epoch": 21.753101520181723,
+      "grad_norm": 0.3907955586910248,
+      "learning_rate": 0.0003391643356643356,
+      "loss": 3.2258,
+      "step": 74700
+    },
+    {
+      "epoch": 21.767662647795444,
+      "grad_norm": 0.4291480779647827,
+      "learning_rate": 0.00033898951048951046,
+      "loss": 3.2237,
+      "step": 74750
+    },
+    {
+      "epoch": 21.782223775409168,
+      "grad_norm": 0.3837590515613556,
+      "learning_rate": 0.00033881468531468526,
+      "loss": 3.2264,
+      "step": 74800
+    },
+    {
+      "epoch": 21.79678490302289,
+      "grad_norm": 0.4333752989768982,
+      "learning_rate": 0.0003386398601398601,
+      "loss": 3.223,
+      "step": 74850
+    },
+    {
+      "epoch": 21.811346030636614,
+      "grad_norm": 0.41991353034973145,
+      "learning_rate": 0.00033846503496503497,
+      "loss": 3.2259,
+      "step": 74900
+    },
+    {
+      "epoch": 21.825907158250335,
+      "grad_norm": 0.4055936634540558,
+      "learning_rate": 0.00033829020979020977,
+      "loss": 3.2374,
+      "step": 74950
+    },
+    {
+      "epoch": 21.840468285864056,
+      "grad_norm": 0.39700213074684143,
+      "learning_rate": 0.0003381153846153846,
+      "loss": 3.2287,
+      "step": 75000
+    },
+    {
+      "epoch": 21.840468285864056,
+      "eval_accuracy": 0.3741963714066107,
+      "eval_loss": 3.534473180770874,
+      "eval_runtime": 179.2224,
+      "eval_samples_per_second": 92.879,
+      "eval_steps_per_second": 5.808,
+      "step": 75000
+    },
+    {
+      "epoch": 21.85502941347778,
+      "grad_norm": 0.43522578477859497,
+      "learning_rate": 0.0003379405594405594,
+      "loss": 3.2402,
+      "step": 75050
+    },
+    {
+      "epoch": 21.8695905410915,
+      "grad_norm": 0.38809141516685486,
+      "learning_rate": 0.0003377657342657343,
+      "loss": 3.2153,
+      "step": 75100
+    },
+    {
+      "epoch": 21.884151668705226,
+      "grad_norm": 0.3915172219276428,
+      "learning_rate": 0.000337590909090909,
+      "loss": 3.2482,
+      "step": 75150
+    },
+    {
+      "epoch": 21.898712796318947,
+      "grad_norm": 0.3965000510215759,
+      "learning_rate": 0.0003374160839160839,
+      "loss": 3.2331,
+      "step": 75200
+    },
+    {
+      "epoch": 21.91327392393267,
+      "grad_norm": 0.39834538102149963,
+      "learning_rate": 0.0003372412587412587,
+      "loss": 3.2402,
+      "step": 75250
+    },
+    {
+      "epoch": 21.927835051546392,
+      "grad_norm": 0.3902565836906433,
+      "learning_rate": 0.00033706643356643353,
+      "loss": 3.2401,
+      "step": 75300
+    },
+    {
+      "epoch": 21.942396179160113,
+      "grad_norm": 0.38109835982322693,
+      "learning_rate": 0.00033689160839160833,
+      "loss": 3.2298,
+      "step": 75350
+    },
+    {
+      "epoch": 21.956957306773838,
+      "grad_norm": 0.3719768524169922,
+      "learning_rate": 0.0003367167832167832,
+      "loss": 3.2509,
+      "step": 75400
+    },
+    {
+      "epoch": 21.97151843438756,
+      "grad_norm": 0.42772918939590454,
+      "learning_rate": 0.000336541958041958,
+      "loss": 3.2426,
+      "step": 75450
+    },
+    {
+      "epoch": 21.986079562001283,
+      "grad_norm": 0.36617153882980347,
+      "learning_rate": 0.00033636713286713284,
+      "loss": 3.2474,
+      "step": 75500
+    },
+    {
+      "epoch": 22.00058244510455,
+      "grad_norm": 0.39934152364730835,
+      "learning_rate": 0.0003361923076923077,
+      "loss": 3.238,
+      "step": 75550
+    },
+    {
+      "epoch": 22.015143572718273,
+      "grad_norm": 0.4263432025909424,
+      "learning_rate": 0.0003360174825174825,
+      "loss": 3.1354,
+      "step": 75600
+    },
+    {
+      "epoch": 22.029704700331994,
+      "grad_norm": 0.4135393500328064,
+      "learning_rate": 0.00033584265734265734,
+      "loss": 3.1378,
+      "step": 75650
+    },
+    {
+      "epoch": 22.044265827945715,
+      "grad_norm": 0.41830018162727356,
+      "learning_rate": 0.00033566783216783214,
+      "loss": 3.1443,
+      "step": 75700
+    },
+    {
+      "epoch": 22.05882695555944,
+      "grad_norm": 0.41102901101112366,
+      "learning_rate": 0.000335493006993007,
+      "loss": 3.1422,
+      "step": 75750
+    },
+    {
+      "epoch": 22.07338808317316,
+      "grad_norm": 0.40229088068008423,
+      "learning_rate": 0.0003353181818181818,
+      "loss": 3.1599,
+      "step": 75800
+    },
+    {
+      "epoch": 22.087949210786885,
+      "grad_norm": 0.3901878297328949,
+      "learning_rate": 0.00033514335664335665,
+      "loss": 3.139,
+      "step": 75850
+    },
+    {
+      "epoch": 22.102510338400606,
+      "grad_norm": 0.3897287845611572,
+      "learning_rate": 0.0003349685314685314,
+      "loss": 3.1714,
+      "step": 75900
+    },
+    {
+      "epoch": 22.117071466014327,
+      "grad_norm": 0.3760414719581604,
+      "learning_rate": 0.00033479370629370625,
+      "loss": 3.156,
+      "step": 75950
+    },
+    {
+      "epoch": 22.13163259362805,
+      "grad_norm": 0.40278053283691406,
+      "learning_rate": 0.00033461888111888105,
+      "loss": 3.1662,
+      "step": 76000
+    },
+    {
+      "epoch": 22.13163259362805,
+      "eval_accuracy": 0.3727959656275151,
+      "eval_loss": 3.555933713912964,
+      "eval_runtime": 178.8942,
+      "eval_samples_per_second": 93.049,
+      "eval_steps_per_second": 5.819,
+      "step": 76000
+    },
+    {
+      "epoch": 22.146193721241772,
+      "grad_norm": 0.41445207595825195,
+      "learning_rate": 0.0003344440559440559,
+      "loss": 3.1631,
+      "step": 76050
+    },
+    {
+      "epoch": 22.160754848855497,
+      "grad_norm": 0.3912239074707031,
+      "learning_rate": 0.0003342692307692307,
+      "loss": 3.1725,
+      "step": 76100
+    },
+    {
+      "epoch": 22.175315976469218,
+      "grad_norm": 0.3929377794265747,
+      "learning_rate": 0.00033409440559440556,
+      "loss": 3.1578,
+      "step": 76150
+    },
+    {
+      "epoch": 22.18987710408294,
+      "grad_norm": 0.40977317094802856,
+      "learning_rate": 0.00033391958041958036,
+      "loss": 3.1738,
+      "step": 76200
+    },
+    {
+      "epoch": 22.204438231696663,
+      "grad_norm": 0.39516210556030273,
+      "learning_rate": 0.0003337447552447552,
+      "loss": 3.1761,
+      "step": 76250
+    },
+    {
+      "epoch": 22.218999359310384,
+      "grad_norm": 0.3958999216556549,
+      "learning_rate": 0.00033356993006993007,
+      "loss": 3.1694,
+      "step": 76300
+    },
+    {
+      "epoch": 22.23356048692411,
+      "grad_norm": 0.40409860014915466,
+      "learning_rate": 0.00033339510489510487,
+      "loss": 3.1816,
+      "step": 76350
+    },
+    {
+      "epoch": 22.24812161453783,
+      "grad_norm": 0.43948835134506226,
+      "learning_rate": 0.0003332202797202797,
+      "loss": 3.1861,
+      "step": 76400
+    },
+    {
+      "epoch": 22.26268274215155,
+      "grad_norm": 0.4331420660018921,
+      "learning_rate": 0.0003330454545454545,
+      "loss": 3.1809,
+      "step": 76450
+    },
+    {
+      "epoch": 22.277243869765275,
+      "grad_norm": 0.4208989441394806,
+      "learning_rate": 0.0003328706293706294,
+      "loss": 3.1821,
+      "step": 76500
+    },
+    {
+      "epoch": 22.291804997378996,
+      "grad_norm": 0.4600921869277954,
+      "learning_rate": 0.00033269580419580417,
+      "loss": 3.1938,
+      "step": 76550
+    },
+    {
+      "epoch": 22.30636612499272,
+      "grad_norm": 0.4141092598438263,
+      "learning_rate": 0.000332520979020979,
+      "loss": 3.196,
+      "step": 76600
+    },
+    {
+      "epoch": 22.32092725260644,
+      "grad_norm": 0.4022761881351471,
+      "learning_rate": 0.00033234615384615377,
+      "loss": 3.1862,
+      "step": 76650
+    },
+    {
+      "epoch": 22.335488380220163,
+      "grad_norm": 0.39649698138237,
+      "learning_rate": 0.0003321713286713286,
+      "loss": 3.1896,
+      "step": 76700
+    },
+    {
+      "epoch": 22.350049507833887,
+      "grad_norm": 0.4082602858543396,
+      "learning_rate": 0.0003319965034965034,
+      "loss": 3.1956,
+      "step": 76750
+    },
+    {
+      "epoch": 22.364610635447608,
+      "grad_norm": 0.3875502943992615,
+      "learning_rate": 0.0003318216783216783,
+      "loss": 3.1923,
+      "step": 76800
+    },
+    {
+      "epoch": 22.379171763061333,
+      "grad_norm": 0.40803077816963196,
+      "learning_rate": 0.0003316468531468531,
+      "loss": 3.1881,
+      "step": 76850
+    },
+    {
+      "epoch": 22.393732890675054,
+      "grad_norm": 0.44484943151474,
+      "learning_rate": 0.00033147202797202793,
+      "loss": 3.1941,
+      "step": 76900
+    },
+    {
+      "epoch": 22.408294018288775,
+      "grad_norm": 0.39733248949050903,
+      "learning_rate": 0.0003312972027972028,
+      "loss": 3.1924,
+      "step": 76950
+    },
+    {
+      "epoch": 22.4228551459025,
+      "grad_norm": 0.41683346033096313,
+      "learning_rate": 0.0003311223776223776,
+      "loss": 3.1907,
+      "step": 77000
+    },
+    {
+      "epoch": 22.4228551459025,
+      "eval_accuracy": 0.37371119052595864,
+      "eval_loss": 3.5440673828125,
+      "eval_runtime": 179.0008,
+      "eval_samples_per_second": 92.994,
+      "eval_steps_per_second": 5.816,
+      "step": 77000
+    },
+    {
+      "epoch": 22.43741627351622,
+      "grad_norm": 0.3999550938606262,
+      "learning_rate": 0.00033094755244755244,
+      "loss": 3.1993,
+      "step": 77050
+    },
+    {
+      "epoch": 22.451977401129945,
+      "grad_norm": 0.43484437465667725,
+      "learning_rate": 0.00033077272727272724,
+      "loss": 3.202,
+      "step": 77100
+    },
+    {
+      "epoch": 22.466538528743666,
+      "grad_norm": 0.4053047001361847,
+      "learning_rate": 0.0003305979020979021,
+      "loss": 3.1986,
+      "step": 77150
+    },
+    {
+      "epoch": 22.481099656357387,
+      "grad_norm": 0.42602503299713135,
+      "learning_rate": 0.0003304230769230769,
+      "loss": 3.1993,
+      "step": 77200
+    },
+    {
+      "epoch": 22.49566078397111,
+      "grad_norm": 0.406269907951355,
+      "learning_rate": 0.00033024825174825175,
+      "loss": 3.201,
+      "step": 77250
+    },
+    {
+      "epoch": 22.510221911584832,
+      "grad_norm": 0.38955157995224,
+      "learning_rate": 0.00033007342657342655,
+      "loss": 3.2068,
+      "step": 77300
+    },
+    {
+      "epoch": 22.524783039198557,
+      "grad_norm": 0.4480753540992737,
+      "learning_rate": 0.0003298986013986014,
+      "loss": 3.2082,
+      "step": 77350
+    },
+    {
+      "epoch": 22.539344166812278,
+      "grad_norm": 0.42242157459259033,
+      "learning_rate": 0.00032972377622377615,
+      "loss": 3.2064,
+      "step": 77400
+    },
+    {
+      "epoch": 22.553905294426002,
+      "grad_norm": 0.4058876931667328,
+      "learning_rate": 0.000329548951048951,
+      "loss": 3.2034,
+      "step": 77450
+    },
+    {
+      "epoch": 22.568466422039723,
+      "grad_norm": 0.4551337659358978,
+      "learning_rate": 0.0003293741258741258,
+      "loss": 3.2124,
+      "step": 77500
+    },
+    {
+      "epoch": 22.583027549653444,
+      "grad_norm": 0.4230758547782898,
+      "learning_rate": 0.00032919930069930065,
+      "loss": 3.2141,
+      "step": 77550
+    },
+    {
+      "epoch": 22.59758867726717,
+      "grad_norm": 0.3794611692428589,
+      "learning_rate": 0.0003290244755244755,
+      "loss": 3.2088,
+      "step": 77600
+    },
+    {
+      "epoch": 22.61214980488089,
+      "grad_norm": 0.43101099133491516,
+      "learning_rate": 0.0003288496503496503,
+      "loss": 3.2062,
+      "step": 77650
+    },
+    {
+      "epoch": 22.626710932494614,
+      "grad_norm": 0.3839839696884155,
+      "learning_rate": 0.00032867482517482516,
+      "loss": 3.2087,
+      "step": 77700
+    },
+    {
+      "epoch": 22.641272060108335,
+      "grad_norm": 0.40371859073638916,
+      "learning_rate": 0.00032849999999999996,
+      "loss": 3.2173,
+      "step": 77750
+    },
+    {
+      "epoch": 22.655833187722056,
+      "grad_norm": 0.41439804434776306,
+      "learning_rate": 0.0003283251748251748,
+      "loss": 3.2186,
+      "step": 77800
+    },
+    {
+      "epoch": 22.67039431533578,
+      "grad_norm": 0.4533017575740814,
+      "learning_rate": 0.0003281503496503496,
+      "loss": 3.2081,
+      "step": 77850
