diff --git "a/last_to_push_frequency_3591/checkpoint-70000/trainer_state.json" "b/last_to_push_frequency_3591/checkpoint-70000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/last_to_push_frequency_3591/checkpoint-70000/trainer_state.json"
@@ -0,0 +1,10473 @@
+{
+  "best_global_step": 65000,
+  "best_metric": 3.5290017127990723,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/last_to_push_frequency_3591/checkpoint-30000",
+  "epoch": 20.384413769002272,
+  "eval_steps": 1000,
+  "global_step": 70000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014561127613722406,
+      "grad_norm": 0.8466977477073669,
+      "learning_rate": 0.000294,
+      "loss": 8.4098,
+      "step": 50
+    },
+    {
+      "epoch": 0.029122255227444813,
+      "grad_norm": 0.6842089295387268,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7014,
+      "step": 100
+    },
+    {
+      "epoch": 0.04368338284116722,
+      "grad_norm": 0.7142037749290466,
+      "learning_rate": 0.0005998286713286713,
+      "loss": 6.34,
+      "step": 150
+    },
+    {
+      "epoch": 0.058244510454889625,
+      "grad_norm": 0.5938196778297424,
+      "learning_rate": 0.0005996538461538461,
+      "loss": 6.139,
+      "step": 200
+    },
+    {
+      "epoch": 0.07280563806861204,
+      "grad_norm": 0.3856406807899475,
+      "learning_rate": 0.0005994790209790209,
+      "loss": 5.9921,
+      "step": 250
+    },
+    {
+      "epoch": 0.08736676568233444,
+      "grad_norm": 0.39986154437065125,
+      "learning_rate": 0.0005993041958041958,
+      "loss": 5.8431,
+      "step": 300
+    },
+    {
+      "epoch": 0.10192789329605685,
+      "grad_norm": 0.5341420769691467,
+      "learning_rate": 0.0005991293706293705,
+      "loss": 5.7043,
+      "step": 350
+    },
+    {
+      "epoch": 0.11648902090977925,
+      "grad_norm": 0.4427669644355774,
+      "learning_rate": 0.0005989545454545454,
+      "loss": 5.6019,
+      "step": 400
+    },
+    {
+      "epoch": 0.13105014852350166,
+      "grad_norm": 0.45295220613479614,
+      "learning_rate": 0.0005987797202797202,
+      "loss": 5.4996,
+      "step": 450
+    },
+    {
+      "epoch": 0.14561127613722408,
+      "grad_norm": 0.5014562606811523,
+      "learning_rate": 0.000598604895104895,
+      "loss": 5.3991,
+      "step": 500
+    },
+    {
+      "epoch": 0.16017240375094646,
+      "grad_norm": 0.4532122015953064,
+      "learning_rate": 0.0005984300699300698,
+      "loss": 5.3103,
+      "step": 550
+    },
+    {
+      "epoch": 0.17473353136466888,
+      "grad_norm": 0.45960476994514465,
+      "learning_rate": 0.0005982552447552447,
+      "loss": 5.2526,
+      "step": 600
+    },
+    {
+      "epoch": 0.1892946589783913,
+      "grad_norm": 0.4369155764579773,
+      "learning_rate": 0.0005980804195804195,
+      "loss": 5.1734,
+      "step": 650
+    },
+    {
+      "epoch": 0.2038557865921137,
+      "grad_norm": 0.48178285360336304,
+      "learning_rate": 0.0005979055944055943,
+      "loss": 5.1075,
+      "step": 700
+    },
+    {
+      "epoch": 0.2184169142058361,
+      "grad_norm": 0.44058653712272644,
+      "learning_rate": 0.0005977307692307691,
+      "loss": 5.0671,
+      "step": 750
+    },
+    {
+      "epoch": 0.2329780418195585,
+      "grad_norm": 0.45520836114883423,
+      "learning_rate": 0.000597555944055944,
+      "loss": 5.0176,
+      "step": 800
+    },
+    {
+      "epoch": 0.24753916943328091,
+      "grad_norm": 0.4496179521083832,
+      "learning_rate": 0.0005973811188811188,
+      "loss": 4.972,
+      "step": 850
+    },
+    {
+      "epoch": 0.2621002970470033,
+      "grad_norm": 0.5189023613929749,
+      "learning_rate": 0.0005972062937062936,
+      "loss": 4.9247,
+      "step": 900
+    },
+    {
+      "epoch": 0.27666142466072574,
+      "grad_norm": 0.542473554611206,
+      "learning_rate": 0.0005970314685314685,
+      "loss": 4.8671,
+      "step": 950
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "grad_norm": 0.45608243346214294,
+      "learning_rate": 0.0005968566433566433,
+      "loss": 4.8193,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29122255227444815,
+      "eval_accuracy": 0.2555685292424054,
+      "eval_loss": 4.740924835205078,
+      "eval_runtime": 182.37,
+      "eval_samples_per_second": 91.276,
+      "eval_steps_per_second": 5.708,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30578367988817057,
+      "grad_norm": 0.4338395893573761,
+      "learning_rate": 0.0005966818181818181,
+      "loss": 4.7728,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3203448075018929,
+      "grad_norm": 0.5276835560798645,
+      "learning_rate": 0.0005965069930069929,
+      "loss": 4.7444,
+      "step": 1100
+    },
+    {
+      "epoch": 0.33490593511561534,
+      "grad_norm": 0.44655585289001465,
+      "learning_rate": 0.0005963321678321677,
+      "loss": 4.6967,
+      "step": 1150
+    },
+    {
+      "epoch": 0.34946706272933775,
+      "grad_norm": 0.38414475321769714,
+      "learning_rate": 0.0005961573426573425,
+      "loss": 4.6614,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36402819034306017,
+      "grad_norm": 0.4639468193054199,
+      "learning_rate": 0.0005959825174825174,
+      "loss": 4.6234,
+      "step": 1250
+    },
+    {
+      "epoch": 0.3785893179567826,
+      "grad_norm": 0.4747146964073181,
+      "learning_rate": 0.0005958076923076922,
+      "loss": 4.6,
+      "step": 1300
+    },
+    {
+      "epoch": 0.393150445570505,
+      "grad_norm": 0.5148531198501587,
+      "learning_rate": 0.000595632867132867,
+      "loss": 4.5744,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4077115731842274,
+      "grad_norm": 0.4540764391422272,
+      "learning_rate": 0.0005954580419580418,
+      "loss": 4.5458,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4222727007979498,
+      "grad_norm": 0.4469960033893585,
+      "learning_rate": 0.0005952832167832168,
+      "loss": 4.5213,
+      "step": 1450
+    },
+    {
+      "epoch": 0.4368338284116722,
+      "grad_norm": 0.4068995416164398,
+      "learning_rate": 0.0005951083916083916,
+      "loss": 4.4989,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4513949560253946,
+      "grad_norm": 0.36514100432395935,
+      "learning_rate": 0.0005949335664335664,
+      "loss": 4.4959,
+      "step": 1550
+    },
+    {
+      "epoch": 0.465956083639117,
+      "grad_norm": 0.41950780153274536,
+      "learning_rate": 0.0005947587412587413,
+      "loss": 4.4468,
+      "step": 1600
+    },
+    {
+      "epoch": 0.4805172112528394,
+      "grad_norm": 0.42646339535713196,
+      "learning_rate": 0.0005945839160839161,
+      "loss": 4.4366,
+      "step": 1650
+    },
+    {
+      "epoch": 0.49507833886656183,
+      "grad_norm": 0.3759300112724304,
+      "learning_rate": 0.0005944090909090909,
+      "loss": 4.425,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5096394664802842,
+      "grad_norm": 0.40312129259109497,
+      "learning_rate": 0.0005942342657342657,
+      "loss": 4.3986,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5242005940940067,
+      "grad_norm": 0.3916018009185791,
+      "learning_rate": 0.0005940594405594406,
+      "loss": 4.3877,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5387617217077291,
+      "grad_norm": 0.36997199058532715,
+      "learning_rate": 0.0005938846153846153,
+      "loss": 4.3737,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5533228493214515,
+      "grad_norm": 0.4766369163990021,
+      "learning_rate": 0.0005937097902097902,
+      "loss": 4.3561,
+      "step": 1900
+    },
+    {
+      "epoch": 0.5678839769351739,
+      "grad_norm": 0.4964778423309326,
+      "learning_rate": 0.000593534965034965,
+      "loss": 4.3618,
+      "step": 1950
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "grad_norm": 0.43863603472709656,
+      "learning_rate": 0.0005933601398601398,
+      "loss": 4.3232,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5824451045488963,
+      "eval_accuracy": 0.29954776016193546,
+      "eval_loss": 4.281495094299316,
+      "eval_runtime": 182.5417,
+      "eval_samples_per_second": 91.19,
+      "eval_steps_per_second": 5.703,
+      "step": 2000
+    },
+    {
+      "epoch": 0.5970062321626187,
+      "grad_norm": 0.3895237147808075,
+      "learning_rate": 0.0005931853146853146,
+      "loss": 4.3153,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6115673597763411,
+      "grad_norm": 0.35456663370132446,
+      "learning_rate": 0.0005930104895104895,
+      "loss": 4.3017,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6261284873900634,
+      "grad_norm": 0.36657285690307617,
+      "learning_rate": 0.0005928356643356643,
+      "loss": 4.3021,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6406896150037859,
+      "grad_norm": 0.3920918107032776,
+      "learning_rate": 0.0005926608391608391,
+      "loss": 4.2813,
+      "step": 2200
+    },
+    {
+      "epoch": 0.6552507426175083,
+      "grad_norm": 0.3744465708732605,
+      "learning_rate": 0.000592486013986014,
+      "loss": 4.2666,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6698118702312307,
+      "grad_norm": 0.38081395626068115,
+      "learning_rate": 0.0005923111888111888,
+      "loss": 4.2704,
+      "step": 2300
+    },
+    {
+      "epoch": 0.6843729978449531,
+      "grad_norm": 0.36270445585250854,
+      "learning_rate": 0.0005921363636363636,
+      "loss": 4.24,
+      "step": 2350
+    },
+    {
+      "epoch": 0.6989341254586755,
+      "grad_norm": 0.40953528881073,
+      "learning_rate": 0.0005919615384615384,
+      "loss": 4.2401,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7134952530723979,
+      "grad_norm": 0.36897608637809753,
+      "learning_rate": 0.0005917867132867133,
+      "loss": 4.2372,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7280563806861203,
+      "grad_norm": 0.35832679271698,
+      "learning_rate": 0.0005916118881118881,
+      "loss": 4.2331,
+      "step": 2500
+    },
+    {
+      "epoch": 0.7426175082998427,
+      "grad_norm": 0.38073498010635376,
+      "learning_rate": 0.0005914370629370629,
+      "loss": 4.2068,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7571786359135652,
+      "grad_norm": 0.4003191888332367,
+      "learning_rate": 0.0005912622377622377,
+      "loss": 4.2046,
+      "step": 2600
+    },
+    {
+      "epoch": 0.7717397635272876,
+      "grad_norm": 0.3635654151439667,
+      "learning_rate": 0.0005910874125874125,
+      "loss": 4.1995,
+      "step": 2650
+    },
+    {
+      "epoch": 0.78630089114101,
+      "grad_norm": 0.35873937606811523,
+      "learning_rate": 0.0005909125874125873,
+      "loss": 4.2033,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8008620187547324,
+      "grad_norm": 0.3539746105670929,
+      "learning_rate": 0.0005907377622377622,
+      "loss": 4.1872,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8154231463684548,
+      "grad_norm": 0.37548962235450745,
+      "learning_rate": 0.000590562937062937,
+      "loss": 4.168,
+      "step": 2800
+    },
+    {
+      "epoch": 0.8299842739821772,
+      "grad_norm": 0.3538251221179962,
+      "learning_rate": 0.0005903881118881118,
+      "loss": 4.179,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8445454015958996,
+      "grad_norm": 0.3582804501056671,
+      "learning_rate": 0.0005902132867132867,
+      "loss": 4.1529,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8591065292096219,
+      "grad_norm": 0.3666636049747467,
+      "learning_rate": 0.0005900384615384615,
+      "loss": 4.1502,
+      "step": 2950
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "grad_norm": 0.37483951449394226,
+      "learning_rate": 0.0005898636363636363,
+      "loss": 4.1292,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8736676568233444,
+      "eval_accuracy": 0.31617228847136397,
+      "eval_loss": 4.089555263519287,
+      "eval_runtime": 182.8843,
+      "eval_samples_per_second": 91.019,
+      "eval_steps_per_second": 5.692,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8882287844370668,
+      "grad_norm": 0.3864665925502777,
+      "learning_rate": 0.0005896888111888111,
+      "loss": 4.1193,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9027899120507892,
+      "grad_norm": 0.347319632768631,
+      "learning_rate": 0.000589513986013986,
+      "loss": 4.1179,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9173510396645116,
+      "grad_norm": 0.3676200807094574,
+      "learning_rate": 0.0005893391608391608,
+      "loss": 4.1154,
+      "step": 3150
+    },
+    {
+      "epoch": 0.931912167278234,
+      "grad_norm": 0.3425697088241577,
+      "learning_rate": 0.0005891643356643356,
+      "loss": 4.1291,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9464732948919564,
+      "grad_norm": 0.3959852457046509,
+      "learning_rate": 0.0005889895104895104,
+      "loss": 4.1053,
+      "step": 3250
+    },
+    {
+      "epoch": 0.9610344225056788,
+      "grad_norm": 0.3836718499660492,
+      "learning_rate": 0.0005888146853146853,
+      "loss": 4.1068,
+      "step": 3300
+    },
+    {
+      "epoch": 0.9755955501194012,
+      "grad_norm": 0.4188532531261444,
+      "learning_rate": 0.00058863986013986,
+      "loss": 4.0989,
+      "step": 3350
+    },
+    {
+      "epoch": 0.9901566777331237,
+      "grad_norm": 0.33719179034233093,
+      "learning_rate": 0.0005884650349650349,
+      "loss": 4.0963,
+      "step": 3400
+    },
+    {
+      "epoch": 1.004659560836391,
+      "grad_norm": 0.36592167615890503,
+      "learning_rate": 0.0005882902097902097,
+      "loss": 4.0577,
+      "step": 3450
+    },
+    {
+      "epoch": 1.0192206884501136,
+      "grad_norm": 0.3381110727787018,
+      "learning_rate": 0.0005881153846153845,
+      "loss": 4.0016,
+      "step": 3500
+    },
+    {
+      "epoch": 1.033781816063836,
+      "grad_norm": 0.35888952016830444,
+      "learning_rate": 0.0005879405594405594,
+      "loss": 4.0032,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0483429436775584,
+      "grad_norm": 0.35463494062423706,
+      "learning_rate": 0.0005877657342657342,
+      "loss": 4.0018,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0629040712912807,
+      "grad_norm": 0.3339248299598694,
+      "learning_rate": 0.000587590909090909,
+      "loss": 4.0116,
+      "step": 3650
+    },
+    {
+      "epoch": 1.0774651989050033,
+      "grad_norm": 0.37050846219062805,
+      "learning_rate": 0.0005874160839160838,
+      "loss": 4.0031,
+      "step": 3700
+    },
+    {
+      "epoch": 1.0920263265187256,
+      "grad_norm": 0.37132593989372253,
+      "learning_rate": 0.0005872412587412587,
+      "loss": 3.996,
+      "step": 3750
+    },
+    {
+      "epoch": 1.106587454132448,
+      "grad_norm": 0.3583703339099884,
+      "learning_rate": 0.0005870664335664335,
+      "loss": 3.9926,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1211485817461704,
+      "grad_norm": 0.32278576493263245,
+      "learning_rate": 0.0005868916083916083,
+      "loss": 3.9928,
+      "step": 3850
+    },
+    {
+      "epoch": 1.135709709359893,
+      "grad_norm": 0.46608036756515503,
+      "learning_rate": 0.0005867167832167831,
+      "loss": 3.9823,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1502708369736152,
+      "grad_norm": 0.3258165717124939,
+      "learning_rate": 0.000586541958041958,
+      "loss": 3.99,
+      "step": 3950
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "grad_norm": 0.36865976452827454,
+      "learning_rate": 0.0005863671328671328,
+      "loss": 3.9822,
+      "step": 4000
+    },
+    {
+      "epoch": 1.1648319645873377,
+      "eval_accuracy": 0.3258375806744003,
+      "eval_loss": 3.9846527576446533,
+      "eval_runtime": 182.7722,
+      "eval_samples_per_second": 91.075,
+      "eval_steps_per_second": 5.696,
+      "step": 4000
+    },
+    {
+      "epoch": 1.17939309220106,
+      "grad_norm": 0.3421526551246643,
+      "learning_rate": 0.0005861923076923076,
+      "loss": 3.995,
+      "step": 4050
+    },
+    {
+      "epoch": 1.1939542198147826,
+      "grad_norm": 0.34738972783088684,
+      "learning_rate": 0.0005860174825174824,
+      "loss": 3.9821,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2085153474285049,
+      "grad_norm": 0.33670172095298767,
+      "learning_rate": 0.0005858426573426573,
+      "loss": 3.9627,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2230764750422272,
+      "grad_norm": 0.3570351004600525,
+      "learning_rate": 0.000585667832167832,
+      "loss": 3.9683,
+      "step": 4200
+    },
+    {
+      "epoch": 1.2376376026559497,
+      "grad_norm": 0.34402894973754883,
+      "learning_rate": 0.000585493006993007,
+      "loss": 3.9625,
+      "step": 4250
+    },
+    {
+      "epoch": 1.2521987302696722,
+      "grad_norm": 0.3514033555984497,
+      "learning_rate": 0.0005853181818181817,
+      "loss": 3.9698,
+      "step": 4300
+    },
+    {
+      "epoch": 1.2667598578833945,
+      "grad_norm": 0.34714460372924805,
+      "learning_rate": 0.0005851433566433565,
+      "loss": 3.9567,
+      "step": 4350
+    },
+    {
+      "epoch": 1.2813209854971168,
+      "grad_norm": 0.34446805715560913,
+      "learning_rate": 0.0005849685314685315,
+      "loss": 3.9564,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2958821131108393,
+      "grad_norm": 0.3477458655834198,
+      "learning_rate": 0.0005847937062937063,
+      "loss": 3.9627,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3104432407245616,
+      "grad_norm": 0.32496753334999084,
+      "learning_rate": 0.0005846188811188811,
+      "loss": 3.9557,
+      "step": 4500
+    },
+    {
+      "epoch": 1.3250043683382842,
+      "grad_norm": 0.37617602944374084,
+      "learning_rate": 0.0005844440559440559,
+      "loss": 3.9518,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3395654959520065,
+      "grad_norm": 0.35174936056137085,
+      "learning_rate": 0.0005842692307692308,
+      "loss": 3.9558,
+      "step": 4600
+    },
+    {
+      "epoch": 1.354126623565729,
+      "grad_norm": 0.36389508843421936,
+      "learning_rate": 0.0005840944055944056,
+      "loss": 3.9488,
+      "step": 4650
+    },
+    {
+      "epoch": 1.3686877511794513,
+      "grad_norm": 0.3404442071914673,
+      "learning_rate": 0.0005839195804195804,