+    },
+    {
+      "epoch": 22.6849554429495,
+      "grad_norm": 0.40645161271095276,
+      "learning_rate": 0.00032797552447552447,
+      "loss": 3.2124,
+      "step": 77900
+    },
+    {
+      "epoch": 22.699516570563226,
+      "grad_norm": 0.411747545003891,
+      "learning_rate": 0.00032780069930069927,
+      "loss": 3.2292,
+      "step": 77950
+    },
+    {
+      "epoch": 22.714077698176947,
+      "grad_norm": 0.3930788040161133,
+      "learning_rate": 0.0003276258741258741,
+      "loss": 3.2248,
+      "step": 78000
+    },
+    {
+      "epoch": 22.714077698176947,
+      "eval_accuracy": 0.3740154425538548,
+      "eval_loss": 3.537041664123535,
+      "eval_runtime": 177.136,
+      "eval_samples_per_second": 93.973,
+      "eval_steps_per_second": 5.877,
+      "step": 78000
+    },
+    {
+      "epoch": 22.728638825790668,
+      "grad_norm": 0.39840519428253174,
+      "learning_rate": 0.0003274510489510489,
+      "loss": 3.2236,
+      "step": 78050
+    },
+    {
+      "epoch": 22.743199953404392,
+      "grad_norm": 0.3945090174674988,
+      "learning_rate": 0.0003272762237762238,
+      "loss": 3.2243,
+      "step": 78100
+    },
+    {
+      "epoch": 22.757761081018113,
+      "grad_norm": 0.40629005432128906,
+      "learning_rate": 0.0003271013986013985,
+      "loss": 3.2116,
+      "step": 78150
+    },
+    {
+      "epoch": 22.772322208631838,
+      "grad_norm": 0.3936091363430023,
+      "learning_rate": 0.0003269265734265734,
+      "loss": 3.2266,
+      "step": 78200
+    },
+    {
+      "epoch": 22.78688333624556,
+      "grad_norm": 0.4197155237197876,
+      "learning_rate": 0.0003267517482517482,
+      "loss": 3.2297,
+      "step": 78250
+    },
+    {
+      "epoch": 22.80144446385928,
+      "grad_norm": 0.4412369132041931,
+      "learning_rate": 0.00032657692307692303,
+      "loss": 3.2289,
+      "step": 78300
+    },
+    {
+      "epoch": 22.816005591473004,
+      "grad_norm": 0.3846014142036438,
+      "learning_rate": 0.0003264020979020979,
+      "loss": 3.2186,
+      "step": 78350
+    },
+    {
+      "epoch": 22.830566719086725,
+      "grad_norm": 0.40719151496887207,
+      "learning_rate": 0.0003262272727272727,
+      "loss": 3.2149,
+      "step": 78400
+    },
+    {
+      "epoch": 22.84512784670045,
+      "grad_norm": 0.39673471450805664,
+      "learning_rate": 0.00032605244755244754,
+      "loss": 3.2335,
+      "step": 78450
+    },
+    {
+      "epoch": 22.85968897431417,
+      "grad_norm": 0.40791651606559753,
+      "learning_rate": 0.00032587762237762234,
+      "loss": 3.2268,
+      "step": 78500
+    },
+    {
+      "epoch": 22.874250101927892,
+      "grad_norm": 0.3985656201839447,
+      "learning_rate": 0.0003257027972027972,
+      "loss": 3.2262,
+      "step": 78550
+    },
+    {
+      "epoch": 22.888811229541616,
+      "grad_norm": 0.37513139843940735,
+      "learning_rate": 0.000325527972027972,
+      "loss": 3.2292,
+      "step": 78600
+    },
+    {
+      "epoch": 22.903372357155337,
+      "grad_norm": 0.4130067825317383,
+      "learning_rate": 0.00032535314685314684,
+      "loss": 3.2287,
+      "step": 78650
+    },
+    {
+      "epoch": 22.917933484769062,
+      "grad_norm": 0.40743470191955566,
+      "learning_rate": 0.00032517832167832164,
+      "loss": 3.2296,
+      "step": 78700
+    },
+    {
+      "epoch": 22.932494612382783,
+      "grad_norm": 0.4178810119628906,
+      "learning_rate": 0.0003250034965034965,
+      "loss": 3.2217,
+      "step": 78750
+    },
+    {
+      "epoch": 22.947055739996504,
+      "grad_norm": 0.3990694582462311,
+      "learning_rate": 0.0003248286713286713,
+      "loss": 3.2269,
+      "step": 78800
+    },
+    {
+      "epoch": 22.96161686761023,
+      "grad_norm": 0.46170514822006226,
+      "learning_rate": 0.00032465384615384615,
+      "loss": 3.224,
+      "step": 78850
+    },
+    {
+      "epoch": 22.97617799522395,
+      "grad_norm": 0.42402830719947815,
+      "learning_rate": 0.0003244790209790209,
+      "loss": 3.2342,
+      "step": 78900
+    },
+    {
+      "epoch": 22.990739122837674,
+      "grad_norm": 0.37818992137908936,
+      "learning_rate": 0.00032430419580419575,
+      "loss": 3.2272,
+      "step": 78950
+    },
+    {
+      "epoch": 23.00524200594094,
+      "grad_norm": 0.40692174434661865,
+      "learning_rate": 0.00032412937062937066,
+      "loss": 3.1964,
+      "step": 79000
+    },
+    {
+      "epoch": 23.00524200594094,
+      "eval_accuracy": 0.3737579804436954,
+      "eval_loss": 3.5501067638397217,
+      "eval_runtime": 176.3895,
+      "eval_samples_per_second": 94.371,
+      "eval_steps_per_second": 5.902,
+      "step": 79000
+    },
+    {
+      "epoch": 23.019803133554664,
+      "grad_norm": 0.3965844213962555,
+      "learning_rate": 0.0003239545454545454,
+      "loss": 3.1108,
+      "step": 79050
+    },
+    {
+      "epoch": 23.034364261168385,
+      "grad_norm": 0.40269994735717773,
+      "learning_rate": 0.00032377972027972026,
+      "loss": 3.1373,
+      "step": 79100
+    },
+    {
+      "epoch": 23.048925388782106,
+      "grad_norm": 0.4315015971660614,
+      "learning_rate": 0.00032360489510489506,
+      "loss": 3.1402,
+      "step": 79150
+    },
+    {
+      "epoch": 23.06348651639583,
+      "grad_norm": 0.3808068037033081,
+      "learning_rate": 0.0003234300699300699,
+      "loss": 3.1429,
+      "step": 79200
+    },
+    {
+      "epoch": 23.07804764400955,
+      "grad_norm": 0.4001609981060028,
+      "learning_rate": 0.0003232552447552447,
+      "loss": 3.1472,
+      "step": 79250
+    },
+    {
+      "epoch": 23.092608771623276,
+      "grad_norm": 0.4144390821456909,
+      "learning_rate": 0.00032308041958041957,
+      "loss": 3.159,
+      "step": 79300
+    },
+    {
+      "epoch": 23.107169899236997,
+      "grad_norm": 0.4327448308467865,
+      "learning_rate": 0.00032290559440559437,
+      "loss": 3.1542,
+      "step": 79350
+    },
+    {
+      "epoch": 23.121731026850718,
+      "grad_norm": 0.3874760866165161,
+      "learning_rate": 0.0003227307692307692,
+      "loss": 3.158,
+      "step": 79400
+    },
+    {
+      "epoch": 23.136292154464442,
+      "grad_norm": 0.40993136167526245,
+      "learning_rate": 0.000322555944055944,
+      "loss": 3.1433,
+      "step": 79450
+    },
+    {
+      "epoch": 23.150853282078163,
+      "grad_norm": 0.421434611082077,
+      "learning_rate": 0.00032238111888111887,
+      "loss": 3.1411,
+      "step": 79500
+    },
+    {
+      "epoch": 23.165414409691888,
+      "grad_norm": 0.4032493829727173,
+      "learning_rate": 0.00032220629370629367,
+      "loss": 3.1729,
+      "step": 79550
+    },
+    {
+      "epoch": 23.17997553730561,
+      "grad_norm": 0.4178277552127838,
+      "learning_rate": 0.0003220314685314685,
+      "loss": 3.1528,
+      "step": 79600
+    },
+    {
+      "epoch": 23.19453666491933,
+      "grad_norm": 0.3926478326320648,
+      "learning_rate": 0.00032185664335664327,
+      "loss": 3.1495,
+      "step": 79650
+    },
+    {
+      "epoch": 23.209097792533054,
+      "grad_norm": 0.38440054655075073,
+      "learning_rate": 0.0003216818181818181,
+      "loss": 3.1602,
+      "step": 79700
+    },
+    {
+      "epoch": 23.223658920146775,
+      "grad_norm": 0.42098525166511536,
+      "learning_rate": 0.00032150699300699303,
+      "loss": 3.1784,
+      "step": 79750
+    },
+    {
+      "epoch": 23.2382200477605,
+      "grad_norm": 0.4336051046848297,
+      "learning_rate": 0.0003213321678321678,
+      "loss": 3.1667,
+      "step": 79800
+    },
+    {
+      "epoch": 23.25278117537422,
+      "grad_norm": 0.4008271396160126,
+      "learning_rate": 0.00032115734265734263,
+      "loss": 3.1702,
+      "step": 79850
+    },
+    {
+      "epoch": 23.26734230298794,
+      "grad_norm": 0.427082359790802,
+      "learning_rate": 0.00032098251748251743,
+      "loss": 3.1729,
+      "step": 79900
+    },
+    {
+      "epoch": 23.281903430601666,
+      "grad_norm": 0.46416816115379333,
+      "learning_rate": 0.0003208076923076923,
+      "loss": 3.1725,
+      "step": 79950
+    },
+    {
+      "epoch": 23.296464558215387,
+      "grad_norm": 0.37470319867134094,
+      "learning_rate": 0.0003206328671328671,
+      "loss": 3.1882,
+      "step": 80000
+    },
+    {
+      "epoch": 23.296464558215387,
+      "eval_accuracy": 0.3735514229425309,
+      "eval_loss": 3.5463123321533203,
+      "eval_runtime": 176.9835,
+      "eval_samples_per_second": 94.054,
+      "eval_steps_per_second": 5.882,
+      "step": 80000
+    },
+    {
+      "epoch": 23.31102568582911,
+      "grad_norm": 0.45111221075057983,
+      "learning_rate": 0.00032045804195804194,
+      "loss": 3.1322,
+      "step": 80050
+    },
+    {
+      "epoch": 23.325586813442833,
+      "grad_norm": 0.427644819021225,
+      "learning_rate": 0.00032028321678321674,
+      "loss": 3.1348,
+      "step": 80100
+    },
+    {
+      "epoch": 23.340147941056557,
+      "grad_norm": 0.4981081783771515,
+      "learning_rate": 0.0003201083916083916,
+      "loss": 3.1441,
+      "step": 80150
+    },
+    {
+      "epoch": 23.354709068670278,
+      "grad_norm": 0.42692831158638,
+      "learning_rate": 0.0003199335664335664,
+      "loss": 3.1436,
+      "step": 80200
+    },
+    {
+      "epoch": 23.369270196284,
+      "grad_norm": 0.43491700291633606,
+      "learning_rate": 0.00031975874125874125,
+      "loss": 3.144,
+      "step": 80250
+    },
+    {
+      "epoch": 23.383831323897724,
+      "grad_norm": 0.4590279757976532,
+      "learning_rate": 0.00031958391608391605,
+      "loss": 3.1532,
+      "step": 80300
+    },
+    {
+      "epoch": 23.398392451511445,
+      "grad_norm": 0.42716917395591736,
+      "learning_rate": 0.0003194090909090909,
+      "loss": 3.1646,
+      "step": 80350
+    },
+    {
+      "epoch": 23.41295357912517,
+      "grad_norm": 0.4419504702091217,
+      "learning_rate": 0.00031923426573426576,
+      "loss": 3.156,
+      "step": 80400
+    },
+    {
+      "epoch": 23.42751470673889,
+      "grad_norm": 0.42690083384513855,
+      "learning_rate": 0.0003190594405594405,
+      "loss": 3.1625,
+      "step": 80450
+    },
+    {
+      "epoch": 23.44207583435261,
+      "grad_norm": 0.41010886430740356,
+      "learning_rate": 0.0003188846153846154,
+      "loss": 3.1491,
+      "step": 80500
+    },
+    {
+      "epoch": 23.456636961966336,
+      "grad_norm": 0.41106072068214417,
+      "learning_rate": 0.00031870979020979015,
+      "loss": 3.1736,
+      "step": 80550
+    },
+    {
+      "epoch": 23.471198089580056,
+      "grad_norm": 0.4519209861755371,
+      "learning_rate": 0.000318534965034965,
+      "loss": 3.1633,
+      "step": 80600
+    },
+    {
+      "epoch": 23.48575921719378,
+      "grad_norm": 0.41818633675575256,
+      "learning_rate": 0.0003183601398601398,
+      "loss": 3.1721,
+      "step": 80650
+    },
+    {
+      "epoch": 23.500320344807502,
+      "grad_norm": 0.38743162155151367,
+      "learning_rate": 0.00031818531468531466,
+      "loss": 3.1703,
+      "step": 80700
+    },
+    {
+      "epoch": 23.514881472421223,
+      "grad_norm": 0.4087889790534973,
+      "learning_rate": 0.00031801048951048946,
+      "loss": 3.1617,
+      "step": 80750
+    },
+    {
+      "epoch": 23.529442600034947,
+      "grad_norm": 0.4340457320213318,
+      "learning_rate": 0.0003178356643356643,
+      "loss": 3.1721,
+      "step": 80800
+    },
+    {
+      "epoch": 23.54400372764867,
+      "grad_norm": 0.3856426477432251,
+      "learning_rate": 0.0003176608391608391,
+      "loss": 3.1715,
+      "step": 80850
+    },
+    {
+      "epoch": 23.558564855262393,
+      "grad_norm": 0.3838791847229004,
+      "learning_rate": 0.00031748601398601397,
+      "loss": 3.1676,
+      "step": 80900
+    },
+    {
+      "epoch": 23.573125982876114,
+      "grad_norm": 0.41913339495658875,
+      "learning_rate": 0.00031731118881118877,
+      "loss": 3.1758,
+      "step": 80950
+    },
+    {
+      "epoch": 23.587687110489835,
+      "grad_norm": 0.4226961135864258,
+      "learning_rate": 0.0003171363636363636,
+      "loss": 3.1855,
+      "step": 81000
+    },
+    {
+      "epoch": 23.587687110489835,
+      "eval_accuracy": 0.3734603119218124,
+      "eval_loss": 3.5508110523223877,
+      "eval_runtime": 182.9847,