+      "loss": 3.9436,
+      "step": 4700
+    },
+    {
+      "epoch": 1.3832488787931738,
+      "grad_norm": 0.3432350158691406,
+      "learning_rate": 0.0005837447552447552,
+      "loss": 3.934,
+      "step": 4750
+    },
+    {
+      "epoch": 1.3978100064068961,
+      "grad_norm": 0.3505391776561737,
+      "learning_rate": 0.0005835699300699301,
+      "loss": 3.9345,
+      "step": 4800
+    },
+    {
+      "epoch": 1.4123711340206184,
+      "grad_norm": 0.3216480016708374,
+      "learning_rate": 0.0005833951048951048,
+      "loss": 3.9266,
+      "step": 4850
+    },
+    {
+      "epoch": 1.426932261634341,
+      "grad_norm": 0.3275972902774811,
+      "learning_rate": 0.0005832202797202797,
+      "loss": 3.9393,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4414933892480635,
+      "grad_norm": 0.35102906823158264,
+      "learning_rate": 0.0005830454545454546,
+      "loss": 3.9359,
+      "step": 4950
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "grad_norm": 0.3343009650707245,
+      "learning_rate": 0.0005828706293706293,
+      "loss": 3.9236,
+      "step": 5000
+    },
+    {
+      "epoch": 1.4560545168617858,
+      "eval_accuracy": 0.33192285635753893,
+      "eval_loss": 3.9100613594055176,
+      "eval_runtime": 184.3964,
+      "eval_samples_per_second": 90.273,
+      "eval_steps_per_second": 5.645,
+      "step": 5000
+    },
+    {
+      "epoch": 1.470615644475508,
+      "grad_norm": 0.32075586915016174,
+      "learning_rate": 0.0005826958041958042,
+      "loss": 3.923,
+      "step": 5050
+    },
+    {
+      "epoch": 1.4851767720892306,
+      "grad_norm": 0.3327905237674713,
+      "learning_rate": 0.000582520979020979,
+      "loss": 3.9322,
+      "step": 5100
+    },
+    {
+      "epoch": 1.4997378997029531,
+      "grad_norm": 0.34814631938934326,
+      "learning_rate": 0.0005823461538461538,
+      "loss": 3.9174,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5142990273166754,
+      "grad_norm": 0.3348972797393799,
+      "learning_rate": 0.0005821713286713286,
+      "loss": 3.9043,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5288601549303977,
+      "grad_norm": 0.3347136676311493,
+      "learning_rate": 0.0005819965034965035,
+      "loss": 3.9046,
+      "step": 5250
+    },
+    {
+      "epoch": 1.5434212825441203,
+      "grad_norm": 0.336887001991272,
+      "learning_rate": 0.0005818216783216783,
+      "loss": 3.915,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5579824101578428,
+      "grad_norm": 0.3282018005847931,
+      "learning_rate": 0.0005816468531468531,
+      "loss": 3.9147,
+      "step": 5350
+    },
+    {
+      "epoch": 1.572543537771565,
+      "grad_norm": 0.34454506635665894,
+      "learning_rate": 0.0005814720279720279,
+      "loss": 3.9073,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5871046653852874,
+      "grad_norm": 0.3495652377605438,
+      "learning_rate": 0.0005812972027972028,
+      "loss": 3.8947,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6016657929990097,
+      "grad_norm": 0.33489686250686646,
+      "learning_rate": 0.0005811223776223776,
+      "loss": 3.9016,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6162269206127322,
+      "grad_norm": 0.3139170706272125,
+      "learning_rate": 0.0005809475524475524,
+      "loss": 3.8813,
+      "step": 5550
+    },
+    {
+      "epoch": 1.6307880482264547,
+      "grad_norm": 0.3473421335220337,
+      "learning_rate": 0.0005807727272727272,
+      "loss": 3.8912,
+      "step": 5600
+    },
+    {
+      "epoch": 1.645349175840177,
+      "grad_norm": 0.3421488106250763,
+      "learning_rate": 0.0005805979020979021,
+      "loss": 3.8781,
+      "step": 5650
+    },
+    {
+      "epoch": 1.6599103034538993,
+      "grad_norm": 0.3326078951358795,
+      "learning_rate": 0.0005804230769230769,
+      "loss": 3.8902,
+      "step": 5700
+    },
+    {
+      "epoch": 1.6744714310676219,
+      "grad_norm": 0.34903430938720703,
+      "learning_rate": 0.0005802482517482517,
+      "loss": 3.8875,
+      "step": 5750
+    },
+    {
+      "epoch": 1.6890325586813444,
+      "grad_norm": 0.337380051612854,
+      "learning_rate": 0.0005800734265734265,
+      "loss": 3.8923,
+      "step": 5800
+    },
+    {
+      "epoch": 1.7035936862950667,
+      "grad_norm": 0.36713698506355286,
+      "learning_rate": 0.0005798986013986013,
+      "loss": 3.8702,
+      "step": 5850
+    },
+    {
+      "epoch": 1.718154813908789,
+      "grad_norm": 0.34020107984542847,
+      "learning_rate": 0.0005797237762237762,
+      "loss": 3.8835,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7327159415225115,
+      "grad_norm": 0.325308233499527,
+      "learning_rate": 0.000579548951048951,
+      "loss": 3.8765,
+      "step": 5950
+    },
+    {
+      "epoch": 1.747277069136234,
+      "grad_norm": 0.3312442898750305,
+      "learning_rate": 0.0005793741258741258,
+      "loss": 3.8783,
+      "step": 6000
+    },
+    {
+      "epoch": 1.747277069136234,
+      "eval_accuracy": 0.3375568091909506,
+      "eval_loss": 3.8538968563079834,
+      "eval_runtime": 182.6837,
+      "eval_samples_per_second": 91.119,
+      "eval_steps_per_second": 5.698,
+      "step": 6000
+    },
+    {
+      "epoch": 1.7618381967499563,
+      "grad_norm": 0.3400241732597351,
+      "learning_rate": 0.0005791993006993006,
+      "loss": 3.8657,
+      "step": 6050
+    },
+    {
+      "epoch": 1.7763993243636786,
+      "grad_norm": 0.3185133934020996,
+      "learning_rate": 0.0005790244755244755,
+      "loss": 3.8761,
+      "step": 6100
+    },
+    {
+      "epoch": 1.7909604519774012,
+      "grad_norm": 0.3425745964050293,
+      "learning_rate": 0.0005788496503496503,
+      "loss": 3.8653,
+      "step": 6150
+    },
+    {
+      "epoch": 1.8055215795911237,
+      "grad_norm": 0.31458574533462524,
+      "learning_rate": 0.0005786748251748251,
+      "loss": 3.8693,
+      "step": 6200
+    },
+    {
+      "epoch": 1.820082707204846,
+      "grad_norm": 0.33563944697380066,
+      "learning_rate": 0.0005784999999999999,
+      "loss": 3.8654,
+      "step": 6250
+    },
+    {
+      "epoch": 1.8346438348185683,
+      "grad_norm": 0.3224489092826843,
+      "learning_rate": 0.0005783251748251748,
+      "loss": 3.8461,
+      "step": 6300
+    },
+    {
+      "epoch": 1.8492049624322906,
+      "grad_norm": 0.3258609175682068,
+      "learning_rate": 0.0005781503496503496,
+      "loss": 3.857,
+      "step": 6350
+    },
+    {
+      "epoch": 1.8637660900460131,
+      "grad_norm": 0.31683316826820374,
+      "learning_rate": 0.0005779755244755244,
+      "loss": 3.8613,
+      "step": 6400
+    },
+    {
+      "epoch": 1.8783272176597356,
+      "grad_norm": 0.3589027523994446,
+      "learning_rate": 0.0005778006993006993,
+      "loss": 3.8568,
+      "step": 6450
+    },
+    {
+      "epoch": 1.892888345273458,
+      "grad_norm": 0.3233237862586975,
+      "learning_rate": 0.000577625874125874,
+      "loss": 3.8619,
+      "step": 6500
+    },
+    {
+      "epoch": 1.9074494728871803,
+      "grad_norm": 0.3178718388080597,
+      "learning_rate": 0.0005774510489510489,
+      "loss": 3.8528,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9220106005009028,
+      "grad_norm": 0.3323863446712494,
+      "learning_rate": 0.0005772762237762237,
+      "loss": 3.8432,
+      "step": 6600
+    },
+    {
+      "epoch": 1.9365717281146253,
+      "grad_norm": 0.3357718586921692,
+      "learning_rate": 0.0005771013986013985,
+      "loss": 3.8408,
+      "step": 6650
+    },
+    {
+      "epoch": 1.9511328557283476,
+      "grad_norm": 0.3276062607765198,
+      "learning_rate": 0.0005769265734265733,
+      "loss": 3.8554,
+      "step": 6700
+    },
+    {
+      "epoch": 1.96569398334207,
+      "grad_norm": 0.35431572794914246,
+      "learning_rate": 0.0005767517482517482,
+      "loss": 3.8394,
+      "step": 6750
+    },
+    {
+      "epoch": 1.9802551109557924,
+      "grad_norm": 0.3411652147769928,
+      "learning_rate": 0.000576576923076923,
+      "loss": 3.8523,
+      "step": 6800
+    },
+    {
+      "epoch": 1.994816238569515,
+      "grad_norm": 0.32246044278144836,
+      "learning_rate": 0.0005764020979020978,
+      "loss": 3.8446,
+      "step": 6850
+    },
+    {
+      "epoch": 2.009319121672782,
+      "grad_norm": 0.3211682438850403,
+      "learning_rate": 0.0005762272727272726,
+      "loss": 3.7718,
+      "step": 6900
+    },
+    {
+      "epoch": 2.023880249286505,
+      "grad_norm": 0.327195405960083,
+      "learning_rate": 0.0005760524475524475,
+      "loss": 3.7531,
+      "step": 6950
+    },
+    {
+      "epoch": 2.038441376900227,
+      "grad_norm": 0.327593058347702,
+      "learning_rate": 0.0005758776223776223,
+      "loss": 3.7409,
+      "step": 7000
+    },
+    {
+      "epoch": 2.038441376900227,
+      "eval_accuracy": 0.3418543103036807,
+      "eval_loss": 3.807173013687134,
+      "eval_runtime": 182.7201,
+      "eval_samples_per_second": 91.101,
+      "eval_steps_per_second": 5.697,
+      "step": 7000
+    },
+    {
+      "epoch": 2.0530025045139495,
+      "grad_norm": 0.31732338666915894,