+      "eval_samples_per_second": 90.969,
+      "eval_steps_per_second": 5.689,
+      "step": 81000
+    },
+    {
+      "epoch": 23.60224823810356,
+      "grad_norm": 0.39404693245887756,
+      "learning_rate": 0.0003169615384615385,
+      "loss": 3.1797,
+      "step": 81050
+    },
+    {
+      "epoch": 23.61680936571728,
+      "grad_norm": 0.41680702567100525,
+      "learning_rate": 0.0003167867132867133,
+      "loss": 3.1846,
+      "step": 81100
+    },
+    {
+      "epoch": 23.631370493331005,
+      "grad_norm": 0.38685619831085205,
+      "learning_rate": 0.00031661188811188813,
+      "loss": 3.1678,
+      "step": 81150
+    },
+    {
+      "epoch": 23.645931620944726,
+      "grad_norm": 0.42534857988357544,
+      "learning_rate": 0.0003164370629370629,
+      "loss": 3.1897,
+      "step": 81200
+    },
+    {
+      "epoch": 23.660492748558447,
+      "grad_norm": 0.42006826400756836,
+      "learning_rate": 0.0003162622377622378,
+      "loss": 3.1834,
+      "step": 81250
+    },
+    {
+      "epoch": 23.67505387617217,
+      "grad_norm": 0.40510988235473633,
+      "learning_rate": 0.00031608741258741253,
+      "loss": 3.1783,
+      "step": 81300
+    },
+    {
+      "epoch": 23.689615003785892,
+      "grad_norm": 0.4101458191871643,
+      "learning_rate": 0.0003159125874125874,
+      "loss": 3.1803,
+      "step": 81350
+    },
+    {
+      "epoch": 23.704176131399617,
+      "grad_norm": 0.40217939019203186,
+      "learning_rate": 0.0003157377622377622,
+      "loss": 3.1829,
+      "step": 81400
+    },
+    {
+      "epoch": 23.718737259013338,
+      "grad_norm": 0.397582083940506,
+      "learning_rate": 0.00031556293706293704,
+      "loss": 3.1848,
+      "step": 81450
+    },
+    {
+      "epoch": 23.73329838662706,
+      "grad_norm": 0.4027913808822632,
+      "learning_rate": 0.00031538811188811184,
+      "loss": 3.1967,
+      "step": 81500
+    },
+    {
+      "epoch": 23.747859514240783,
+      "grad_norm": 0.43434393405914307,
+      "learning_rate": 0.0003152132867132867,
+      "loss": 3.1933,
+      "step": 81550
+    },
+    {
+      "epoch": 23.762420641854504,
+      "grad_norm": 0.43189942836761475,
+      "learning_rate": 0.0003150384615384615,
+      "loss": 3.1976,
+      "step": 81600
+    },
+    {
+      "epoch": 23.77698176946823,
+      "grad_norm": 0.4070805013179779,
+      "learning_rate": 0.00031486363636363634,
+      "loss": 3.1962,
+      "step": 81650
+    },
+    {
+      "epoch": 23.79154289708195,
+      "grad_norm": 0.44222292304039,
+      "learning_rate": 0.00031468881118881114,
+      "loss": 3.1887,
+      "step": 81700
+    },
+    {
+      "epoch": 23.80610402469567,
+      "grad_norm": 0.45916780829429626,
+      "learning_rate": 0.000314513986013986,
+      "loss": 3.1934,
+      "step": 81750
+    },
+    {
+      "epoch": 23.820665152309395,
+      "grad_norm": 0.42806705832481384,
+      "learning_rate": 0.00031433916083916085,
+      "loss": 3.1996,
+      "step": 81800
+    },
+    {
+      "epoch": 23.835226279923116,
+      "grad_norm": 0.42251259088516235,
+      "learning_rate": 0.00031416433566433565,
+      "loss": 3.2017,
+      "step": 81850
+    },
+    {
+      "epoch": 23.84978740753684,
+      "grad_norm": 0.4552309811115265,
+      "learning_rate": 0.0003139895104895105,
+      "loss": 3.1962,
+      "step": 81900
+    },
+    {
+      "epoch": 23.86434853515056,
+      "grad_norm": 0.41143572330474854,
+      "learning_rate": 0.00031381468531468525,
+      "loss": 3.1874,
+      "step": 81950
+    },
+    {
+      "epoch": 23.878909662764286,
+      "grad_norm": 0.3958553075790405,
+      "learning_rate": 0.00031363986013986016,
+      "loss": 3.1989,
+      "step": 82000
+    },
+    {
+      "epoch": 23.878909662764286,
+      "eval_accuracy": 0.3738077094266166,
+      "eval_loss": 3.545875310897827,
+      "eval_runtime": 182.4158,
+      "eval_samples_per_second": 91.253,
+      "eval_steps_per_second": 5.707,
+      "step": 82000
+    },
+    {
+      "epoch": 23.893470790378007,
+      "grad_norm": 0.4320649802684784,
+      "learning_rate": 0.0003134650349650349,
+      "loss": 3.1949,
+      "step": 82050
+    },
+    {
+      "epoch": 23.908031917991728,
+      "grad_norm": 0.4198630750179291,
+      "learning_rate": 0.00031329020979020976,
+      "loss": 3.2004,
+      "step": 82100
+    },
+    {
+      "epoch": 23.922593045605453,
+      "grad_norm": 0.41041895747184753,
+      "learning_rate": 0.00031311538461538456,
+      "loss": 3.1914,
+      "step": 82150
+    },
+    {
+      "epoch": 23.937154173219174,
+      "grad_norm": 0.3980288803577423,
+      "learning_rate": 0.0003129405594405594,
+      "loss": 3.2107,
+      "step": 82200
+    },
+    {
+      "epoch": 23.951715300832895,
+      "grad_norm": 0.46288254857063293,
+      "learning_rate": 0.0003127657342657342,
+      "loss": 3.2086,
+      "step": 82250
+    },
+    {
+      "epoch": 23.96627642844662,
+      "grad_norm": 0.4133715033531189,
+      "learning_rate": 0.00031259090909090907,
+      "loss": 3.1936,
+      "step": 82300
+    },
+    {
+      "epoch": 23.98083755606034,
+      "grad_norm": 0.4262646734714508,
+      "learning_rate": 0.00031241608391608386,
+      "loss": 3.1925,
+      "step": 82350
+    },
+    {
+      "epoch": 23.995398683674065,
+      "grad_norm": 0.39082765579223633,
+      "learning_rate": 0.0003122412587412587,
+      "loss": 3.2045,
+      "step": 82400
+    },
+    {
+      "epoch": 24.010192789329604,
+      "grad_norm": 0.42849040031433105,
+      "learning_rate": 0.00031206643356643357,
+      "loss": 3.2134,
+      "step": 82450
+    },
+    {
+      "epoch": 24.02475391694333,
+      "grad_norm": 0.41999033093452454,
+      "learning_rate": 0.00031189160839160837,
+      "loss": 3.1282,
+      "step": 82500
+    },
+    {
+      "epoch": 24.03931504455705,
+      "grad_norm": 0.42225944995880127,
+      "learning_rate": 0.0003117167832167832,
+      "loss": 3.1347,
+      "step": 82550
+    },
+    {
+      "epoch": 24.053876172170774,
+      "grad_norm": 0.3997180163860321,
+      "learning_rate": 0.000311541958041958,
+      "loss": 3.1311,
+      "step": 82600
+    },
+    {
+      "epoch": 24.068437299784495,
+      "grad_norm": 0.4154646694660187,
+      "learning_rate": 0.0003113671328671329,
+      "loss": 3.1451,
+      "step": 82650
+    },
+    {
+      "epoch": 24.082998427398216,
+      "grad_norm": 0.42067858576774597,
+      "learning_rate": 0.0003111923076923076,
+      "loss": 3.129,
+      "step": 82700
+    },
+    {
+      "epoch": 24.09755955501194,
+      "grad_norm": 0.4242795705795288,
+      "learning_rate": 0.00031101748251748253,
+      "loss": 3.149,
+      "step": 82750
+    },
+    {
+      "epoch": 24.11212068262566,
+      "grad_norm": 0.4412245452404022,
+      "learning_rate": 0.0003108426573426573,
+      "loss": 3.1334,
+      "step": 82800
+    },
+    {
+      "epoch": 24.126681810239386,
+      "grad_norm": 0.4352100193500519,
+      "learning_rate": 0.00031066783216783213,
+      "loss": 3.1476,
+      "step": 82850
+    },
+    {
+      "epoch": 24.141242937853107,
+      "grad_norm": 0.452750563621521,
+      "learning_rate": 0.00031049300699300693,
+      "loss": 3.1419,
+      "step": 82900
+    },
+    {
+      "epoch": 24.15580406546683,
+      "grad_norm": 0.39354854822158813,
+      "learning_rate": 0.0003103181818181818,
+      "loss": 3.1623,
+      "step": 82950
+    },
+    {
+      "epoch": 24.170365193080553,
+      "grad_norm": 0.41943788528442383,
+      "learning_rate": 0.0003101433566433566,
+      "loss": 3.1544,
+      "step": 83000
+    },
+    {
+      "epoch": 24.170365193080553,
+      "eval_accuracy": 0.37311115097789754,
+      "eval_loss": 3.558357000350952,
+      "eval_runtime": 181.8256,
+      "eval_samples_per_second": 91.549,
+      "eval_steps_per_second": 5.725,
+      "step": 83000
+    },
+    {
+      "epoch": 24.184926320694274,
+      "grad_norm": 0.40159091353416443,
+      "learning_rate": 0.00030996853146853144,
+      "loss": 3.1446,
+      "step": 83050
+    },
+    {
+      "epoch": 24.199487448308,
+      "grad_norm": 0.39169952273368835,
+      "learning_rate": 0.00030979370629370624,
+      "loss": 3.1614,
+      "step": 83100
+    },
+    {
+      "epoch": 24.21404857592172,
+      "grad_norm": 0.39745545387268066,
+      "learning_rate": 0.0003096188811188811,
+      "loss": 3.1627,
+      "step": 83150
+    },
+    {
+      "epoch": 24.22860970353544,
+      "grad_norm": 0.4220676124095917,
+      "learning_rate": 0.00030944405594405595,
+      "loss": 3.1664,
+      "step": 83200
+    },
+    {
+      "epoch": 24.243170831149165,
+      "grad_norm": 0.4526214301586151,
+      "learning_rate": 0.00030926923076923075,
+      "loss": 3.1581,
+      "step": 83250
+    },
+    {
+      "epoch": 24.257731958762886,
+      "grad_norm": 0.43737250566482544,
+      "learning_rate": 0.0003090944055944056,
+      "loss": 3.1569,
+      "step": 83300
+    },
+    {
+      "epoch": 24.27229308637661,
+      "grad_norm": 0.4575173556804657,
+      "learning_rate": 0.0003089195804195804,
+      "loss": 3.1739,
+      "step": 83350
+    },
+    {
+      "epoch": 24.28685421399033,
+      "grad_norm": 0.4393577575683594,
+      "learning_rate": 0.00030874475524475525,
+      "loss": 3.1765,
+      "step": 83400
+    },
+    {
+      "epoch": 24.301415341604052,
+      "grad_norm": 0.41313502192497253,
+      "learning_rate": 0.00030856993006993,
+      "loss": 3.1567,
+      "step": 83450
+    },
+    {
+      "epoch": 24.315976469217777,
+      "grad_norm": 0.4203622043132782,
+      "learning_rate": 0.0003083951048951049,
+      "loss": 3.163,
+      "step": 83500
+    },
+    {
+      "epoch": 24.330537596831498,
+      "grad_norm": 0.41048669815063477,
+      "learning_rate": 0.00030822027972027965,
+      "loss": 3.174,
+      "step": 83550
+    },
+    {
+      "epoch": 24.345098724445222,
+      "grad_norm": 0.4165532886981964,
+      "learning_rate": 0.0003080454545454545,
+      "loss": 3.1816,
+      "step": 83600
+    },
+    {
+      "epoch": 24.359659852058943,
+      "grad_norm": 0.3939666152000427,
+      "learning_rate": 0.0003078706293706293,
+      "loss": 3.1633,
+      "step": 83650
+    },
+    {
+      "epoch": 24.374220979672664,
+      "grad_norm": 0.4266517460346222,
+      "learning_rate": 0.00030769580419580416,
+      "loss": 3.1838,
+      "step": 83700
+    },
+    {
+      "epoch": 24.38878210728639,
+      "grad_norm": 0.4121813476085663,
+      "learning_rate": 0.00030752097902097896,
+      "loss": 3.1829,
+      "step": 83750
+    },
+    {
+      "epoch": 24.40334323490011,
+      "grad_norm": 0.42097821831703186,
+      "learning_rate": 0.0003073461538461538,
+      "loss": 3.1885,
+      "step": 83800
+    },
+    {
+      "epoch": 24.417904362513834,
+      "grad_norm": 0.4097153842449188,
+      "learning_rate": 0.00030717132867132867,
+      "loss": 3.1914,
+      "step": 83850
+    },
+    {
+      "epoch": 24.432465490127555,
+      "grad_norm": 0.40335527062416077,
+      "learning_rate": 0.00030699650349650347,
+      "loss": 3.1849,
+      "step": 83900
+    },
+    {
+      "epoch": 24.44702661774128,
+      "grad_norm": 0.4203415513038635,
+      "learning_rate": 0.0003068216783216783,
+      "loss": 3.1831,
+      "step": 83950
+    },
+    {
+      "epoch": 24.461587745355,
+      "grad_norm": 0.4227919280529022,
+      "learning_rate": 0.0003066468531468531,
+      "loss": 3.1822,
+      "step": 84000
+    },
+    {
+      "epoch": 24.461587745355,
+      "eval_accuracy": 0.3738135875569855,
+      "eval_loss": 3.5504536628723145,
+      "eval_runtime": 181.79,
+      "eval_samples_per_second": 91.567,
+      "eval_steps_per_second": 5.726,
+      "step": 84000
+    },
+    {
+      "epoch": 24.47614887296872,
+      "grad_norm": 0.4112480580806732,
+      "learning_rate": 0.000306472027972028,
+      "loss": 3.1963,
+      "step": 84050
+    },
+    {
+      "epoch": 24.490710000582446,
+      "grad_norm": 0.436646431684494,
+      "learning_rate": 0.0003062972027972028,
+      "loss": 3.1987,
+      "step": 84100
+    },
+    {
+      "epoch": 24.505271128196167,
+      "grad_norm": 0.44917264580726624,
+      "learning_rate": 0.00030612237762237763,
+      "loss": 3.1892,
+      "step": 84150
+    },
+    {
+      "epoch": 24.51983225580989,