+      "learning_rate": 0.0005757027972027971,
+      "loss": 3.7579,
+      "step": 7050
+    },
+    {
+      "epoch": 2.067563632127672,
+      "grad_norm": 0.31844356656074524,
+      "learning_rate": 0.000575527972027972,
+      "loss": 3.7387,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0821247597413945,
+      "grad_norm": 0.3230411410331726,
+      "learning_rate": 0.0005753531468531468,
+      "loss": 3.751,
+      "step": 7150
+    },
+    {
+      "epoch": 2.096685887355117,
+      "grad_norm": 0.32469573616981506,
+      "learning_rate": 0.0005751783216783216,
+      "loss": 3.7505,
+      "step": 7200
+    },
+    {
+      "epoch": 2.111247014968839,
+      "grad_norm": 0.32758161425590515,
+      "learning_rate": 0.0005750034965034964,
+      "loss": 3.749,
+      "step": 7250
+    },
+    {
+      "epoch": 2.1258081425825615,
+      "grad_norm": 0.32330748438835144,
+      "learning_rate": 0.0005748286713286712,
+      "loss": 3.7479,
+      "step": 7300
+    },
+    {
+      "epoch": 2.140369270196284,
+      "grad_norm": 0.3409372568130493,
+      "learning_rate": 0.000574653846153846,
+      "loss": 3.7407,
+      "step": 7350
+    },
+    {
+      "epoch": 2.1549303978100065,
+      "grad_norm": 0.3466344177722931,
+      "learning_rate": 0.000574479020979021,
+      "loss": 3.7507,
+      "step": 7400
+    },
+    {
+      "epoch": 2.169491525423729,
+      "grad_norm": 0.3374614119529724,
+      "learning_rate": 0.0005743041958041958,
+      "loss": 3.752,
+      "step": 7450
+    },
+    {
+      "epoch": 2.184052653037451,
+      "grad_norm": 0.31741416454315186,
+      "learning_rate": 0.0005741293706293706,
+      "loss": 3.7516,
+      "step": 7500
+    },
+    {
+      "epoch": 2.198613780651174,
+      "grad_norm": 0.3337715268135071,
+      "learning_rate": 0.0005739545454545454,
+      "loss": 3.7466,
+      "step": 7550
+    },
+    {
+      "epoch": 2.213174908264896,
+      "grad_norm": 0.3211740255355835,
+      "learning_rate": 0.0005737797202797203,
+      "loss": 3.7529,
+      "step": 7600
+    },
+    {
+      "epoch": 2.2277360358786185,
+      "grad_norm": 0.3633030652999878,
+      "learning_rate": 0.0005736048951048951,
+      "loss": 3.7411,
+      "step": 7650
+    },
+    {
+      "epoch": 2.2422971634923408,
+      "grad_norm": 0.3291577100753784,
+      "learning_rate": 0.0005734300699300699,
+      "loss": 3.7522,
+      "step": 7700
+    },
+    {
+      "epoch": 2.256858291106063,
+      "grad_norm": 0.32250499725341797,
+      "learning_rate": 0.0005732552447552448,
+      "loss": 3.7576,
+      "step": 7750
+    },
+    {
+      "epoch": 2.271419418719786,
+      "grad_norm": 0.30450233817100525,
+      "learning_rate": 0.0005730804195804196,
+      "loss": 3.7453,
+      "step": 7800
+    },
+    {
+      "epoch": 2.285980546333508,
+      "grad_norm": 0.30964726209640503,
+      "learning_rate": 0.0005729055944055944,
+      "loss": 3.755,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3005416739472304,
+      "grad_norm": 0.31292903423309326,
+      "learning_rate": 0.0005727307692307692,
+      "loss": 3.7539,
+      "step": 7900
+    },
+    {
+      "epoch": 2.3151028015609527,
+      "grad_norm": 0.3473982512950897,
+      "learning_rate": 0.0005725559440559441,
+      "loss": 3.7509,
+      "step": 7950
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "grad_norm": 0.35157695412635803,
+      "learning_rate": 0.0005723811188811188,
+      "loss": 3.7556,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3296639291746755,
+      "eval_accuracy": 0.3448868377610154,
+      "eval_loss": 3.778634786605835,
+      "eval_runtime": 184.554,
+      "eval_samples_per_second": 90.196,
+      "eval_steps_per_second": 5.641,
+      "step": 8000
+    },
+    {
+      "epoch": 2.3442250567883978,
+      "grad_norm": 0.3254748582839966,
+      "learning_rate": 0.0005722062937062937,
+      "loss": 3.7366,
+      "step": 8050
+    },
+    {
+      "epoch": 2.35878618440212,
+      "grad_norm": 0.33637189865112305,
+      "learning_rate": 0.0005720314685314685,
+      "loss": 3.749,
+      "step": 8100
+    },
+    {
+      "epoch": 2.3733473120158424,
+      "grad_norm": 0.3259914517402649,
+      "learning_rate": 0.0005718566433566433,
+      "loss": 3.7439,
+      "step": 8150
+    },
+    {
+      "epoch": 2.387908439629565,
+      "grad_norm": 0.31481459736824036,
+      "learning_rate": 0.0005716818181818181,
+      "loss": 3.7376,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4024695672432874,
+      "grad_norm": 0.3136726915836334,
+      "learning_rate": 0.000571506993006993,
+      "loss": 3.7284,
+      "step": 8250
+    },
+    {
+      "epoch": 2.4170306948570097,
+      "grad_norm": 0.32896652817726135,
+      "learning_rate": 0.0005713321678321678,
+      "loss": 3.7478,
+      "step": 8300
+    },
+    {
+      "epoch": 2.431591822470732,
+      "grad_norm": 0.3367721140384674,
+      "learning_rate": 0.0005711573426573426,
+      "loss": 3.7418,
+      "step": 8350
+    },
+    {
+      "epoch": 2.4461529500844543,
+      "grad_norm": 0.32438716292381287,
+      "learning_rate": 0.0005709825174825175,
+      "loss": 3.752,
+      "step": 8400
+    },
+    {
+      "epoch": 2.460714077698177,
+      "grad_norm": 0.3197824954986572,
+      "learning_rate": 0.0005708076923076923,
+      "loss": 3.7425,
+      "step": 8450
+    },
+    {
+      "epoch": 2.4752752053118994,
+      "grad_norm": 0.3408988416194916,
+      "learning_rate": 0.0005706328671328671,
+      "loss": 3.7424,
+      "step": 8500
+    },
+    {
+      "epoch": 2.4898363329256217,
+      "grad_norm": 0.3419415354728699,
+      "learning_rate": 0.0005704580419580419,
+      "loss": 3.7546,
+      "step": 8550
+    },
+    {
+      "epoch": 2.5043974605393444,
+      "grad_norm": 0.3166530132293701,
+      "learning_rate": 0.0005702832167832168,
+      "loss": 3.7339,
+      "step": 8600
+    },
+    {
+      "epoch": 2.5189585881530667,
+      "grad_norm": 0.3352281451225281,
+      "learning_rate": 0.0005701083916083916,
+      "loss": 3.7278,
+      "step": 8650
+    },
+    {
+      "epoch": 2.533519715766789,
+      "grad_norm": 0.31808385252952576,
+      "learning_rate": 0.0005699335664335664,
+      "loss": 3.7421,
+      "step": 8700
+    },
+    {
+      "epoch": 2.5480808433805113,
+      "grad_norm": 0.3182724714279175,
+      "learning_rate": 0.0005697587412587412,
+      "loss": 3.7366,
+      "step": 8750
+    },
+    {
+      "epoch": 2.5626419709942336,
+      "grad_norm": 0.3110598623752594,
+      "learning_rate": 0.000569583916083916,
+      "loss": 3.7345,
+      "step": 8800
+    },
+    {
+      "epoch": 2.5772030986079564,
+      "grad_norm": 0.3083634674549103,
+      "learning_rate": 0.0005694090909090908,
+      "loss": 3.7335,
+      "step": 8850
+    },
+    {
+      "epoch": 2.5917642262216787,
+      "grad_norm": 0.31325507164001465,
+      "learning_rate": 0.0005692342657342657,
+      "loss": 3.735,
+      "step": 8900
+    },
+    {
+      "epoch": 2.606325353835401,
+      "grad_norm": 0.325335830450058,
+      "learning_rate": 0.0005690594405594405,
+      "loss": 3.7352,
+      "step": 8950
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "grad_norm": 0.3095828592777252,
+      "learning_rate": 0.0005688846153846153,
+      "loss": 3.7318,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6208864814491233,
+      "eval_accuracy": 0.34765026441006025,
+      "eval_loss": 3.7488150596618652,
+      "eval_runtime": 181.151,
+      "eval_samples_per_second": 91.89,
+      "eval_steps_per_second": 5.747,
+      "step": 9000
+    },
+    {
+      "epoch": 2.6354476090628456,
+      "grad_norm": 0.32213926315307617,
+      "learning_rate": 0.0005687097902097901,
+      "loss": 3.7308,
+      "step": 9050
+    },
+    {
+      "epoch": 2.6500087366765683,
+      "grad_norm": 0.3269072473049164,
+      "learning_rate": 0.000568534965034965,
+      "loss": 3.7258,
+      "step": 9100
+    },
+    {
+      "epoch": 2.6645698642902906,
+      "grad_norm": 0.3074745833873749,
+      "learning_rate": 0.0005683601398601398,
+      "loss": 3.7307,
+      "step": 9150
+    },
+    {
+      "epoch": 2.679130991904013,
+      "grad_norm": 0.3062356412410736,
+      "learning_rate": 0.0005681853146853146,
+      "loss": 3.7292,
+      "step": 9200
+    },
+    {
+      "epoch": 2.6936921195177357,
+      "grad_norm": 0.3062867820262909,
+      "learning_rate": 0.0005680104895104895,
+      "loss": 3.7162,
+      "step": 9250
+    },
+    {
+      "epoch": 2.708253247131458,
+      "grad_norm": 0.3553263545036316,
+      "learning_rate": 0.0005678356643356643,
+      "loss": 3.7239,
+      "step": 9300
+    },
+    {
+      "epoch": 2.7228143747451803,
+      "grad_norm": 0.31543877720832825,
+      "learning_rate": 0.0005676608391608391,
+      "loss": 3.7271,
+      "step": 9350
+    },
+    {
+      "epoch": 2.7373755023589026,
+      "grad_norm": 0.3128660023212433,
+      "learning_rate": 0.0005674860139860139,
+      "loss": 3.7304,
+      "step": 9400
+    },
+    {
+      "epoch": 2.751936629972625,
+      "grad_norm": 0.31978827714920044,
+      "learning_rate": 0.0005673111888111888,
+      "loss": 3.7252,