+      "grad_norm": 0.40418100357055664,
+      "learning_rate": 0.0003059475524475524,
+      "loss": 3.1847,
+      "step": 84200
+    },
+    {
+      "epoch": 24.534393383423613,
+      "grad_norm": 0.4214847981929779,
+      "learning_rate": 0.0003057727272727273,
+      "loss": 3.189,
+      "step": 84250
+    },
+    {
+      "epoch": 24.548954511037334,
+      "grad_norm": 0.4034685790538788,
+      "learning_rate": 0.00030559790209790203,
+      "loss": 3.1885,
+      "step": 84300
+    },
+    {
+      "epoch": 24.563515638651058,
+      "grad_norm": 0.42310208082199097,
+      "learning_rate": 0.0003054230769230769,
+      "loss": 3.1911,
+      "step": 84350
+    },
+    {
+      "epoch": 24.57807676626478,
+      "grad_norm": 0.42262396216392517,
+      "learning_rate": 0.0003052482517482517,
+      "loss": 3.1973,
+      "step": 84400
+    },
+    {
+      "epoch": 24.592637893878504,
+      "grad_norm": 0.42703932523727417,
+      "learning_rate": 0.00030507342657342654,
+      "loss": 3.2048,
+      "step": 84450
+    },
+    {
+      "epoch": 24.607199021492224,
+      "grad_norm": 0.40316981077194214,
+      "learning_rate": 0.00030489860139860134,
+      "loss": 3.1964,
+      "step": 84500
+    },
+    {
+      "epoch": 24.621760149105945,
+      "grad_norm": 0.4285880923271179,
+      "learning_rate": 0.0003047237762237762,
+      "loss": 3.2069,
+      "step": 84550
+    },
+    {
+      "epoch": 24.63632127671967,
+      "grad_norm": 0.43559470772743225,
+      "learning_rate": 0.00030454895104895104,
+      "loss": 3.1981,
+      "step": 84600
+    },
+    {
+      "epoch": 24.65088240433339,
+      "grad_norm": 0.424099862575531,
+      "learning_rate": 0.00030437412587412584,
+      "loss": 3.1876,
+      "step": 84650
+    },
+    {
+      "epoch": 24.665443531947115,
+      "grad_norm": 0.4152607321739197,
+      "learning_rate": 0.0003041993006993007,
+      "loss": 3.2057,
+      "step": 84700
+    },
+    {
+      "epoch": 24.680004659560836,
+      "grad_norm": 0.40943652391433716,
+      "learning_rate": 0.0003040244755244755,
+      "loss": 3.2007,
+      "step": 84750
+    },
+    {
+      "epoch": 24.694565787174557,
+      "grad_norm": 0.3958180248737335,
+      "learning_rate": 0.00030384965034965035,
+      "loss": 3.2014,
+      "step": 84800
+    },
+    {
+      "epoch": 24.709126914788282,
+      "grad_norm": 0.41481125354766846,
+      "learning_rate": 0.00030367482517482515,
+      "loss": 3.2057,
+      "step": 84850
+    },
+    {
+      "epoch": 24.723688042402003,
+      "grad_norm": 0.3919677436351776,
+      "learning_rate": 0.0003035,
+      "loss": 3.2067,
+      "step": 84900
+    },
+    {
+      "epoch": 24.738249170015727,
+      "grad_norm": 0.4104112386703491,
+      "learning_rate": 0.00030332517482517475,
+      "loss": 3.2026,
+      "step": 84950
+    },
+    {
+      "epoch": 24.75281029762945,
+      "grad_norm": 0.40666452050209045,
+      "learning_rate": 0.00030315034965034966,
+      "loss": 3.2122,
+      "step": 85000
+    },
+    {
+      "epoch": 24.75281029762945,
+      "eval_accuracy": 0.3740955026894798,
+      "eval_loss": 3.5392696857452393,
+      "eval_runtime": 182.2209,
+      "eval_samples_per_second": 91.351,
+      "eval_steps_per_second": 5.713,
+      "step": 85000
+    },
+    {
+      "epoch": 24.76737142524317,
+      "grad_norm": 0.40630656480789185,
+      "learning_rate": 0.0003029755244755244,
+      "loss": 3.2014,
+      "step": 85050
+    },
+    {
+      "epoch": 24.781932552856894,
+      "grad_norm": 0.42266345024108887,
+      "learning_rate": 0.00030280069930069926,
+      "loss": 3.1997,
+      "step": 85100
+    },
+    {
+      "epoch": 24.796493680470615,
+      "grad_norm": 0.3945619463920593,
+      "learning_rate": 0.00030262587412587406,
+      "loss": 3.203,
+      "step": 85150
+    },
+    {
+      "epoch": 24.81105480808434,
+      "grad_norm": 0.415470153093338,
+      "learning_rate": 0.0003024510489510489,
+      "loss": 3.1974,
+      "step": 85200
+    },
+    {
+      "epoch": 24.82561593569806,
+      "grad_norm": 0.4282160997390747,
+      "learning_rate": 0.00030227622377622377,
+      "loss": 3.2002,
+      "step": 85250
+    },
+    {
+      "epoch": 24.84017706331178,
+      "grad_norm": 0.4053383767604828,
+      "learning_rate": 0.00030210139860139856,
+      "loss": 3.2013,
+      "step": 85300
+    },
+    {
+      "epoch": 24.854738190925506,
+      "grad_norm": 0.3949108421802521,
+      "learning_rate": 0.0003019265734265734,
+      "loss": 3.2068,
+      "step": 85350
+    },
+    {
+      "epoch": 24.869299318539227,
+      "grad_norm": 0.4064263105392456,
+      "learning_rate": 0.0003017517482517482,
+      "loss": 3.1976,
+      "step": 85400
+    },
+    {
+      "epoch": 24.88386044615295,
+      "grad_norm": 0.4585686922073364,
+      "learning_rate": 0.00030157692307692307,
+      "loss": 3.2119,
+      "step": 85450
+    },
+    {
+      "epoch": 24.898421573766672,
+      "grad_norm": 0.424032062292099,
+      "learning_rate": 0.00030140209790209787,
+      "loss": 3.2085,
+      "step": 85500
+    },
+    {
+      "epoch": 24.912982701380393,
+      "grad_norm": 0.4028070271015167,
+      "learning_rate": 0.0003012272727272727,
+      "loss": 3.2179,
+      "step": 85550
+    },
+    {
+      "epoch": 24.927543828994118,
+      "grad_norm": 0.41038432717323303,
+      "learning_rate": 0.0003010524475524475,
+      "loss": 3.2107,
+      "step": 85600
+    },
+    {
+      "epoch": 24.94210495660784,
+      "grad_norm": 0.4204416573047638,
+      "learning_rate": 0.0003008776223776224,
+      "loss": 3.2166,
+      "step": 85650
+    },
+    {
+      "epoch": 24.956666084221563,
+      "grad_norm": 0.4086936116218567,
+      "learning_rate": 0.0003007027972027972,
+      "loss": 3.2159,
+      "step": 85700
+    },
+    {
+      "epoch": 24.971227211835284,
+      "grad_norm": 0.39162677526474,
+      "learning_rate": 0.00030052797202797203,
+      "loss": 3.2221,
+      "step": 85750
+    },
+    {
+      "epoch": 24.985788339449005,
+      "grad_norm": 0.4139779508113861,
+      "learning_rate": 0.0003003531468531468,
+      "loss": 3.2221,
+      "step": 85800
+    },
+    {
+      "epoch": 25.000291222552274,
+      "grad_norm": 0.4680265784263611,
+      "learning_rate": 0.00030017832167832163,
+      "loss": 3.2179,
+      "step": 85850
+    },
+    {
+      "epoch": 25.014852350165995,
+      "grad_norm": 0.4011513888835907,
+      "learning_rate": 0.0003000034965034965,
+      "loss": 3.1148,
+      "step": 85900
+    },
+    {
+      "epoch": 25.02941347777972,
+      "grad_norm": 0.413959801197052,
+      "learning_rate": 0.0002998286713286713,
+      "loss": 3.113,
+      "step": 85950
+    },
+    {
+      "epoch": 25.04397460539344,
+      "grad_norm": 0.41489461064338684,
+      "learning_rate": 0.00029965384615384614,
+      "loss": 3.1227,
+      "step": 86000
+    },
+    {
+      "epoch": 25.04397460539344,
+      "eval_accuracy": 0.3735611806389434,
+      "eval_loss": 3.550731897354126,
+      "eval_runtime": 181.3258,
+      "eval_samples_per_second": 91.802,
+      "eval_steps_per_second": 5.741,
+      "step": 86000
+    },
+    {
+      "epoch": 25.058535733007165,
+      "grad_norm": 0.4093635380268097,
+      "learning_rate": 0.00029947902097902094,
+      "loss": 3.1202,
+      "step": 86050
+    },
+    {
+      "epoch": 25.073096860620886,
+      "grad_norm": 0.4468414783477783,
+      "learning_rate": 0.0002993041958041958,
+      "loss": 3.1258,
+      "step": 86100
+    },
+    {
+      "epoch": 25.087657988234607,
+      "grad_norm": 0.452058881521225,
+      "learning_rate": 0.0002991293706293706,
+      "loss": 3.1351,
+      "step": 86150
+    },
+    {
+      "epoch": 25.10221911584833,
+      "grad_norm": 0.44358283281326294,
+      "learning_rate": 0.0002989545454545454,
+      "loss": 3.1289,
+      "step": 86200
+    },
+    {
+      "epoch": 25.116780243462053,
+      "grad_norm": 0.40111038088798523,
+      "learning_rate": 0.00029877972027972025,
+      "loss": 3.1312,
+      "step": 86250
+    },
+    {
+      "epoch": 25.131341371075777,
+      "grad_norm": 0.4112999439239502,
+      "learning_rate": 0.0002986048951048951,
+      "loss": 3.1404,
+      "step": 86300
+    },
+    {
+      "epoch": 25.145902498689498,
+      "grad_norm": 0.44098880887031555,
+      "learning_rate": 0.0002984300699300699,
+      "loss": 3.1299,
+      "step": 86350
+    },
+    {
+      "epoch": 25.160463626303223,
+      "grad_norm": 0.4231182932853699,
+      "learning_rate": 0.00029825524475524475,
+      "loss": 3.1419,
+      "step": 86400
+    },
+    {
+      "epoch": 25.175024753916944,
+      "grad_norm": 0.407163143157959,
+      "learning_rate": 0.00029808041958041955,
+      "loss": 3.1487,
+      "step": 86450
+    },
+    {
+      "epoch": 25.189585881530665,
+      "grad_norm": 0.4498389959335327,
+      "learning_rate": 0.0002979055944055944,
+      "loss": 3.1463,
+      "step": 86500
+    },
+    {
+      "epoch": 25.20414700914439,
+      "grad_norm": 0.4448164105415344,
+      "learning_rate": 0.0002977307692307692,
+      "loss": 3.1494,
+      "step": 86550
+    },
+    {
+      "epoch": 25.21870813675811,
+      "grad_norm": 0.40641462802886963,
+      "learning_rate": 0.000297555944055944,
+      "loss": 3.1534,
+      "step": 86600
+    },
+    {
+      "epoch": 25.233269264371835,
+      "grad_norm": 0.4092680811882019,
+      "learning_rate": 0.00029738111888111886,
+      "loss": 3.1535,
+      "step": 86650
+    },
+    {
+      "epoch": 25.247830391985556,
+      "grad_norm": 0.41941606998443604,
+      "learning_rate": 0.00029720629370629366,
+      "loss": 3.1544,
+      "step": 86700
+    },
+    {
+      "epoch": 25.262391519599277,
+      "grad_norm": 0.43551942706108093,
+      "learning_rate": 0.0002970314685314685,
+      "loss": 3.1584,
+      "step": 86750
+    },
+    {
+      "epoch": 25.276952647213,
+      "grad_norm": 0.41969966888427734,
+      "learning_rate": 0.0002968566433566433,
+      "loss": 3.1581,
+      "step": 86800
+    },
+    {
+      "epoch": 25.291513774826722,
+      "grad_norm": 0.4162365794181824,
+      "learning_rate": 0.00029668181818181817,
+      "loss": 3.1493,
+      "step": 86850
+    },
+    {
+      "epoch": 25.306074902440447,
+      "grad_norm": 0.43325042724609375,
+      "learning_rate": 0.00029650699300699297,
+      "loss": 3.1591,
+      "step": 86900
+    },
+    {
+      "epoch": 25.320636030054168,
+      "grad_norm": 0.4242747724056244,
+      "learning_rate": 0.0002963321678321678,
+      "loss": 3.1761,
+      "step": 86950
+    },
+    {
+      "epoch": 25.33519715766789,
+      "grad_norm": 0.5050255656242371,
+      "learning_rate": 0.0002961573426573426,
+      "loss": 3.1596,
+      "step": 87000
+    },
+    {
+      "epoch": 25.33519715766789,
+      "eval_accuracy": 0.37365311459791356,
+      "eval_loss": 3.5509300231933594,
+      "eval_runtime": 181.2051,
+      "eval_samples_per_second": 91.863,
+      "eval_steps_per_second": 5.745,
+      "step": 87000
+    },
+    {
+      "epoch": 25.349758285281613,
+      "grad_norm": 0.43530839681625366,
+      "learning_rate": 0.0002959825174825175,
+      "loss": 3.16,
+      "step": 87050
+    },
+    {
+      "epoch": 25.364319412895334,
+      "grad_norm": 0.4388258755207062,
+      "learning_rate": 0.0002958076923076923,
+      "loss": 3.1526,
+      "step": 87100
+    },
+    {
+      "epoch": 25.37888054050906,
+      "grad_norm": 0.4287799596786499,
+      "learning_rate": 0.00029563286713286713,
+      "loss": 3.1797,
+      "step": 87150
+    },
+    {
+      "epoch": 25.39344166812278,
+      "grad_norm": 0.44115379452705383,
+      "learning_rate": 0.00029545804195804193,
+      "loss": 3.1749,
+      "step": 87200
+    },
+    {
+      "epoch": 25.4080027957365,
+      "grad_norm": 0.4438251256942749,
+      "learning_rate": 0.0002952832167832168,
+      "loss": 3.1676,
+      "step": 87250
+    },
+    {
+      "epoch": 25.422563923350225,
+      "grad_norm": 0.41780340671539307,
+      "learning_rate": 0.0002951083916083916,
+      "loss": 3.1613,
+      "step": 87300
+    },
+    {
+      "epoch": 25.437125050963946,
+      "grad_norm": 0.4080037772655487,
+      "learning_rate": 0.0002949335664335664,
+      "loss": 3.17,
+      "step": 87350
+    },
+    {
+      "epoch": 25.45168617857767,
+      "grad_norm": 0.44067227840423584,
+      "learning_rate": 0.00029475874125874124,
+      "loss": 3.1738,
+      "step": 87400
+    },
+    {
+      "epoch": 25.46624730619139,
+      "grad_norm": 0.4406086206436157,
+      "learning_rate": 0.00029458391608391604,
+      "loss": 3.1834,
+      "step": 87450
+    },
+    {
+      "epoch": 25.480808433805112,
+      "grad_norm": 0.3974784016609192,
+      "learning_rate": 0.0002944090909090909,
+      "loss": 3.181,
+      "step": 87500
+    },
+    {
+      "epoch": 25.495369561418837,
+      "grad_norm": 0.41356295347213745,
+      "learning_rate": 0.0002942342657342657,
+      "loss": 3.1728,
+      "step": 87550
+    },
+    {
+      "epoch": 25.509930689032558,
+      "grad_norm": 0.4187457859516144,
+      "learning_rate": 0.00029405944055944054,
+      "loss": 3.1659,
+      "step": 87600
+    },
+    {
+      "epoch": 25.524491816646282,
+      "grad_norm": 0.41289469599723816,
+      "learning_rate": 0.0002938846153846154,
+      "loss": 3.1753,
+      "step": 87650
+    },
+    {
+      "epoch": 25.539052944260003,
+      "grad_norm": 0.3981243669986725,
+      "learning_rate": 0.0002937097902097902,
+      "loss": 3.1649,
+      "step": 87700
+    },
+    {
+      "epoch": 25.553614071873724,
+      "grad_norm": 0.42982515692710876,
+      "learning_rate": 0.000293534965034965,
+      "loss": 3.1872,
+      "step": 87750
+    },
+    {
+      "epoch": 25.56817519948745,
+      "grad_norm": 0.4279654324054718,
+      "learning_rate": 0.00029336013986013985,
+      "loss": 3.1956,
+      "step": 87800
+    },
+    {
+      "epoch": 25.58273632710117,
+      "grad_norm": 0.38420966267585754,
+      "learning_rate": 0.00029318531468531465,
+      "loss": 3.1845,
+      "step": 87850
+    },
+    {
+      "epoch": 25.597297454714894,
+      "grad_norm": 0.4386531114578247,
+      "learning_rate": 0.0002930104895104895,
+      "loss": 3.1866,
+      "step": 87900
+    },
+    {
+      "epoch": 25.611858582328615,
+      "grad_norm": 0.4742956757545471,
+      "learning_rate": 0.0002928356643356643,
+      "loss": 3.1911,
+      "step": 87950
+    },
+    {
+      "epoch": 25.626419709942336,
+      "grad_norm": 0.411289244890213,
+      "learning_rate": 0.00029266083916083916,
+      "loss": 3.1876,
+      "step": 88000
+    },
+    {
+      "epoch": 25.626419709942336,
+      "eval_accuracy": 0.3742827799230341,
+      "eval_loss": 3.541804313659668,
+      "eval_runtime": 181.327,
+      "eval_samples_per_second": 91.801,
+      "eval_steps_per_second": 5.741,
+      "step": 88000
+    },
+    {
+      "epoch": 25.64098083755606,
+      "grad_norm": 0.4100117087364197,
+      "learning_rate": 0.00029248601398601396,
+      "loss": 3.196,
+      "step": 88050
+    },
+    {
+      "epoch": 25.655541965169782,
+      "grad_norm": 0.41513684391975403,
+      "learning_rate": 0.00029231118881118876,
+      "loss": 3.1885,
+      "step": 88100
+    },
+    {
+      "epoch": 25.670103092783506,
+      "grad_norm": 0.40996086597442627,
+      "learning_rate": 0.0002921363636363636,
+      "loss": 3.1824,
+      "step": 88150
+    },
+    {
+      "epoch": 25.684664220397227,
+      "grad_norm": 0.40860715508461,
+      "learning_rate": 0.0002919615384615384,
+      "loss": 3.1817,
+      "step": 88200
+    },
+    {
+      "epoch": 25.69922534801095,
+      "grad_norm": 0.4030529260635376,
+      "learning_rate": 0.00029178671328671326,
+      "loss": 3.2061,
+      "step": 88250
+    },
+    {
+      "epoch": 25.713786475624673,
+      "grad_norm": 0.40431034564971924,
+      "learning_rate": 0.00029161188811188806,
+      "loss": 3.2011,
+      "step": 88300
+    },
+    {
+      "epoch": 25.728347603238394,
+      "grad_norm": 0.4311061501502991,
+      "learning_rate": 0.0002914370629370629,
+      "loss": 3.1881,
+      "step": 88350
+    },
+    {
+      "epoch": 25.74290873085212,
+      "grad_norm": 0.4359666407108307,
+      "learning_rate": 0.00029126223776223777,
+      "loss": 3.1922,
+      "step": 88400
+    },
+    {
+      "epoch": 25.75746985846584,
+      "grad_norm": 0.43195220828056335,
+      "learning_rate": 0.00029108741258741257,
+      "loss": 3.1969,
+      "step": 88450
+    },
+    {
+      "epoch": 25.772030986079564,
+      "grad_norm": 0.4211742877960205,
+      "learning_rate": 0.00029091258741258737,
+      "loss": 3.1924,
+      "step": 88500
+    },
+    {
+      "epoch": 25.786592113693285,
+      "grad_norm": 0.40398746728897095,
+      "learning_rate": 0.0002907377622377622,
+      "loss": 3.1982,
+      "step": 88550
+    },
+    {
+      "epoch": 25.801153241307006,
+      "grad_norm": 0.4384838938713074,
+      "learning_rate": 0.000290562937062937,
+      "loss": 3.1991,
+      "step": 88600
+    },
+    {
+      "epoch": 25.81571436892073,
+      "grad_norm": 0.431005597114563,
+      "learning_rate": 0.0002903881118881119,
+      "loss": 3.2046,
+      "step": 88650
+    },
+    {
+      "epoch": 25.83027549653445,
+      "grad_norm": 0.4143125116825104,
+      "learning_rate": 0.0002902132867132867,
+      "loss": 3.1902,
+      "step": 88700
+    },
+    {
+      "epoch": 25.844836624148176,
+      "grad_norm": 0.39128798246383667,
+      "learning_rate": 0.00029003846153846153,
+      "loss": 3.2102,
+      "step": 88750
+    },
+    {
+      "epoch": 25.859397751761897,
+      "grad_norm": 0.41086676716804504,
+      "learning_rate": 0.00028986363636363633,
+      "loss": 3.1996,
+      "step": 88800
+    },
+    {
+      "epoch": 25.873958879375618,
+      "grad_norm": 0.40953710675239563,
+      "learning_rate": 0.00028968881118881113,
+      "loss": 3.1907,
+      "step": 88850
+    },
+    {
+      "epoch": 25.888520006989342,
+      "grad_norm": 0.4275113344192505,
+      "learning_rate": 0.000289513986013986,
+      "loss": 3.2055,
+      "step": 88900
+    },
+    {
+      "epoch": 25.903081134603063,
+      "grad_norm": 0.4466639757156372,
+      "learning_rate": 0.0002893391608391608,
+      "loss": 3.1968,
+      "step": 88950
+    },
+    {
+      "epoch": 25.917642262216788,
+      "grad_norm": 0.4273034632205963,
+      "learning_rate": 0.00028916433566433564,
+      "loss": 3.1966,
+      "step": 89000
+    },
+    {
+      "epoch": 25.917642262216788,
+      "eval_accuracy": 0.37498956631859515,
+      "eval_loss": 3.530644655227661,
+      "eval_runtime": 181.3443,
+      "eval_samples_per_second": 91.792,
+      "eval_steps_per_second": 5.74,
+      "step": 89000
+    },
+    {
+      "epoch": 25.93220338983051,
+      "grad_norm": 0.4303942322731018,
+      "learning_rate": 0.0002889895104895105,
+      "loss": 3.204,
+      "step": 89050
+    },
+    {
+      "epoch": 25.94676451744423,
+      "grad_norm": 0.4104004502296448,
+      "learning_rate": 0.0002888146853146853,
+      "loss": 3.2073,
+      "step": 89100
+    },
+    {
+      "epoch": 25.961325645057954,
+      "grad_norm": 0.4273345172405243,
+      "learning_rate": 0.00028863986013986015,
+      "loss": 3.1909,
+      "step": 89150
+    },
+    {
+      "epoch": 25.975886772671675,
+      "grad_norm": 0.4050820469856262,
+      "learning_rate": 0.00028846503496503495,
+      "loss": 3.207,
+      "step": 89200
+    },
+    {
+      "epoch": 25.9904479002854,
+      "grad_norm": 0.4026417136192322,
+      "learning_rate": 0.00028829020979020975,
+      "loss": 3.1941,
+      "step": 89250
+    },
+    {
+      "epoch": 26.004950783388665,
+      "grad_norm": 0.4386459290981293,
+      "learning_rate": 0.0002881153846153846,
+      "loss": 3.1719,
+      "step": 89300
+    },
+    {
+      "epoch": 26.01951191100239,
+      "grad_norm": 0.4158802330493927,
+      "learning_rate": 0.0002879405594405594,
+      "loss": 3.0844,
+      "step": 89350
+    },
+    {
+      "epoch": 26.03407303861611,
+      "grad_norm": 0.44361788034439087,
+      "learning_rate": 0.00028776573426573425,
+      "loss": 3.1136,
+      "step": 89400
+    },
+    {
+      "epoch": 26.04863416622983,
+      "grad_norm": 0.4118858277797699,
+      "learning_rate": 0.00028759090909090905,
+      "loss": 3.1005,
+      "step": 89450
+    },
+    {
+      "epoch": 26.063195293843556,
+      "grad_norm": 0.4068799316883087,
+      "learning_rate": 0.0002874160839160839,
+      "loss": 3.0999,
+      "step": 89500
+    },
+    {
+      "epoch": 26.077756421457277,
+      "grad_norm": 0.4220964312553406,
+      "learning_rate": 0.0002872412587412587,
+      "loss": 3.124,
+      "step": 89550
+    },
+    {
+      "epoch": 26.092317549071,
+      "grad_norm": 0.4196264147758484,
+      "learning_rate": 0.0002870664335664335,
+      "loss": 3.1175,
+      "step": 89600
+    },
+    {
+      "epoch": 26.106878676684723,
+      "grad_norm": 0.43385326862335205,
+      "learning_rate": 0.00028689160839160836,
+      "loss": 3.1222,
+      "step": 89650
+    },
+    {
+      "epoch": 26.121439804298443,
+      "grad_norm": 0.45031943917274475,
+      "learning_rate": 0.0002867167832167832,
+      "loss": 3.1352,
+      "step": 89700
+    },
+    {
+      "epoch": 26.136000931912168,
+      "grad_norm": 0.41545194387435913,
+      "learning_rate": 0.000286541958041958,
+      "loss": 3.1274,
+      "step": 89750
+    },
+    {
+      "epoch": 26.15056205952589,
+      "grad_norm": 0.42113929986953735,
+      "learning_rate": 0.00028636713286713287,
+      "loss": 3.1266,
+      "step": 89800
+    },
+    {
+      "epoch": 26.165123187139613,
+      "grad_norm": 0.4348122179508209,
+      "learning_rate": 0.00028619230769230767,
+      "loss": 3.1279,
+      "step": 89850
+    },
+    {
+      "epoch": 26.179684314753334,
+      "grad_norm": 0.45158812403678894,
+      "learning_rate": 0.0002860174825174825,
+      "loss": 3.125,
+      "step": 89900
+    },
+    {
+      "epoch": 26.194245442367055,
+      "grad_norm": 0.4448310136795044,
+      "learning_rate": 0.0002858426573426573,
+      "loss": 3.1502,
+      "step": 89950
+    },
+    {
+      "epoch": 26.20880656998078,
+      "grad_norm": 0.4374328553676605,
+      "learning_rate": 0.0002856678321678321,
+      "loss": 3.1457,
+      "step": 90000
+    },
+    {
+      "epoch": 26.20880656998078,
+      "eval_accuracy": 0.3740267285641632,
+      "eval_loss": 3.5505359172821045,
+      "eval_runtime": 181.0087,
+      "eval_samples_per_second": 91.962,
+      "eval_steps_per_second": 5.751,
+      "step": 90000
+    },
+    {
+      "epoch": 26.2233676975945,
+      "grad_norm": 0.42174583673477173,
+      "learning_rate": 0.000285493006993007,
+      "loss": 3.1509,
+      "step": 90050
+    },
+    {
+      "epoch": 26.237928825208225,
+      "grad_norm": 0.44193533062934875,
+      "learning_rate": 0.0002853181818181818,
+      "loss": 3.1557,
+      "step": 90100
+    },
+    {
+      "epoch": 26.252489952821946,
+      "grad_norm": 0.42890018224716187,
+      "learning_rate": 0.00028514335664335663,
+      "loss": 3.1485,
+      "step": 90150
+    },
+    {
+      "epoch": 26.267051080435667,
+      "grad_norm": 0.41337689757347107,
+      "learning_rate": 0.00028496853146853143,
+      "loss": 3.1492,
+      "step": 90200
+    },
+    {
+      "epoch": 26.281612208049392,
+      "grad_norm": 0.4084251821041107,
+      "learning_rate": 0.0002847937062937063,
+      "loss": 3.1507,
+      "step": 90250
+    },
+    {
+      "epoch": 26.296173335663113,
+      "grad_norm": 0.42145994305610657,
+      "learning_rate": 0.0002846188811188811,
+      "loss": 3.1506,
+      "step": 90300
+    },
+    {
+      "epoch": 26.310734463276837,
+      "grad_norm": 0.4321393668651581,
+      "learning_rate": 0.0002844440559440559,
+      "loss": 3.162,
+      "step": 90350
+    },
+    {
+      "epoch": 26.32529559089056,
+      "grad_norm": 0.4253658950328827,
+      "learning_rate": 0.00028426923076923074,
+      "loss": 3.1556,
+      "step": 90400
+    },
+    {
+      "epoch": 26.33985671850428,
+      "grad_norm": 0.456359326839447,
+      "learning_rate": 0.0002840944055944056,
+      "loss": 3.1634,
+      "step": 90450
+    },
+    {
+      "epoch": 26.354417846118004,
+      "grad_norm": 0.4094768166542053,
+      "learning_rate": 0.0002839195804195804,
+      "loss": 3.1461,
+      "step": 90500
+    },
+    {
+      "epoch": 26.368978973731725,
+      "grad_norm": 0.4343591034412384,
+      "learning_rate": 0.00028374475524475524,
+      "loss": 3.1494,
+      "step": 90550
+    },