+      "step": 9450
+    },
+    {
+      "epoch": 2.7664977575863476,
+      "grad_norm": 0.3351915776729584,
+      "learning_rate": 0.0005671363636363635,
+      "loss": 3.7215,
+      "step": 9500
+    },
+    {
+      "epoch": 2.78105888520007,
+      "grad_norm": 0.32556962966918945,
+      "learning_rate": 0.0005669615384615384,
+      "loss": 3.7109,
+      "step": 9550
+    },
+    {
+      "epoch": 2.7956200128137922,
+      "grad_norm": 0.30667567253112793,
+      "learning_rate": 0.0005667867132867132,
+      "loss": 3.7309,
+      "step": 9600
+    },
+    {
+      "epoch": 2.8101811404275145,
+      "grad_norm": 0.2958901524543762,
+      "learning_rate": 0.000566611888111888,
+      "loss": 3.7281,
+      "step": 9650
+    },
+    {
+      "epoch": 2.824742268041237,
+      "grad_norm": 0.3355960249900818,
+      "learning_rate": 0.0005664370629370628,
+      "loss": 3.7114,
+      "step": 9700
+    },
+    {
+      "epoch": 2.8393033956549596,
+      "grad_norm": 0.3073790967464447,
+      "learning_rate": 0.0005662622377622377,
+      "loss": 3.7258,
+      "step": 9750
+    },
+    {
+      "epoch": 2.853864523268682,
+      "grad_norm": 0.30925753712654114,
+      "learning_rate": 0.0005660874125874125,
+      "loss": 3.721,
+      "step": 9800
+    },
+    {
+      "epoch": 2.868425650882404,
+      "grad_norm": 0.3250259459018707,
+      "learning_rate": 0.0005659125874125873,
+      "loss": 3.7338,
+      "step": 9850
+    },
+    {
+      "epoch": 2.882986778496127,
+      "grad_norm": 0.3341335356235504,
+      "learning_rate": 0.0005657377622377622,
+      "loss": 3.7089,
+      "step": 9900
+    },
+    {
+      "epoch": 2.8975479061098492,
+      "grad_norm": 0.325226366519928,
+      "learning_rate": 0.000565562937062937,
+      "loss": 3.7195,
+      "step": 9950
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "grad_norm": 0.32413652539253235,
+      "learning_rate": 0.0005653881118881118,
+      "loss": 3.7152,
+      "step": 10000
+    },
+    {
+      "epoch": 2.9121090337235715,
+      "eval_accuracy": 0.3496235527749125,
+      "eval_loss": 3.7240853309631348,
+      "eval_runtime": 180.9488,
+      "eval_samples_per_second": 91.993,
+      "eval_steps_per_second": 5.753,
+      "step": 10000
+    },
+    {
+      "epoch": 2.926670161337294,
+      "grad_norm": 0.3107230067253113,
+      "learning_rate": 0.0005652132867132866,
+      "loss": 3.7198,
+      "step": 10050
+    },
+    {
+      "epoch": 2.941231288951016,
+      "grad_norm": 0.3085416257381439,
+      "learning_rate": 0.0005650384615384615,
+      "loss": 3.7137,
+      "step": 10100
+    },
+    {
+      "epoch": 2.955792416564739,
+      "grad_norm": 0.3245227634906769,
+      "learning_rate": 0.0005648636363636363,
+      "loss": 3.7077,
+      "step": 10150
+    },
+    {
+      "epoch": 2.970353544178461,
+      "grad_norm": 0.31815105676651,
+      "learning_rate": 0.0005646888111888111,
+      "loss": 3.7133,
+      "step": 10200
+    },
+    {
+      "epoch": 2.9849146717921835,
+      "grad_norm": 0.3235960900783539,
+      "learning_rate": 0.000564513986013986,
+      "loss": 3.7025,
+      "step": 10250
+    },
+    {
+      "epoch": 2.9994757994059063,
+      "grad_norm": 0.34538596868515015,
+      "learning_rate": 0.0005643391608391607,
+      "loss": 3.7129,
+      "step": 10300
+    },
+    {
+      "epoch": 3.0139786825091734,
+      "grad_norm": 0.31785765290260315,
+      "learning_rate": 0.0005641643356643355,
+      "loss": 3.6054,
+      "step": 10350
+    },
+    {
+      "epoch": 3.0285398101228957,
+      "grad_norm": 0.3313532769680023,
+      "learning_rate": 0.0005639895104895105,
+      "loss": 3.6084,
+      "step": 10400
+    },
+    {
+      "epoch": 3.0431009377366185,
+      "grad_norm": 0.3206842839717865,
+      "learning_rate": 0.0005638146853146853,
+      "loss": 3.6094,
+      "step": 10450
+    },
+    {
+      "epoch": 3.057662065350341,
+      "grad_norm": 0.3367646336555481,
+      "learning_rate": 0.0005636398601398601,
+      "loss": 3.6184,
+      "step": 10500
+    },
+    {
+      "epoch": 3.072223192964063,
+      "grad_norm": 0.3168485164642334,
+      "learning_rate": 0.000563465034965035,
+      "loss": 3.6267,
+      "step": 10550
+    },
+    {
+      "epoch": 3.0867843205777854,
+      "grad_norm": 0.3227006196975708,
+      "learning_rate": 0.0005632902097902098,
+      "loss": 3.6165,
+      "step": 10600
+    },
+    {
+      "epoch": 3.101345448191508,
+      "grad_norm": 0.3266565799713135,
+      "learning_rate": 0.0005631153846153846,
+      "loss": 3.6185,
+      "step": 10650
+    },
+    {
+      "epoch": 3.1159065758052304,
+      "grad_norm": 0.3158946633338928,
+      "learning_rate": 0.0005629405594405594,
+      "loss": 3.6104,
+      "step": 10700
+    },
+    {
+      "epoch": 3.1304677034189528,
+      "grad_norm": 0.32722142338752747,
+      "learning_rate": 0.0005627657342657343,
+      "loss": 3.619,
+      "step": 10750
+    },
+    {
+      "epoch": 3.145028831032675,
+      "grad_norm": 0.30216503143310547,
+      "learning_rate": 0.0005625909090909091,
+      "loss": 3.629,
+      "step": 10800
+    },
+    {
+      "epoch": 3.1595899586463974,
+      "grad_norm": 0.3186177909374237,
+      "learning_rate": 0.0005624160839160839,
+      "loss": 3.6195,
+      "step": 10850
+    },
+    {
+      "epoch": 3.17415108626012,
+      "grad_norm": 0.3162807822227478,
+      "learning_rate": 0.0005622412587412587,
+      "loss": 3.6169,
+      "step": 10900
+    },
+    {
+      "epoch": 3.1887122138738424,
+      "grad_norm": 0.33053115010261536,
+      "learning_rate": 0.0005620664335664336,
+      "loss": 3.6338,
+      "step": 10950
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "grad_norm": 0.30891671776771545,
+      "learning_rate": 0.0005618916083916083,
+      "loss": 3.6271,
+      "step": 11000
+    },
+    {
+      "epoch": 3.2032733414875647,
+      "eval_accuracy": 0.3519776264250645,
+      "eval_loss": 3.709869384765625,
+      "eval_runtime": 180.919,
+      "eval_samples_per_second": 92.008,
+      "eval_steps_per_second": 5.754,
+      "step": 11000
+    },
+    {
+      "epoch": 3.217834469101287,
+      "grad_norm": 0.33202117681503296,
+      "learning_rate": 0.0005617167832167832,
+      "loss": 3.6235,
+      "step": 11050
+    },
+    {
+      "epoch": 3.2323955967150098,
+      "grad_norm": 0.3139243423938751,
+      "learning_rate": 0.000561541958041958,
+      "loss": 3.6294,
+      "step": 11100
+    },
+    {
+      "epoch": 3.246956724328732,
+      "grad_norm": 0.3314415216445923,
+      "learning_rate": 0.0005613671328671328,
+      "loss": 3.6461,
+      "step": 11150
+    },
+    {
+      "epoch": 3.2615178519424544,
+      "grad_norm": 0.3135373592376709,
+      "learning_rate": 0.0005611923076923077,
+      "loss": 3.6442,
+      "step": 11200
+    },
+    {
+      "epoch": 3.2760789795561767,
+      "grad_norm": 0.32920390367507935,
+      "learning_rate": 0.0005610174825174825,
+      "loss": 3.6266,
+      "step": 11250
+    },
+    {
+      "epoch": 3.2906401071698994,
+      "grad_norm": 0.3141324520111084,
+      "learning_rate": 0.0005608426573426573,
+      "loss": 3.641,
+      "step": 11300
+    },
+    {
+      "epoch": 3.3052012347836217,
+      "grad_norm": 0.3209163546562195,
+      "learning_rate": 0.0005606678321678321,
+      "loss": 3.6214,
+      "step": 11350
+    },
+    {
+      "epoch": 3.319762362397344,
+      "grad_norm": 0.33475250005722046,
+      "learning_rate": 0.000560493006993007,
+      "loss": 3.6157,
+      "step": 11400
+    },
+    {
+      "epoch": 3.3343234900110663,
+      "grad_norm": 0.33062008023262024,
+      "learning_rate": 0.0005603181818181818,
+      "loss": 3.6323,
+      "step": 11450
+    },
+    {
+      "epoch": 3.3488846176247886,
+      "grad_norm": 0.32883119583129883,
+      "learning_rate": 0.0005601433566433566,
+      "loss": 3.6385,
+      "step": 11500
+    },
+    {
+      "epoch": 3.3634457452385114,
+      "grad_norm": 0.31883203983306885,
+      "learning_rate": 0.0005599685314685314,
+      "loss": 3.6394,
+      "step": 11550
+    },
+    {
+      "epoch": 3.3780068728522337,
+      "grad_norm": 0.30644893646240234,
+      "learning_rate": 0.0005597937062937063,
+      "loss": 3.6358,
+      "step": 11600
+    },
+    {
+      "epoch": 3.392568000465956,
+      "grad_norm": 0.30966854095458984,
+      "learning_rate": 0.0005596188811188811,
+      "loss": 3.6315,
+      "step": 11650
+    },
+    {
+      "epoch": 3.4071291280796787,
+      "grad_norm": 0.33119064569473267,
+      "learning_rate": 0.0005594440559440559,
+      "loss": 3.6342,
+      "step": 11700
+    },
+    {
+      "epoch": 3.421690255693401,
+      "grad_norm": 0.340206116437912,
+      "learning_rate": 0.0005592692307692307,
+      "loss": 3.6359,
+      "step": 11750
+    },
+    {
+      "epoch": 3.4362513833071233,
+      "grad_norm": 0.3317243754863739,
+      "learning_rate": 0.0005590944055944055,
+      "loss": 3.6418,
+      "step": 11800
+    },
+    {
+      "epoch": 3.4508125109208456,
+      "grad_norm": 0.35198959708213806,
+      "learning_rate": 0.0005589195804195803,
+      "loss": 3.6433,
+      "step": 11850