+    {
+      "epoch": 26.38354010134545,
+      "grad_norm": 0.4192764163017273,
+      "learning_rate": 0.00028356993006993004,
+      "loss": 3.1609,
+      "step": 90600
+    },
+    {
+      "epoch": 26.39810122895917,
+      "grad_norm": 0.41687625646591187,
+      "learning_rate": 0.0002833951048951049,
+      "loss": 3.1549,
+      "step": 90650
+    },
+    {
+      "epoch": 26.41266235657289,
+      "grad_norm": 0.42564550042152405,
+      "learning_rate": 0.0002832202797202797,
+      "loss": 3.1641,
+      "step": 90700
+    },
+    {
+      "epoch": 26.427223484186616,
+      "grad_norm": 0.4025818705558777,
+      "learning_rate": 0.0002830454545454545,
+      "loss": 3.1625,
+      "step": 90750
+    },
+    {
+      "epoch": 26.441784611800337,
+      "grad_norm": 0.4263351261615753,
+      "learning_rate": 0.00028287062937062935,
+      "loss": 3.164,
+      "step": 90800
+    },
+    {
+      "epoch": 26.45634573941406,
+      "grad_norm": 0.4589676558971405,
+      "learning_rate": 0.00028269580419580415,
+      "loss": 3.1734,
+      "step": 90850
+    },
+    {
+      "epoch": 26.470906867027782,
+      "grad_norm": 0.45051443576812744,
+      "learning_rate": 0.000282520979020979,
+      "loss": 3.1644,
+      "step": 90900
+    },
+    {
+      "epoch": 26.485467994641503,
+      "grad_norm": 0.4175613820552826,
+      "learning_rate": 0.0002823461538461538,
+      "loss": 3.174,
+      "step": 90950
+    },
+    {
+      "epoch": 26.500029122255228,
+      "grad_norm": 0.4551471769809723,
+      "learning_rate": 0.00028217132867132866,
+      "loss": 3.1694,
+      "step": 91000
+    },
+    {
+      "epoch": 26.500029122255228,
+      "eval_accuracy": 0.37420977354385193,
+      "eval_loss": 3.5417368412017822,
+      "eval_runtime": 180.9394,
+      "eval_samples_per_second": 91.998,
+      "eval_steps_per_second": 5.753,
+      "step": 91000
+    },
+    {
+      "epoch": 26.51459024986895,
+      "grad_norm": 0.397365003824234,
+      "learning_rate": 0.00028199650349650346,
+      "loss": 3.166,
+      "step": 91050
+    },
+    {
+      "epoch": 26.529151377482673,
+      "grad_norm": 0.4480145573616028,
+      "learning_rate": 0.0002818216783216783,
+      "loss": 3.1662,
+      "step": 91100
+    },
+    {
+      "epoch": 26.543712505096394,
+      "grad_norm": 0.42891454696655273,
+      "learning_rate": 0.0002816468531468531,
+      "loss": 3.1695,
+      "step": 91150
+    },
+    {
+      "epoch": 26.55827363271012,
+      "grad_norm": 0.41012558341026306,
+      "learning_rate": 0.00028147202797202796,
+      "loss": 3.1757,
+      "step": 91200
+    },
+    {
+      "epoch": 26.57283476032384,
+      "grad_norm": 0.4216759502887726,
+      "learning_rate": 0.00028129720279720276,
+      "loss": 3.1814,
+      "step": 91250
+    },
+    {
+      "epoch": 26.58739588793756,
+      "grad_norm": 0.41179201006889343,
+      "learning_rate": 0.0002811223776223776,
+      "loss": 3.1793,
+      "step": 91300
+    },
+    {
+      "epoch": 26.601957015551285,
+      "grad_norm": 0.4413478374481201,
+      "learning_rate": 0.0002809475524475524,
+      "loss": 3.1764,
+      "step": 91350
+    },
+    {
+      "epoch": 26.616518143165006,
+      "grad_norm": 0.39704814553260803,
+      "learning_rate": 0.00028077272727272727,
+      "loss": 3.1747,
+      "step": 91400
+    },
+    {
+      "epoch": 26.63107927077873,
+      "grad_norm": 0.41823717951774597,
+      "learning_rate": 0.00028059790209790207,
+      "loss": 3.1668,
+      "step": 91450
+    },
+    {
+      "epoch": 26.64564039839245,
+      "grad_norm": 0.43079426884651184,
+      "learning_rate": 0.00028042307692307687,
+      "loss": 3.1834,
+      "step": 91500
+    },
+    {
+      "epoch": 26.660201526006173,
+      "grad_norm": 0.42470020055770874,
+      "learning_rate": 0.0002802482517482517,
+      "loss": 3.1938,
+      "step": 91550
+    },
+    {
+      "epoch": 26.674762653619897,
+      "grad_norm": 0.42501482367515564,
+      "learning_rate": 0.0002800734265734265,
+      "loss": 3.1755,
+      "step": 91600
+    },
+    {
+      "epoch": 26.689323781233618,
+      "grad_norm": 0.4389747679233551,
+      "learning_rate": 0.0002798986013986014,
+      "loss": 3.1832,
+      "step": 91650
+    },
+    {
+      "epoch": 26.703884908847343,
+      "grad_norm": 0.40251684188842773,
+      "learning_rate": 0.0002797237762237762,
+      "loss": 3.171,
+      "step": 91700
+    },
+    {
+      "epoch": 26.718446036461064,
+      "grad_norm": 0.4349626898765564,
+      "learning_rate": 0.00027954895104895103,
+      "loss": 3.1768,
+      "step": 91750
+    },
+    {
+      "epoch": 26.733007164074785,
+      "grad_norm": 0.41921675205230713,
+      "learning_rate": 0.0002793741258741259,
+      "loss": 3.1882,
+      "step": 91800
+    },
+    {
+      "epoch": 26.74756829168851,
+      "grad_norm": 0.41158103942871094,
+      "learning_rate": 0.0002791993006993007,
+      "loss": 3.1867,
+      "step": 91850
+    },
+    {
+      "epoch": 26.76212941930223,
+      "grad_norm": 0.4082656502723694,
+      "learning_rate": 0.0002790244755244755,
+      "loss": 3.1754,
+      "step": 91900
+    },
+    {
+      "epoch": 26.776690546915955,
+      "grad_norm": 0.44691285490989685,
+      "learning_rate": 0.00027884965034965034,
+      "loss": 3.1832,
+      "step": 91950
+    },
+    {
+      "epoch": 26.791251674529676,
+      "grad_norm": 0.42668673396110535,
+      "learning_rate": 0.00027867482517482514,
+      "loss": 3.1857,
+      "step": 92000
+    },
+    {
+      "epoch": 26.791251674529676,
+      "eval_accuracy": 0.3747800697522462,
+      "eval_loss": 3.537883996963501,
+      "eval_runtime": 181.0914,
+      "eval_samples_per_second": 91.92,
+      "eval_steps_per_second": 5.748,
+      "step": 92000
+    },
+    {
+      "epoch": 26.805812802143397,
+      "grad_norm": 0.42242157459259033,
+      "learning_rate": 0.0002785,
+      "loss": 3.1905,
+      "step": 92050
+    },
+    {
+      "epoch": 26.82037392975712,
+      "grad_norm": 0.42586851119995117,
+      "learning_rate": 0.0002783251748251748,
+      "loss": 3.1868,
+      "step": 92100
+    },
+    {
+      "epoch": 26.834935057370842,
+      "grad_norm": 0.43328920006752014,
+      "learning_rate": 0.00027815034965034965,
+      "loss": 3.191,
+      "step": 92150
+    },
+    {
+      "epoch": 26.849496184984567,
+      "grad_norm": 0.44402876496315,
+      "learning_rate": 0.00027797552447552445,
+      "loss": 3.187,
+      "step": 92200
+    },
+    {
+      "epoch": 26.864057312598288,
+      "grad_norm": 0.450183242559433,
+      "learning_rate": 0.00027780069930069925,
+      "loss": 3.1948,
+      "step": 92250
+    },
+    {
+      "epoch": 26.87861844021201,
+      "grad_norm": 0.42371490597724915,
+      "learning_rate": 0.0002776258741258741,
+      "loss": 3.1943,
+      "step": 92300
+    },
+    {
+      "epoch": 26.893179567825733,
+      "grad_norm": 0.40242645144462585,
+      "learning_rate": 0.0002774510489510489,
+      "loss": 3.1945,
+      "step": 92350
+    },
+    {
+      "epoch": 26.907740695439454,
+      "grad_norm": 0.4435162842273712,
+      "learning_rate": 0.00027727622377622375,
+      "loss": 3.1853,
+      "step": 92400
+    },
+    {
+      "epoch": 26.92230182305318,
+      "grad_norm": 0.4274364411830902,
+      "learning_rate": 0.00027710139860139855,
+      "loss": 3.1863,
+      "step": 92450
+    },
+    {
+      "epoch": 26.9368629506669,
+      "grad_norm": 0.4162507653236389,
+      "learning_rate": 0.0002769265734265734,
+      "loss": 3.1857,
+      "step": 92500
+    },
+    {
+      "epoch": 26.95142407828062,
+      "grad_norm": 0.40001586079597473,
+      "learning_rate": 0.00027675174825174826,
+      "loss": 3.2034,
+      "step": 92550
+    },
+    {
+      "epoch": 26.965985205894345,
+      "grad_norm": 0.4307563304901123,
+      "learning_rate": 0.00027657692307692306,
+      "loss": 3.183,
+      "step": 92600
+    },
+    {
+      "epoch": 26.980546333508066,
+      "grad_norm": 0.4021975100040436,
+      "learning_rate": 0.00027640209790209786,
+      "loss": 3.1857,
+      "step": 92650
+    },
+    {
+      "epoch": 26.99510746112179,
+      "grad_norm": 0.4598715901374817,
+      "learning_rate": 0.0002762272727272727,
+      "loss": 3.2017,
+      "step": 92700
+    },
+    {
+      "epoch": 27.009610344225056,
+      "grad_norm": 0.4127676784992218,
+      "learning_rate": 0.0002760524475524475,
+      "loss": 3.1279,
+      "step": 92750
+    },
+    {
+      "epoch": 27.02417147183878,
+      "grad_norm": 0.4335608184337616,
+      "learning_rate": 0.00027587762237762237,
+      "loss": 3.1031,
+      "step": 92800
+    },
+    {
+      "epoch": 27.0387325994525,
+      "grad_norm": 0.41142764687538147,
+      "learning_rate": 0.00027570279720279717,
+      "loss": 3.1039,
+      "step": 92850
+    },
+    {
+      "epoch": 27.053293727066222,
+      "grad_norm": 0.42291176319122314,
+      "learning_rate": 0.000275527972027972,
+      "loss": 3.0966,
+      "step": 92900
+    },
+    {
+      "epoch": 27.067854854679947,
+      "grad_norm": 0.41084083914756775,
+      "learning_rate": 0.0002753531468531468,
+      "loss": 3.1131,
+      "step": 92950
+    },
+    {
+      "epoch": 27.082415982293668,
+      "grad_norm": 0.45117121934890747,
+      "learning_rate": 0.0002751783216783216,
+      "loss": 3.1146,
+      "step": 93000
+    },
+    {
+      "epoch": 27.082415982293668,
+      "eval_accuracy": 0.37372600341448836,
+      "eval_loss": 3.5564258098602295,
+      "eval_runtime": 181.0502,
+      "eval_samples_per_second": 91.941,
+      "eval_steps_per_second": 5.75,
+      "step": 93000
+    },
+    {
+      "epoch": 27.096977109907392,
+      "grad_norm": 0.4476958215236664,
+      "learning_rate": 0.0002750034965034965,
+      "loss": 3.1217,
+      "step": 93050
+    },
+    {
+      "epoch": 27.111538237521113,
+      "grad_norm": 0.43004608154296875,
+      "learning_rate": 0.0002748286713286713,
+      "loss": 3.1158,
+      "step": 93100
+    },
+    {
+      "epoch": 27.126099365134834,
+      "grad_norm": 0.4326183795928955,
+      "learning_rate": 0.00027465384615384613,
+      "loss": 3.122,
+      "step": 93150
+    },
+    {
+      "epoch": 27.14066049274856,
+      "grad_norm": 0.46706339716911316,
+      "learning_rate": 0.000274479020979021,
+      "loss": 3.1226,
+      "step": 93200
+    },
+    {
+      "epoch": 27.15522162036228,
+      "grad_norm": 0.42264524102211,
+      "learning_rate": 0.0002743041958041958,
+      "loss": 3.1192,
+      "step": 93250
+    },
+    {
+      "epoch": 27.169782747976004,
+      "grad_norm": 0.45010286569595337,
+      "learning_rate": 0.00027412937062937064,
+      "loss": 3.1194,
+      "step": 93300
+    },
+    {
+      "epoch": 27.184343875589725,
+      "grad_norm": 0.44601142406463623,
+      "learning_rate": 0.00027395454545454544,
+      "loss": 3.117,
+      "step": 93350
+    },
+    {
+      "epoch": 27.19890500320345,
+      "grad_norm": 0.44697099924087524,
+      "learning_rate": 0.00027377972027972024,
+      "loss": 3.1289,
+      "step": 93400
+    },
+    {
+      "epoch": 27.21346613081717,
+      "grad_norm": 0.41724687814712524,
+      "learning_rate": 0.0002736048951048951,
+      "loss": 3.1254,
+      "step": 93450
+    },
+    {
+      "epoch": 27.228027258430892,
+      "grad_norm": 0.4229699969291687,
+      "learning_rate": 0.0002734300699300699,
+      "loss": 3.1306,
+      "step": 93500
+    },
+    {
+      "epoch": 27.242588386044616,
+      "grad_norm": 0.42787185311317444,
+      "learning_rate": 0.00027325524475524474,
+      "loss": 3.1234,
+      "step": 93550
+    },
+    {
+      "epoch": 27.257149513658337,
+      "grad_norm": 0.43008866906166077,
+      "learning_rate": 0.00027308041958041954,
+      "loss": 3.1421,
+      "step": 93600
+    },
+    {
+      "epoch": 27.271710641272062,
+      "grad_norm": 0.40983253717422485,
+      "learning_rate": 0.0002729055944055944,
+      "loss": 3.136,
+      "step": 93650
+    },
+    {
+      "epoch": 27.286271768885783,
+      "grad_norm": 0.4102419316768646,
+      "learning_rate": 0.0002727307692307692,
+      "loss": 3.1395,
+      "step": 93700
+    },
+    {
+      "epoch": 27.300832896499504,
+      "grad_norm": 0.447038471698761,
+      "learning_rate": 0.000272555944055944,
+      "loss": 3.1426,