+    },
+    {
+      "epoch": 3.465373638534568,
+      "grad_norm": 0.3379652202129364,
+      "learning_rate": 0.0005587447552447552,
+      "loss": 3.63,
+      "step": 11900
+    },
+    {
+      "epoch": 3.4799347661482907,
+      "grad_norm": 0.3148704171180725,
+      "learning_rate": 0.00055856993006993,
+      "loss": 3.6433,
+      "step": 11950
+    },
+    {
+      "epoch": 3.494495893762013,
+      "grad_norm": 0.32295045256614685,
+      "learning_rate": 0.0005583951048951048,
+      "loss": 3.637,
+      "step": 12000
+    },
+    {
+      "epoch": 3.494495893762013,
+      "eval_accuracy": 0.3538141894775353,
+      "eval_loss": 3.6922500133514404,
+      "eval_runtime": 181.0231,
+      "eval_samples_per_second": 91.955,
+      "eval_steps_per_second": 5.751,
+      "step": 12000
+    },
+    {
+      "epoch": 3.5090570213757353,
+      "grad_norm": 0.33041051030158997,
+      "learning_rate": 0.0005582202797202797,
+      "loss": 3.6393,
+      "step": 12050
+    },
+    {
+      "epoch": 3.523618148989458,
+      "grad_norm": 0.30944448709487915,
+      "learning_rate": 0.0005580454545454545,
+      "loss": 3.636,
+      "step": 12100
+    },
+    {
+      "epoch": 3.53817927660318,
+      "grad_norm": 0.3318426311016083,
+      "learning_rate": 0.0005578706293706293,
+      "loss": 3.6265,
+      "step": 12150
+    },
+    {
+      "epoch": 3.5527404042169026,
+      "grad_norm": 0.31078043580055237,
+      "learning_rate": 0.0005576958041958041,
+      "loss": 3.6335,
+      "step": 12200
+    },
+    {
+      "epoch": 3.567301531830625,
+      "grad_norm": 0.31694719195365906,
+      "learning_rate": 0.000557520979020979,
+      "loss": 3.6387,
+      "step": 12250
+    },
+    {
+      "epoch": 3.5818626594443472,
+      "grad_norm": 0.32518377900123596,
+      "learning_rate": 0.0005573461538461538,
+      "loss": 3.6292,
+      "step": 12300
+    },
+    {
+      "epoch": 3.59642378705807,
+      "grad_norm": 0.33264607191085815,
+      "learning_rate": 0.0005571713286713286,
+      "loss": 3.6424,
+      "step": 12350
+    },
+    {
+      "epoch": 3.6109849146717923,
+      "grad_norm": 0.31201255321502686,
+      "learning_rate": 0.0005569965034965034,
+      "loss": 3.6499,
+      "step": 12400
+    },
+    {
+      "epoch": 3.6255460422855146,
+      "grad_norm": 0.32355648279190063,
+      "learning_rate": 0.0005568216783216783,
+      "loss": 3.6397,
+      "step": 12450
+    },
+    {
+      "epoch": 3.640107169899237,
+      "grad_norm": 0.3250090479850769,
+      "learning_rate": 0.000556646853146853,
+      "loss": 3.6408,
+      "step": 12500
+    },
+    {
+      "epoch": 3.654668297512959,
+      "grad_norm": 0.3053436577320099,
+      "learning_rate": 0.0005564720279720279,
+      "loss": 3.6246,
+      "step": 12550
+    },
+    {
+      "epoch": 3.669229425126682,
+      "grad_norm": 0.3021223545074463,
+      "learning_rate": 0.0005562972027972027,
+      "loss": 3.6422,
+      "step": 12600
+    },
+    {
+      "epoch": 3.6837905527404042,
+      "grad_norm": 0.30464479327201843,
+      "learning_rate": 0.0005561223776223775,
+      "loss": 3.6258,
+      "step": 12650
+    },
+    {
+      "epoch": 3.6983516803541265,
+      "grad_norm": 0.32816681265830994,
+      "learning_rate": 0.0005559475524475524,
+      "loss": 3.6374,
+      "step": 12700
+    },
+    {
+      "epoch": 3.7129128079678493,
+      "grad_norm": 0.3056129217147827,
+      "learning_rate": 0.0005557727272727272,
+      "loss": 3.6422,
+      "step": 12750
+    },
+    {
+      "epoch": 3.7274739355815716,
+      "grad_norm": 0.3144519031047821,
+      "learning_rate": 0.000555597902097902,
+      "loss": 3.6329,
+      "step": 12800
+    },
+    {
+      "epoch": 3.742035063195294,
+      "grad_norm": 0.3042009472846985,
+      "learning_rate": 0.0005554230769230768,
+      "loss": 3.6296,
+      "step": 12850
+    },
+    {
+      "epoch": 3.756596190809016,
+      "grad_norm": 0.3230903446674347,
+      "learning_rate": 0.0005552482517482517,
+      "loss": 3.6342,
+      "step": 12900
+    },
+    {
+      "epoch": 3.7711573184227385,
+      "grad_norm": 0.30217060446739197,
+      "learning_rate": 0.0005550734265734265,
+      "loss": 3.638,
+      "step": 12950
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "grad_norm": 0.3292492628097534,
+      "learning_rate": 0.0005548986013986013,
+      "loss": 3.6368,
+      "step": 13000
+    },
+    {
+      "epoch": 3.7857184460364612,
+      "eval_accuracy": 0.35537424527745126,
+      "eval_loss": 3.6732499599456787,
+      "eval_runtime": 181.1768,
+      "eval_samples_per_second": 91.877,
+      "eval_steps_per_second": 5.746,
+      "step": 13000
+    },
+    {
+      "epoch": 3.8002795736501835,
+      "grad_norm": 0.3120541274547577,
+      "learning_rate": 0.0005547237762237761,
+      "loss": 3.6445,
+      "step": 13050
+    },
+    {
+      "epoch": 3.814840701263906,
+      "grad_norm": 0.3175022006034851,
+      "learning_rate": 0.000554548951048951,
+      "loss": 3.6454,
+      "step": 13100
+    },
+    {
+      "epoch": 3.829401828877628,
+      "grad_norm": 0.3213273286819458,
+      "learning_rate": 0.0005543741258741258,
+      "loss": 3.6236,
+      "step": 13150
+    },
+    {
+      "epoch": 3.8439629564913504,
+      "grad_norm": 0.3022514283657074,
+      "learning_rate": 0.0005541993006993006,
+      "loss": 3.6383,
+      "step": 13200
+    },
+    {
+      "epoch": 3.858524084105073,
+      "grad_norm": 0.3136868476867676,
+      "learning_rate": 0.0005540244755244756,
+      "loss": 3.6342,
+      "step": 13250
+    },
+    {
+      "epoch": 3.8730852117187955,
+      "grad_norm": 0.3210989236831665,
+      "learning_rate": 0.0005538496503496502,
+      "loss": 3.636,
+      "step": 13300
+    },
+    {
+      "epoch": 3.887646339332518,
+      "grad_norm": 0.31793737411499023,
+      "learning_rate": 0.0005536748251748252,
+      "loss": 3.6384,
+      "step": 13350
+    },
+    {
+      "epoch": 3.9022074669462405,
+      "grad_norm": 0.3172999620437622,
+      "learning_rate": 0.0005535,
+      "loss": 3.6322,
+      "step": 13400
+    },
+    {
+      "epoch": 3.916768594559963,
+      "grad_norm": 0.3040863573551178,
+      "learning_rate": 0.0005533251748251748,
+      "loss": 3.6361,
+      "step": 13450
+    },
+    {
+      "epoch": 3.931329722173685,
+      "grad_norm": 0.2977621555328369,
+      "learning_rate": 0.0005531503496503496,
+      "loss": 3.6293,
+      "step": 13500
+    },
+    {
+      "epoch": 3.9458908497874075,
+      "grad_norm": 0.32015374302864075,
+      "learning_rate": 0.0005529755244755245,
+      "loss": 3.6437,
+      "step": 13550
+    },
+    {
+      "epoch": 3.9604519774011298,
+      "grad_norm": 0.30774080753326416,
+      "learning_rate": 0.0005528006993006993,
+      "loss": 3.6305,
+      "step": 13600
+    },
+    {
+      "epoch": 3.9750131050148525,
+      "grad_norm": 0.32130375504493713,
+      "learning_rate": 0.0005526258741258741,
+      "loss": 3.6176,
+      "step": 13650
+    },
+    {
+      "epoch": 3.989574232628575,
+      "grad_norm": 0.3188425600528717,
+      "learning_rate": 0.0005524510489510489,
+      "loss": 3.6412,
+      "step": 13700
+    },
+    {
+      "epoch": 4.004077115731842,
+      "grad_norm": 0.31326737999916077,
+      "learning_rate": 0.0005522762237762238,
+      "loss": 3.599,
+      "step": 13750
+    },
+    {
+      "epoch": 4.018638243345564,
+      "grad_norm": 0.30300453305244446,
+      "learning_rate": 0.0005521013986013986,
+      "loss": 3.5165,
+      "step": 13800
+    },
+    {
+      "epoch": 4.033199370959287,
+      "grad_norm": 0.3266518712043762,
+      "learning_rate": 0.0005519265734265734,
+      "loss": 3.5381,
+      "step": 13850
+    },
+    {
+      "epoch": 4.04776049857301,
+      "grad_norm": 0.3203023076057434,
+      "learning_rate": 0.0005517517482517482,
+      "loss": 3.5187,
+      "step": 13900
+    },
+    {
+      "epoch": 4.062321626186732,
+      "grad_norm": 0.3111482262611389,
+      "learning_rate": 0.0005515769230769231,
+      "loss": 3.5362,
+      "step": 13950
+    },
+    {
+      "epoch": 4.076882753800454,
+      "grad_norm": 0.34344446659088135,
+      "learning_rate": 0.0005514020979020979,
+      "loss": 3.5417,
+      "step": 14000
+    },
+    {
+      "epoch": 4.076882753800454,
+      "eval_accuracy": 0.35658572794648924,
+      "eval_loss": 3.6685900688171387,
+      "eval_runtime": 180.9843,
+      "eval_samples_per_second": 91.975,
+      "eval_steps_per_second": 5.752,
+      "step": 14000
+    },
+    {
+      "epoch": 4.091443881414177,
+      "grad_norm": 0.3067654073238373,
+      "learning_rate": 0.0005512272727272727,
+      "loss": 3.539,
+      "step": 14050
+    },
+    {
+      "epoch": 4.106005009027899,
+      "grad_norm": 0.30828049778938293,
+      "learning_rate": 0.0005510524475524475,
+      "loss": 3.5469,
+      "step": 14100
+    },
+    {
+      "epoch": 4.120566136641622,
+      "grad_norm": 0.31522074341773987,
+      "learning_rate": 0.0005508776223776223,
+      "loss": 3.5362,
+      "step": 14150
+    },
+    {
+      "epoch": 4.135127264255344,
+      "grad_norm": 0.3391481637954712,