+      "step": 93750
+    },
+    {
+      "epoch": 27.31539402411323,
+      "grad_norm": 0.44245871901512146,
+      "learning_rate": 0.00027238111888111885,
+      "loss": 3.1448,
+      "step": 93800
+    },
+    {
+      "epoch": 27.32995515172695,
+      "grad_norm": 0.442560613155365,
+      "learning_rate": 0.0002722062937062937,
+      "loss": 3.1514,
+      "step": 93850
+    },
+    {
+      "epoch": 27.344516279340674,
+      "grad_norm": 0.4968518614768982,
+      "learning_rate": 0.0002720314685314685,
+      "loss": 3.1468,
+      "step": 93900
+    },
+    {
+      "epoch": 27.359077406954395,
+      "grad_norm": 0.4427065849304199,
+      "learning_rate": 0.00027185664335664336,
+      "loss": 3.1354,
+      "step": 93950
+    },
+    {
+      "epoch": 27.373638534568116,
+      "grad_norm": 0.44690605998039246,
+      "learning_rate": 0.00027168181818181816,
+      "loss": 3.1478,
+      "step": 94000
+    },
+    {
+      "epoch": 27.373638534568116,
+      "eval_accuracy": 0.3741124317049423,
+      "eval_loss": 3.5478107929229736,
+      "eval_runtime": 181.2792,
+      "eval_samples_per_second": 91.825,
+      "eval_steps_per_second": 5.743,
+      "step": 94000
+    },
+    {
+      "epoch": 27.38819966218184,
+      "grad_norm": 0.4106026291847229,
+      "learning_rate": 0.000271506993006993,
+      "loss": 3.1407,
+      "step": 94050
+    },
+    {
+      "epoch": 27.40276078979556,
+      "grad_norm": 0.4304497539997101,
+      "learning_rate": 0.0002713321678321678,
+      "loss": 3.1485,
+      "step": 94100
+    },
+    {
+      "epoch": 27.417321917409286,
+      "grad_norm": 0.4128531217575073,
+      "learning_rate": 0.0002711573426573426,
+      "loss": 3.1525,
+      "step": 94150
+    },
+    {
+      "epoch": 27.431883045023007,
+      "grad_norm": 0.43160468339920044,
+      "learning_rate": 0.00027098251748251746,
+      "loss": 3.1553,
+      "step": 94200
+    },
+    {
+      "epoch": 27.446444172636728,
+      "grad_norm": 0.42581573128700256,
+      "learning_rate": 0.00027080769230769226,
+      "loss": 3.1617,
+      "step": 94250
+    },
+    {
+      "epoch": 27.461005300250452,
+      "grad_norm": 0.4796488285064697,
+      "learning_rate": 0.0002706328671328671,
+      "loss": 3.1466,
+      "step": 94300
+    },
+    {
+      "epoch": 27.475566427864173,
+      "grad_norm": 0.45811060070991516,
+      "learning_rate": 0.0002704580419580419,
+      "loss": 3.1417,
+      "step": 94350
+    },
+    {
+      "epoch": 27.490127555477898,
+      "grad_norm": 0.4167649447917938,
+      "learning_rate": 0.00027028321678321677,
+      "loss": 3.1777,
+      "step": 94400
+    },
+    {
+      "epoch": 27.50468868309162,
+      "grad_norm": 0.43479686975479126,
+      "learning_rate": 0.00027010839160839157,
+      "loss": 3.1601,
+      "step": 94450
+    },
+    {
+      "epoch": 27.51924981070534,
+      "grad_norm": 0.42193397879600525,
+      "learning_rate": 0.00026993356643356637,
+      "loss": 3.1669,
+      "step": 94500
+    },
+    {
+      "epoch": 27.533810938319064,
+      "grad_norm": 0.47727319598197937,
+      "learning_rate": 0.0002697587412587412,
+      "loss": 3.1698,
+      "step": 94550
+    },
+    {
+      "epoch": 27.548372065932785,
+      "grad_norm": 0.4227813482284546,
+      "learning_rate": 0.0002695839160839161,
+      "loss": 3.1649,
+      "step": 94600
+    },
+    {
+      "epoch": 27.56293319354651,
+      "grad_norm": 0.4389137029647827,
+      "learning_rate": 0.0002694090909090909,
+      "loss": 3.1721,
+      "step": 94650
+    },
+    {
+      "epoch": 27.57749432116023,
+      "grad_norm": 0.4371396601200104,
+      "learning_rate": 0.00026923426573426573,
+      "loss": 3.1665,
+      "step": 94700
+    },
+    {
+      "epoch": 27.59205544877395,
+      "grad_norm": 0.45018795132637024,
+      "learning_rate": 0.00026905944055944053,
+      "loss": 3.1634,
+      "step": 94750
+    },
+    {
+      "epoch": 27.606616576387676,
+      "grad_norm": 0.3988571763038635,
+      "learning_rate": 0.0002688846153846154,
+      "loss": 3.1685,
+      "step": 94800
+    },
+    {
+      "epoch": 27.621177704001397,
+      "grad_norm": 0.43770983815193176,
+      "learning_rate": 0.0002687097902097902,
+      "loss": 3.1807,
+      "step": 94850
+    },
+    {
+      "epoch": 27.63573883161512,
+      "grad_norm": 0.43427303433418274,
+      "learning_rate": 0.000268534965034965,
+      "loss": 3.1776,
+      "step": 94900
+    },
+    {
+      "epoch": 27.650299959228843,
+      "grad_norm": 0.4287642538547516,
+      "learning_rate": 0.00026836013986013984,
+      "loss": 3.1615,
+      "step": 94950
+    },
+    {
+      "epoch": 27.664861086842564,
+      "grad_norm": 0.42970359325408936,
+      "learning_rate": 0.00026818531468531464,
+      "loss": 3.1677,
+      "step": 95000
+    },
+    {
+      "epoch": 27.664861086842564,
+      "eval_accuracy": 0.3748591893870121,
+      "eval_loss": 3.537830114364624,
+      "eval_runtime": 181.1281,
+      "eval_samples_per_second": 91.902,
+      "eval_steps_per_second": 5.747,
+      "step": 95000
+    },
+    {
+      "epoch": 27.679422214456288,
+      "grad_norm": 0.4214085638523102,
+      "learning_rate": 0.0002680104895104895,
+      "loss": 3.1704,
+      "step": 95050
+    },
+    {
+      "epoch": 27.69398334207001,
+      "grad_norm": 0.4056569039821625,
+      "learning_rate": 0.0002678356643356643,
+      "loss": 3.1684,
+      "step": 95100
+    },
+    {
+      "epoch": 27.708544469683734,
+      "grad_norm": 0.4250808656215668,
+      "learning_rate": 0.00026766083916083915,
+      "loss": 3.1809,
+      "step": 95150
+    },
+    {
+      "epoch": 27.723105597297454,
+      "grad_norm": 0.46413102746009827,
+      "learning_rate": 0.00026748601398601395,
+      "loss": 3.1842,
+      "step": 95200
+    },
+    {
+      "epoch": 27.737666724911175,
+      "grad_norm": 0.4296405017375946,
+      "learning_rate": 0.0002673111888111888,
+      "loss": 3.1742,
+      "step": 95250
+    },
+    {
+      "epoch": 27.7522278525249,
+      "grad_norm": 0.41784653067588806,
+      "learning_rate": 0.0002671363636363636,
+      "loss": 3.1853,
+      "step": 95300
+    },
+    {
+      "epoch": 27.76678898013862,
+      "grad_norm": 0.42170286178588867,
+      "learning_rate": 0.00026696153846153845,
+      "loss": 3.173,
+      "step": 95350
+    },
+    {
+      "epoch": 27.781350107752345,
+      "grad_norm": 0.4131033420562744,
+      "learning_rate": 0.00026678671328671325,
+      "loss": 3.1809,
+      "step": 95400
+    },
+    {
+      "epoch": 27.795911235366066,
+      "grad_norm": 0.41183069348335266,
+      "learning_rate": 0.0002666118881118881,
+      "loss": 3.168,
+      "step": 95450
+    },
+    {
+      "epoch": 27.810472362979787,
+      "grad_norm": 0.4307995140552521,
+      "learning_rate": 0.0002664370629370629,
+      "loss": 3.1832,
+      "step": 95500
+    },
+    {
+      "epoch": 27.825033490593512,
+      "grad_norm": 0.4343326687812805,
+      "learning_rate": 0.00026626223776223776,
+      "loss": 3.1806,
+      "step": 95550
+    },
+    {
+      "epoch": 27.839594618207233,
+      "grad_norm": 0.4288672208786011,
+      "learning_rate": 0.00026608741258741256,
+      "loss": 3.1826,
+      "step": 95600
+    },
+    {
+      "epoch": 27.854155745820957,
+      "grad_norm": 0.45248332619667053,
+      "learning_rate": 0.00026591258741258736,
+      "loss": 3.1664,
+      "step": 95650
+    },
+    {
+      "epoch": 27.86871687343468,
+      "grad_norm": 0.4320247769355774,
+      "learning_rate": 0.0002657377622377622,
+      "loss": 3.1905,
+      "step": 95700
+    },
+    {
+      "epoch": 27.883278001048403,
+      "grad_norm": 0.4126224219799042,
+      "learning_rate": 0.000265562937062937,
+      "loss": 3.1741,
+      "step": 95750
+    },
+    {
+      "epoch": 27.897839128662124,
+      "grad_norm": 0.43033668398857117,
+      "learning_rate": 0.00026538811188811187,
+      "loss": 3.1842,
+      "step": 95800
+    },
+    {
+      "epoch": 27.912400256275845,
+      "grad_norm": 0.43303316831588745,
+      "learning_rate": 0.00026521328671328667,
+      "loss": 3.1813,
+      "step": 95850
+    },
+    {
+      "epoch": 27.92696138388957,
+      "grad_norm": 0.4476587176322937,
+      "learning_rate": 0.0002650384615384615,
+      "loss": 3.1903,
+      "step": 95900
+    },
+    {
+      "epoch": 27.94152251150329,
+      "grad_norm": 0.44143158197402954,
+      "learning_rate": 0.0002648636363636364,
+      "loss": 3.1735,
+      "step": 95950
+    },
+    {
+      "epoch": 27.956083639117015,
+      "grad_norm": 0.4395497441291809,
+      "learning_rate": 0.0002646888111888112,
+      "loss": 3.1917,
+      "step": 96000
+    },
+    {
+      "epoch": 27.956083639117015,
+      "eval_accuracy": 0.37492549469757375,
+      "eval_loss": 3.533468008041382,
+      "eval_runtime": 181.422,
+      "eval_samples_per_second": 91.753,
+      "eval_steps_per_second": 5.738,
+      "step": 96000
+    },
+    {
+      "epoch": 27.970644766730736,
+      "grad_norm": 0.39390426874160767,
+      "learning_rate": 0.000264513986013986,
+      "loss": 3.1874,
+      "step": 96050
+    },
+    {
+      "epoch": 27.985205894344457,
+      "grad_norm": 0.4274383783340454,
+      "learning_rate": 0.00026433916083916083,
+      "loss": 3.186,
+      "step": 96100
+    },
+    {
+      "epoch": 27.99976702195818,
+      "grad_norm": 0.43289220333099365,
+      "learning_rate": 0.00026416433566433563,
+      "loss": 3.1902,
+      "step": 96150
+    },
+    {
+      "epoch": 28.014269905061447,
+      "grad_norm": 0.46341463923454285,
+      "learning_rate": 0.0002639895104895105,
+      "loss": 3.0824,
+      "step": 96200
+    },
+    {
+      "epoch": 28.02883103267517,
+      "grad_norm": 0.45257487893104553,
+      "learning_rate": 0.0002638146853146853,
+      "loss": 3.0785,
+      "step": 96250
+    },
+    {
+      "epoch": 28.043392160288892,
+      "grad_norm": 0.4652392566204071,
+      "learning_rate": 0.00026363986013986014,
+      "loss": 3.0985,
+      "step": 96300
+    },
+    {
+      "epoch": 28.057953287902617,
+      "grad_norm": 0.4572628140449524,
+      "learning_rate": 0.00026346503496503494,
+      "loss": 3.1004,
+      "step": 96350
+    },
+    {
+      "epoch": 28.072514415516338,
+      "grad_norm": 0.4591698944568634,
+      "learning_rate": 0.00026329020979020974,
+      "loss": 3.0914,
+      "step": 96400
+    },
+    {
+      "epoch": 28.08707554313006,
+      "grad_norm": 0.44280004501342773,
+      "learning_rate": 0.0002631153846153846,
+      "loss": 3.1029,
+      "step": 96450
+    },
+    {
+      "epoch": 28.101636670743783,
+      "grad_norm": 0.4677381217479706,
+      "learning_rate": 0.0002629405594405594,
+      "loss": 3.1036,
+      "step": 96500
+    },
+    {
+      "epoch": 28.116197798357504,
+      "grad_norm": 0.48905616998672485,
+      "learning_rate": 0.00026276573426573424,
+      "loss": 3.1068,
+      "step": 96550
+    },
+    {
+      "epoch": 28.13075892597123,
+      "grad_norm": 0.456222802400589,
+      "learning_rate": 0.00026259090909090904,
+      "loss": 3.108,
+      "step": 96600
+    },
+    {
+      "epoch": 28.14532005358495,
+      "grad_norm": 0.46212059259414673,
+      "learning_rate": 0.0002624160839160839,
+      "loss": 3.1184,
+      "step": 96650
+    },
+    {
+      "epoch": 28.15988118119867,
+      "grad_norm": 0.4542124271392822,
+      "learning_rate": 0.00026224125874125875,
+      "loss": 3.1212,
+      "step": 96700
+    },
+    {
+      "epoch": 28.174442308812395,
+      "grad_norm": 0.44424667954444885,
+      "learning_rate": 0.00026206643356643355,
+      "loss": 3.1177,
+      "step": 96750
+    },
+    {
+      "epoch": 28.189003436426116,
+      "grad_norm": 0.47115257382392883,
+      "learning_rate": 0.00026189160839160835,
+      "loss": 3.1258,
+      "step": 96800
+    },
+    {
+      "epoch": 28.20356456403984,
+      "grad_norm": 0.4265085458755493,
+      "learning_rate": 0.0002617167832167832,
+      "loss": 3.1237,
+      "step": 96850
+    },
+    {
+      "epoch": 28.21812569165356,
+      "grad_norm": 0.44990769028663635,
+      "learning_rate": 0.000261541958041958,
+      "loss": 3.1388,
+      "step": 96900
+    },
+    {
+      "epoch": 28.232686819267283,
+      "grad_norm": 0.44933226704597473,
+      "learning_rate": 0.00026136713286713286,
+      "loss": 3.1249,
+      "step": 96950
+    },
+    {
+      "epoch": 28.247247946881007,
+      "grad_norm": 0.4495621919631958,
+      "learning_rate": 0.00026119230769230766,
+      "loss": 3.1284,
+      "step": 97000
+    },
+    {
+      "epoch": 28.247247946881007,
+      "eval_accuracy": 0.37415416643056176,
+      "eval_loss": 3.550610303878784,
+      "eval_runtime": 181.4389,
+      "eval_samples_per_second": 91.744,
+      "eval_steps_per_second": 5.737,
+      "step": 97000
+    },
+    {
+      "epoch": 28.261809074494728,
+      "grad_norm": 0.4604361653327942,
+      "learning_rate": 0.0002610174825174825,
+      "loss": 3.1232,
+      "step": 97050
+    },
+    {
+      "epoch": 28.276370202108453,
+      "grad_norm": 0.4367489218711853,
+      "learning_rate": 0.0002608426573426573,
+      "loss": 3.1272,
+      "step": 97100
+    },
+    {
+      "epoch": 28.290931329722174,
+      "grad_norm": 0.4408527612686157,
+      "learning_rate": 0.0002606678321678321,
+      "loss": 3.1239,
+      "step": 97150
+    },
+    {
+      "epoch": 28.305492457335895,
+      "grad_norm": 0.4509691596031189,
+      "learning_rate": 0.00026049300699300696,
+      "loss": 3.1393,
+      "step": 97200
+    },
+    {
+      "epoch": 28.32005358494962,
+      "grad_norm": 0.42577970027923584,
+      "learning_rate": 0.00026031818181818176,
+      "loss": 3.1271,
+      "step": 97250
+    },
+    {
+      "epoch": 28.33461471256334,
+      "grad_norm": 0.4179809093475342,
+      "learning_rate": 0.0002601433566433566,
+      "loss": 3.1299,
+      "step": 97300
+    },
+    {
+      "epoch": 28.349175840177065,
+      "grad_norm": 0.42162272334098816,
+      "learning_rate": 0.00025996853146853147,
+      "loss": 3.1311,
+      "step": 97350
+    },
+    {
+      "epoch": 28.363736967790786,
+      "grad_norm": 0.4421674311161041,
+      "learning_rate": 0.00025979370629370627,
+      "loss": 3.1443,
+      "step": 97400
+    },
+    {
+      "epoch": 28.378298095404507,
+      "grad_norm": 0.45821619033813477,
+      "learning_rate": 0.0002596188811188811,
+      "loss": 3.1496,
+      "step": 97450
+    },
+    {
+      "epoch": 28.39285922301823,
+      "grad_norm": 0.4294704794883728,
+      "learning_rate": 0.0002594440559440559,
+      "loss": 3.1382,
+      "step": 97500
+    },
+    {
+      "epoch": 28.407420350631952,
+      "grad_norm": 0.4437113106250763,
+      "learning_rate": 0.0002592692307692307,
+      "loss": 3.1473,
+      "step": 97550
+    },
+    {
+      "epoch": 28.421981478245677,
+      "grad_norm": 0.449618399143219,
+      "learning_rate": 0.0002590944055944056,
+      "loss": 3.1572,
+      "step": 97600
+    },
+    {
+      "epoch": 28.436542605859398,
+      "grad_norm": 0.43033522367477417,
+      "learning_rate": 0.0002589195804195804,
+      "loss": 3.1457,
+      "step": 97650
+    },
+    {
+      "epoch": 28.45110373347312,
+      "grad_norm": 0.4679327607154846,
+      "learning_rate": 0.00025874475524475523,
+      "loss": 3.1525,
+      "step": 97700
+    },
+    {
+      "epoch": 28.465664861086843,
+      "grad_norm": 0.4457078278064728,
+      "learning_rate": 0.00025856993006993003,
+      "loss": 3.1485,
+      "step": 97750
+    },
+    {
+      "epoch": 28.480225988700564,
+      "grad_norm": 0.43366381525993347,
+      "learning_rate": 0.0002583951048951049,
+      "loss": 3.1554,
+      "step": 97800
+    },
+    {
+      "epoch": 28.49478711631429,
+      "grad_norm": 0.44160208106040955,
+      "learning_rate": 0.0002582202797202797,
+      "loss": 3.1466,
+      "step": 97850
+    },
+    {
+      "epoch": 28.50934824392801,
+      "grad_norm": 0.4116990268230438,
+      "learning_rate": 0.0002580454545454545,
+      "loss": 3.1515,
+      "step": 97900
+    },
+    {
+      "epoch": 28.523909371541734,
+      "grad_norm": 0.4543076455593109,
+      "learning_rate": 0.00025787062937062934,
+      "loss": 3.1502,
+      "step": 97950
+    },
+    {
+      "epoch": 28.538470499155455,
+      "grad_norm": 0.4376068711280823,
+      "learning_rate": 0.0002576958041958042,
+      "loss": 3.1637,
+      "step": 98000
+    },
+    {
+      "epoch": 28.538470499155455,
+      "eval_accuracy": 0.374864127016522,
+      "eval_loss": 3.543623924255371,
+      "eval_runtime": 181.1432,
+      "eval_samples_per_second": 91.894,
+      "eval_steps_per_second": 5.747,
+      "step": 98000
+    },
+    {
+      "epoch": 28.553031626769176,
+      "grad_norm": 0.41571149230003357,
+      "learning_rate": 0.000257520979020979,
+      "loss": 3.1527,
+      "step": 98050
+    },
+    {
+      "epoch": 28.5675927543829,
+      "grad_norm": 0.42478272318840027,
+      "learning_rate": 0.00025734615384615385,
+      "loss": 3.1513,
+      "step": 98100
+    },
+    {
+      "epoch": 28.58215388199662,
+      "grad_norm": 0.446841299533844,
+      "learning_rate": 0.00025717132867132865,
+      "loss": 3.1573,
+      "step": 98150
+    },
+    {
+      "epoch": 28.596715009610342,
+      "grad_norm": 0.4337193965911865,
+      "learning_rate": 0.0002569965034965035,
+      "loss": 3.1565,
+      "step": 98200
+    },
+    {
+      "epoch": 28.611276137224067,
+      "grad_norm": 0.41480013728141785,
+      "learning_rate": 0.0002568216783216783,
+      "loss": 3.1569,
+      "step": 98250
+    },
+    {
+      "epoch": 28.625837264837788,
+      "grad_norm": 0.45146986842155457,
+      "learning_rate": 0.0002566468531468531,
+      "loss": 3.1727,
+      "step": 98300
+    },
+    {
+      "epoch": 28.640398392451512,
+      "grad_norm": 0.44779402017593384,
+      "learning_rate": 0.00025647202797202795,
+      "loss": 3.1558,
+      "step": 98350
+    },
+    {
+      "epoch": 28.654959520065233,
+      "grad_norm": 0.4365552067756653,
+      "learning_rate": 0.00025629720279720275,
+      "loss": 3.1651,
+      "step": 98400
+    },
+    {
+      "epoch": 28.669520647678958,
+      "grad_norm": 0.4349055290222168,
+      "learning_rate": 0.0002561223776223776,
+      "loss": 3.1562,
+      "step": 98450
+    },
+    {
+      "epoch": 28.68408177529268,
+      "grad_norm": 0.4297862946987152,
+      "learning_rate": 0.0002559475524475524,
+      "loss": 3.1711,
+      "step": 98500
+    },
+    {
+      "epoch": 28.6986429029064,
+      "grad_norm": 0.44524985551834106,
+      "learning_rate": 0.00025577272727272726,
+      "loss": 3.1499,
+      "step": 98550
+    },
+    {
+      "epoch": 28.713204030520124,
+      "grad_norm": 0.44011494517326355,
+      "learning_rate": 0.00025559790209790206,
+      "loss": 3.1627,
+      "step": 98600
+    },
+    {
+      "epoch": 28.727765158133845,
+      "grad_norm": 0.44971513748168945,
+      "learning_rate": 0.00025542307692307686,
+      "loss": 3.1663,
+      "step": 98650
+    },
+    {
+      "epoch": 28.74232628574757,
+      "grad_norm": 0.4642869532108307,
+      "learning_rate": 0.00025524825174825177,
+      "loss": 3.1668,
+      "step": 98700
+    },
+    {
+      "epoch": 28.75688741336129,
+      "grad_norm": 0.43200162053108215,
+      "learning_rate": 0.00025507342657342657,
+      "loss": 3.177,
+      "step": 98750
+    },
+    {
+      "epoch": 28.771448540975012,
+      "grad_norm": 0.43370288610458374,
+      "learning_rate": 0.00025489860139860137,
+      "loss": 3.1681,
+      "step": 98800
+    },
+    {
+      "epoch": 28.786009668588736,
+      "grad_norm": 0.46173909306526184,
+      "learning_rate": 0.0002547237762237762,
+      "loss": 3.1736,
+      "step": 98850
+    },
+    {
+      "epoch": 28.800570796202457,
+      "grad_norm": 0.4236147403717041,
+      "learning_rate": 0.000254548951048951,
+      "loss": 3.1683,
+      "step": 98900
+    },
+    {
+      "epoch": 28.815131923816182,
+      "grad_norm": 0.439842164516449,
+      "learning_rate": 0.0002543741258741259,
+      "loss": 3.1618,
+      "step": 98950
+    },
+    {
+      "epoch": 28.829693051429903,
+      "grad_norm": 0.41996750235557556,
+      "learning_rate": 0.0002541993006993007,
+      "loss": 3.1684,
+      "step": 99000
+    },
+    {
+      "epoch": 28.829693051429903,
+      "eval_accuracy": 0.37464710644330085,
+      "eval_loss": 3.5397422313690186,
+      "eval_runtime": 180.731,
+      "eval_samples_per_second": 92.104,
+      "eval_steps_per_second": 5.76,
+      "step": 99000
+    },
+    {
+      "epoch": 28.844254179043624,
+      "grad_norm": 0.4436616003513336,
+      "learning_rate": 0.0002540244755244755,
+      "loss": 3.1716,
+      "step": 99050
+    },
+    {
+      "epoch": 28.85881530665735,
+      "grad_norm": 0.43761077523231506,
+      "learning_rate": 0.00025384965034965033,
+      "loss": 3.1711,
+      "step": 99100
+    },
+    {
+      "epoch": 28.87337643427107,
+      "grad_norm": 0.4494776725769043,
+      "learning_rate": 0.00025367482517482513,
+      "loss": 3.1734,
+      "step": 99150
+    },
+    {
+      "epoch": 28.887937561884794,
+      "grad_norm": 0.4295901954174042,
+      "learning_rate": 0.0002535,
+      "loss": 3.173,
+      "step": 99200
+    },
+    {
+      "epoch": 28.902498689498515,
+      "grad_norm": 0.479999303817749,
+      "learning_rate": 0.0002533251748251748,
+      "loss": 3.1758,
+      "step": 99250
+    },
+    {
+      "epoch": 28.917059817112236,
+      "grad_norm": 0.41707783937454224,
+      "learning_rate": 0.00025315034965034964,
+      "loss": 3.1793,
+      "step": 99300
+    },
+    {
+      "epoch": 28.93162094472596,
+      "grad_norm": 0.41884753108024597,
+      "learning_rate": 0.00025297552447552444,
+      "loss": 3.1899,
+      "step": 99350
+    },
+    {
+      "epoch": 28.94618207233968,
+      "grad_norm": 0.44001612067222595,
+      "learning_rate": 0.0002528006993006993,
+      "loss": 3.1776,
+      "step": 99400
+    },
+    {
+      "epoch": 28.960743199953406,
+      "grad_norm": 0.44126561284065247,
+      "learning_rate": 0.00025262587412587414,
+      "loss": 3.1658,
+      "step": 99450
+    },
+    {
+      "epoch": 28.975304327567127,
+      "grad_norm": 0.46320417523384094,
+      "learning_rate": 0.00025245104895104894,
+      "loss": 3.1692,
+      "step": 99500
+    },
+    {
+      "epoch": 28.989865455180848,
+      "grad_norm": 0.4358890950679779,
+      "learning_rate": 0.00025227622377622374,
+      "loss": 3.1706,
+      "step": 99550
+    },
+    {
+      "epoch": 29.004368338284117,
+      "grad_norm": 0.4340914189815521,
+      "learning_rate": 0.0002521013986013986,
+      "loss": 3.1496,
+      "step": 99600
+    },
+    {
+      "epoch": 29.018929465897838,
+      "grad_norm": 0.4288674592971802,
+      "learning_rate": 0.0002519265734265734,
+      "loss": 3.086,
+      "step": 99650
+    },
+    {
+      "epoch": 29.033490593511562,
+      "grad_norm": 0.44773054122924805,
+      "learning_rate": 0.00025175174825174825,
+      "loss": 3.0863,
+      "step": 99700
+    },
+    {
+      "epoch": 29.048051721125283,
+      "grad_norm": 0.4277186095714569,
+      "learning_rate": 0.00025157692307692305,
+      "loss": 3.0906,
+      "step": 99750
+    },
+    {
+      "epoch": 29.062612848739008,
+      "grad_norm": 0.44387462735176086,
+      "learning_rate": 0.0002514020979020979,
+      "loss": 3.0821,
+      "step": 99800
+    },
+    {
+      "epoch": 29.07717397635273,
+      "grad_norm": 0.45624199509620667,
+      "learning_rate": 0.0002512272727272727,
+      "loss": 3.0912,
+      "step": 99850
+    },
+    {
+      "epoch": 29.09173510396645,
+      "grad_norm": 0.4246789216995239,
+      "learning_rate": 0.0002510524475524475,
+      "loss": 3.107,
+      "step": 99900
+    },
+    {
+      "epoch": 29.106296231580174,
+      "grad_norm": 0.43305352330207825,
+      "learning_rate": 0.00025087762237762236,
+      "loss": 3.1013,
+      "step": 99950
+    },
+    {
+      "epoch": 29.120857359193895,
+      "grad_norm": 0.4623396694660187,
+      "learning_rate": 0.00025070279720279716,
+      "loss": 3.1085,
+      "step": 100000
+    },
+    {
+      "epoch": 29.120857359193895,
+      "eval_accuracy": 0.3738908261900334,
+      "eval_loss": 3.561875581741333,
+      "eval_runtime": 180.7786,
+      "eval_samples_per_second": 92.079,
+      "eval_steps_per_second": 5.758,
+      "step": 100000
+    }
+  ],
+  "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": 11
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.090205609984e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}