+      "learning_rate": 0.0005507027972027972,
+      "loss": 3.5471,
+      "step": 14200
+    },
+    {
+      "epoch": 4.149688391869066,
+      "grad_norm": 0.3163962662220001,
+      "learning_rate": 0.000550527972027972,
+      "loss": 3.5409,
+      "step": 14250
+    },
+    {
+      "epoch": 4.164249519482789,
+      "grad_norm": 0.3650486171245575,
+      "learning_rate": 0.0005503531468531468,
+      "loss": 3.5534,
+      "step": 14300
+    },
+    {
+      "epoch": 4.178810647096511,
+      "grad_norm": 0.3179774284362793,
+      "learning_rate": 0.0005501783216783216,
+      "loss": 3.5485,
+      "step": 14350
+    },
+    {
+      "epoch": 4.193371774710234,
+      "grad_norm": 0.3414445221424103,
+      "learning_rate": 0.0005500034965034965,
+      "loss": 3.5559,
+      "step": 14400
+    },
+    {
+      "epoch": 4.207932902323956,
+      "grad_norm": 0.3165196180343628,
+      "learning_rate": 0.0005498286713286713,
+      "loss": 3.5449,
+      "step": 14450
+    },
+    {
+      "epoch": 4.222494029937678,
+      "grad_norm": 0.2997719347476959,
+      "learning_rate": 0.0005496538461538461,
+      "loss": 3.5552,
+      "step": 14500
+    },
+    {
+      "epoch": 4.237055157551401,
+      "grad_norm": 0.3242793679237366,
+      "learning_rate": 0.0005494790209790209,
+      "loss": 3.5649,
+      "step": 14550
+    },
+    {
+      "epoch": 4.251616285165123,
+      "grad_norm": 0.3411242663860321,
+      "learning_rate": 0.0005493041958041958,
+      "loss": 3.5586,
+      "step": 14600
+    },
+    {
+      "epoch": 4.266177412778846,
+      "grad_norm": 0.3131914734840393,
+      "learning_rate": 0.0005491293706293706,
+      "loss": 3.5638,
+      "step": 14650
+    },
+    {
+      "epoch": 4.280738540392568,
+      "grad_norm": 0.3113696873188019,
+      "learning_rate": 0.0005489545454545454,
+      "loss": 3.5583,
+      "step": 14700
+    },
+    {
+      "epoch": 4.29529966800629,
+      "grad_norm": 0.3232826590538025,
+      "learning_rate": 0.0005487797202797203,
+      "loss": 3.5635,
+      "step": 14750
+    },
+    {
+      "epoch": 4.309860795620013,
+      "grad_norm": 0.3019048273563385,
+      "learning_rate": 0.000548604895104895,
+      "loss": 3.5583,
+      "step": 14800
+    },
+    {
+      "epoch": 4.324421923233735,
+      "grad_norm": 0.30543556809425354,
+      "learning_rate": 0.0005484300699300699,
+      "loss": 3.5636,
+      "step": 14850
+    },
+    {
+      "epoch": 4.338983050847458,
+      "grad_norm": 0.327394038438797,
+      "learning_rate": 0.0005482552447552447,
+      "loss": 3.5602,
+      "step": 14900
+    },
+    {
+      "epoch": 4.35354417846118,
+      "grad_norm": 0.30762121081352234,
+      "learning_rate": 0.0005480804195804195,
+      "loss": 3.555,
+      "step": 14950
+    },
+    {
+      "epoch": 4.368105306074902,
+      "grad_norm": 0.3166114389896393,
+      "learning_rate": 0.0005479055944055943,
+      "loss": 3.5573,
+      "step": 15000
+    },
+    {
+      "epoch": 4.368105306074902,
+      "eval_accuracy": 0.3576027620629228,
+      "eval_loss": 3.6566529273986816,
+      "eval_runtime": 181.0498,
+      "eval_samples_per_second": 91.942,
+      "eval_steps_per_second": 5.75,
+      "step": 15000
+    },
+    {
+      "epoch": 4.382666433688625,
+      "grad_norm": 0.3189477026462555,
+      "learning_rate": 0.0005477307692307692,
+      "loss": 3.5526,
+      "step": 15050
+    },
+    {
+      "epoch": 4.397227561302348,
+      "grad_norm": 0.31075966358184814,
+      "learning_rate": 0.000547555944055944,
+      "loss": 3.5699,
+      "step": 15100
+    },
+    {
+      "epoch": 4.41178868891607,
+      "grad_norm": 0.33162787556648254,
+      "learning_rate": 0.0005473811188811188,
+      "loss": 3.573,
+      "step": 15150
+    },
+    {
+      "epoch": 4.426349816529792,
+      "grad_norm": 0.31193026900291443,
+      "learning_rate": 0.0005472062937062936,
+      "loss": 3.5564,
+      "step": 15200
+    },
+    {
+      "epoch": 4.440910944143514,
+      "grad_norm": 0.30927059054374695,
+      "learning_rate": 0.0005470314685314685,
+      "loss": 3.5524,
+      "step": 15250
+    },
+    {
+      "epoch": 4.455472071757237,
+      "grad_norm": 0.31786641478538513,
+      "learning_rate": 0.0005468566433566433,
+      "loss": 3.5712,
+      "step": 15300
+    },
+    {
+      "epoch": 4.47003319937096,
+      "grad_norm": 0.31433233618736267,
+      "learning_rate": 0.0005466818181818181,
+      "loss": 3.5586,
+      "step": 15350
+    },
+    {
+      "epoch": 4.4845943269846815,
+      "grad_norm": 0.3186981976032257,
+      "learning_rate": 0.000546506993006993,
+      "loss": 3.571,
+      "step": 15400
+    },
+    {
+      "epoch": 4.499155454598404,
+      "grad_norm": 0.31594040989875793,
+      "learning_rate": 0.0005463321678321678,
+      "loss": 3.5668,
+      "step": 15450
+    },
+    {
+      "epoch": 4.513716582212126,
+      "grad_norm": 0.3135336637496948,
+      "learning_rate": 0.0005461573426573426,
+      "loss": 3.5652,
+      "step": 15500
+    },
+    {
+      "epoch": 4.528277709825849,
+      "grad_norm": 0.37292537093162537,
+      "learning_rate": 0.0005459825174825174,
+      "loss": 3.5808,
+      "step": 15550
+    },
+    {
+      "epoch": 4.542838837439572,
+      "grad_norm": 0.3025301694869995,
+      "learning_rate": 0.0005458076923076922,
+      "loss": 3.5782,
+      "step": 15600
+    },
+    {
+      "epoch": 4.5573999650532935,
+      "grad_norm": 0.3217926025390625,
+      "learning_rate": 0.000545632867132867,
+      "loss": 3.567,
+      "step": 15650
+    },
+    {
+      "epoch": 4.571961092667016,
+      "grad_norm": 0.33000338077545166,
+      "learning_rate": 0.0005454580419580419,
+      "loss": 3.5761,
+      "step": 15700
+    },
+    {
+      "epoch": 4.586522220280738,
+      "grad_norm": 0.3380807340145111,
+      "learning_rate": 0.0005452832167832167,
+      "loss": 3.5729,
+      "step": 15750
+    },
+    {
+      "epoch": 4.601083347894461,
+      "grad_norm": 0.2971727252006531,
+      "learning_rate": 0.0005451083916083915,
+      "loss": 3.5666,
+      "step": 15800
+    },
+    {
+      "epoch": 4.615644475508184,
+      "grad_norm": 0.32622653245925903,
+      "learning_rate": 0.0005449335664335663,
+      "loss": 3.5744,
+      "step": 15850
+    },
+    {
+      "epoch": 4.630205603121905,
+      "grad_norm": 0.33031386137008667,
+      "learning_rate": 0.0005447587412587412,
+      "loss": 3.5836,
+      "step": 15900
+    },
+    {
+      "epoch": 4.644766730735628,
+      "grad_norm": 0.31544092297554016,
+      "learning_rate": 0.000544583916083916,
+      "loss": 3.58,
+      "step": 15950
+    },
+    {
+      "epoch": 4.659327858349351,
+      "grad_norm": 0.32821419835090637,
+      "learning_rate": 0.0005444090909090908,
+      "loss": 3.5738,
+      "step": 16000
+    },
+    {
+      "epoch": 4.659327858349351,
+      "eval_accuracy": 0.35873136309375875,
+      "eval_loss": 3.642806053161621,
+      "eval_runtime": 181.1419,
+      "eval_samples_per_second": 91.895,
+      "eval_steps_per_second": 5.747,
+      "step": 16000
+    },
+    {
+      "epoch": 4.673888985963073,
+      "grad_norm": 0.313347727060318,
+      "learning_rate": 0.0005442342657342657,
+      "loss": 3.5561,
+      "step": 16050
+    },
+    {
+      "epoch": 4.6884501135767955,
+      "grad_norm": 0.29170653223991394,
+      "learning_rate": 0.0005440594405594405,
+      "loss": 3.5704,
+      "step": 16100
+    },
+    {
+      "epoch": 4.703011241190518,
+      "grad_norm": 0.3108586370944977,
+      "learning_rate": 0.0005438846153846153,
+      "loss": 3.5616,
+      "step": 16150
+    },
+    {
+      "epoch": 4.71757236880424,
+      "grad_norm": 0.31426018476486206,
+      "learning_rate": 0.0005437097902097901,
+      "loss": 3.569,
+      "step": 16200
+    },
+    {
+      "epoch": 4.732133496417963,
+      "grad_norm": 0.3121700584888458,
+      "learning_rate": 0.0005435349650349651,
+      "loss": 3.5648,
+      "step": 16250
+    },
+    {
+      "epoch": 4.746694624031685,
+      "grad_norm": 0.3235621154308319,
+      "learning_rate": 0.0005433601398601397,
+      "loss": 3.5713,
+      "step": 16300
+    },
+    {
+      "epoch": 4.7612557516454075,
+      "grad_norm": 0.3033587336540222,
+      "learning_rate": 0.0005431853146853147,
+      "loss": 3.5571,
+      "step": 16350
+    },
+    {
+      "epoch": 4.77581687925913,
+      "grad_norm": 0.3489319086074829,
+      "learning_rate": 0.0005430104895104895,
+      "loss": 3.5774,
+      "step": 16400
+    },
+    {
+      "epoch": 4.790378006872852,
+      "grad_norm": 0.3031330108642578,
+      "learning_rate": 0.0005428356643356643,
+      "loss": 3.571,
+      "step": 16450
+    },
+    {
+      "epoch": 4.804939134486575,
+      "grad_norm": 0.31163015961647034,
+      "learning_rate": 0.0005426608391608391,
+      "loss": 3.5646,
+      "step": 16500
+    },
+    {
+      "epoch": 4.819500262100297,
+      "grad_norm": 0.32922112941741943,
+      "learning_rate": 0.000542486013986014,
+      "loss": 3.5692,
+      "step": 16550
+    },
+    {
+      "epoch": 4.834061389714019,
+      "grad_norm": 0.32564711570739746,
+      "learning_rate": 0.0005423111888111888,
+      "loss": 3.5759,
+      "step": 16600
+    },
+    {
+      "epoch": 4.848622517327742,
+      "grad_norm": 0.3296327590942383,
+      "learning_rate": 0.0005421363636363636,
+      "loss": 3.5785,
+      "step": 16650
+    },
+    {
+      "epoch": 4.863183644941464,
+      "grad_norm": 0.3310580253601074,
+      "learning_rate": 0.0005419615384615385,
+      "loss": 3.5685,
+      "step": 16700
+    },
+    {
+      "epoch": 4.877744772555187,
+      "grad_norm": 0.31274470686912537,
+      "learning_rate": 0.0005417867132867133,
+      "loss": 3.5587,
+      "step": 16750
+    },
+    {
+      "epoch": 4.892305900168909,
+      "grad_norm": 0.323668897151947,
+      "learning_rate": 0.0005416118881118881,
+      "loss": 3.5657,
+      "step": 16800
+    },
+    {
+      "epoch": 4.906867027782631,
+      "grad_norm": 0.3014122247695923,
+      "learning_rate": 0.0005414370629370629,
+      "loss": 3.5707,
+      "step": 16850
+    },
+    {
+      "epoch": 4.921428155396354,
+      "grad_norm": 0.32767361402511597,
+      "learning_rate": 0.0005412622377622378,
+      "loss": 3.5724,
+      "step": 16900
+    },
+    {
+      "epoch": 4.935989283010076,
+      "grad_norm": 0.3055601119995117,
+      "learning_rate": 0.0005410874125874126,
+      "loss": 3.56,
+      "step": 16950
+    },
+    {
+      "epoch": 4.950550410623799,
+      "grad_norm": 0.30220872163772583,
+      "learning_rate": 0.0005409125874125874,
+      "loss": 3.585,
+      "step": 17000
+    },
+    {
+      "epoch": 4.950550410623799,
+      "eval_accuracy": 0.360180674917524,
+      "eval_loss": 3.6294431686401367,
+      "eval_runtime": 181.1546,
+      "eval_samples_per_second": 91.888,
+      "eval_steps_per_second": 5.746,
+      "step": 17000
+    },
+    {
+      "epoch": 4.9651115382375215,
+      "grad_norm": 0.34501272439956665,
+      "learning_rate": 0.0005407377622377622,
+      "loss": 3.5698,
+      "step": 17050
+    },
+    {
+      "epoch": 4.979672665851243,
+      "grad_norm": 0.31808799505233765,
+      "learning_rate": 0.000540562937062937,
+      "loss": 3.5684,
+      "step": 17100
+    },
+    {
+      "epoch": 4.994233793464966,
+      "grad_norm": 0.30529871582984924,
+      "learning_rate": 0.0005403881118881118,
+      "loss": 3.5667,
+      "step": 17150
+    },
+    {
+      "epoch": 5.008736676568233,
+      "grad_norm": 0.32560649514198303,
+      "learning_rate": 0.0005402132867132867,
+      "loss": 3.5124,
+      "step": 17200
+    },
+    {
+      "epoch": 5.023297804181956,
+      "grad_norm": 0.33668822050094604,
+      "learning_rate": 0.0005400384615384615,
+      "loss": 3.448,
+      "step": 17250
+    },
+    {
+      "epoch": 5.037858931795678,
+      "grad_norm": 0.3430330753326416,
+      "learning_rate": 0.0005398636363636363,
+      "loss": 3.462,
+      "step": 17300
+    },
+    {
+      "epoch": 5.052420059409401,
+      "grad_norm": 0.32133740186691284,
+      "learning_rate": 0.0005396888111888111,
+      "loss": 3.4736,
+      "step": 17350
+    },
+    {
+      "epoch": 5.066981187023123,
+      "grad_norm": 0.30943241715431213,
+      "learning_rate": 0.000539513986013986,
+      "loss": 3.4657,
+      "step": 17400
+    },
+    {
+      "epoch": 5.081542314636845,
+      "grad_norm": 0.3143688142299652,
+      "learning_rate": 0.0005393391608391608,
+      "loss": 3.4726,
+      "step": 17450
+    },
+    {
+      "epoch": 5.096103442250568,
+      "grad_norm": 0.3386751711368561,
+      "learning_rate": 0.0005391643356643356,
+      "loss": 3.4804,
+      "step": 17500
+    },
+    {
+      "epoch": 5.110664569864291,
+      "grad_norm": 0.31570807099342346,
+      "learning_rate": 0.0005389895104895105,
+      "loss": 3.4793,
+      "step": 17550
+    },
+    {
+      "epoch": 5.125225697478013,
+      "grad_norm": 0.32351601123809814,
+      "learning_rate": 0.0005388146853146853,
+      "loss": 3.4856,
+      "step": 17600
+    },
+    {
+      "epoch": 5.139786825091735,
+      "grad_norm": 0.30925703048706055,
+      "learning_rate": 0.0005386398601398601,
+      "loss": 3.486,
+      "step": 17650
+    },
+    {
+      "epoch": 5.154347952705457,
+      "grad_norm": 0.31706124544143677,
+      "learning_rate": 0.0005384650349650349,
+      "loss": 3.4876,
+      "step": 17700
+    },
+    {
+      "epoch": 5.16890908031918,
+      "grad_norm": 0.3108503818511963,
+      "learning_rate": 0.0005382902097902098,
+      "loss": 3.4841,
+      "step": 17750
+    },
+    {
+      "epoch": 5.183470207932903,
+      "grad_norm": 0.3236485719680786,
+      "learning_rate": 0.0005381153846153845,
+      "loss": 3.4851,
+      "step": 17800
+    },
+    {
+      "epoch": 5.1980313355466246,
+      "grad_norm": 0.30849018692970276,
+      "learning_rate": 0.0005379405594405594,
+      "loss": 3.4838,
+      "step": 17850
+    },
+    {
+      "epoch": 5.212592463160347,
+      "grad_norm": 0.3113718032836914,
+      "learning_rate": 0.0005377657342657342,
+      "loss": 3.476,
+      "step": 17900
+    },
+    {
+      "epoch": 5.227153590774069,
+      "grad_norm": 0.35100990533828735,
+      "learning_rate": 0.000537590909090909,
+      "loss": 3.4976,
+      "step": 17950
+    },
+    {
+      "epoch": 5.241714718387792,
+      "grad_norm": 0.3076687455177307,
+      "learning_rate": 0.0005374160839160838,
+      "loss": 3.5044,
+      "step": 18000
+    },
+    {
+      "epoch": 5.241714718387792,
+      "eval_accuracy": 0.3602995307135839,
+      "eval_loss": 3.6342709064483643,
+      "eval_runtime": 181.2085,
+      "eval_samples_per_second": 91.861,
+      "eval_steps_per_second": 5.745,
+      "step": 18000
+    },
+    {
+      "epoch": 5.256275846001515,
+      "grad_norm": 0.3252745270729065,
+      "learning_rate": 0.0005372412587412587,
+      "loss": 3.5038,
+      "step": 18050
+    },
+    {
+      "epoch": 5.2708369736152365,
+      "grad_norm": 0.31030163168907166,
+      "learning_rate": 0.0005370664335664335,
+      "loss": 3.49,
+      "step": 18100
+    },
+    {
+      "epoch": 5.285398101228959,
+      "grad_norm": 0.36103177070617676,
+      "learning_rate": 0.0005368916083916083,
+      "loss": 3.492,
+      "step": 18150
+    },
+    {
+      "epoch": 5.299959228842681,
+      "grad_norm": 0.32504352927207947,
+      "learning_rate": 0.0005367167832167832,
+      "loss": 3.4948,
+      "step": 18200
+    },
+    {
+      "epoch": 5.314520356456404,
+      "grad_norm": 0.3331771790981293,
+      "learning_rate": 0.000536541958041958,
+      "loss": 3.5126,
+      "step": 18250
+    },
+    {
+      "epoch": 5.329081484070127,
+      "grad_norm": 0.31116366386413574,
+      "learning_rate": 0.0005363671328671328,
+      "loss": 3.5076,
+      "step": 18300
+    },
+    {
+      "epoch": 5.3436426116838485,
+      "grad_norm": 0.3158913850784302,
+      "learning_rate": 0.0005361923076923076,
+      "loss": 3.5178,
+      "step": 18350
+    },
+    {
+      "epoch": 5.358203739297571,
+      "grad_norm": 0.32401207089424133,
+      "learning_rate": 0.0005360174825174825,
+      "loss": 3.5149,
+      "step": 18400
+    },
+    {
+      "epoch": 5.372764866911294,
+      "grad_norm": 0.340380996465683,
+      "learning_rate": 0.0005358426573426573,
+      "loss": 3.5078,
+      "step": 18450
+    },
+    {
+      "epoch": 5.387325994525016,
+      "grad_norm": 0.31110548973083496,
+      "learning_rate": 0.0005356678321678321,
+      "loss": 3.5107,
+      "step": 18500
+    },
+    {
+      "epoch": 5.401887122138739,
+      "grad_norm": 0.32105040550231934,
+      "learning_rate": 0.0005354930069930069,
+      "loss": 3.4964,
+      "step": 18550
+    },
+    {
+      "epoch": 5.41644824975246,
+      "grad_norm": 0.3346503674983978,
+      "learning_rate": 0.0005353181818181817,
+      "loss": 3.5029,
+      "step": 18600
+    },
+    {
+      "epoch": 5.431009377366183,
+      "grad_norm": 0.3195594549179077,
+      "learning_rate": 0.0005351433566433565,
+      "loss": 3.5149,
+      "step": 18650
+    },
+    {
+      "epoch": 5.445570504979906,
+      "grad_norm": 0.3379792273044586,
+      "learning_rate": 0.0005349685314685314,
+      "loss": 3.5073,
+      "step": 18700
+    },
+    {
+      "epoch": 5.460131632593628,
+      "grad_norm": 0.3183291256427765,
+      "learning_rate": 0.0005347937062937062,
+      "loss": 3.5091,
+      "step": 18750
+    },
+    {
+      "epoch": 5.4746927602073505,
+      "grad_norm": 0.3273472785949707,
+      "learning_rate": 0.000534618881118881,
+      "loss": 3.5196,
+      "step": 18800
+    },
+    {
+      "epoch": 5.489253887821073,
+      "grad_norm": 0.3325451910495758,
+      "learning_rate": 0.0005344440559440559,
+      "loss": 3.5159,
+      "step": 18850
+    },
+    {
+      "epoch": 5.503815015434795,
+      "grad_norm": 0.33070364594459534,
+      "learning_rate": 0.0005342692307692307,
+      "loss": 3.518,
+      "step": 18900
+    },
+    {
+      "epoch": 5.518376143048518,
+      "grad_norm": 0.33015328645706177,
+      "learning_rate": 0.0005340944055944055,
+      "loss": 3.5238,
+      "step": 18950
+    },
+    {
+      "epoch": 5.53293727066224,
+      "grad_norm": 0.307849645614624,
+      "learning_rate": 0.0005339195804195803,
+      "loss": 3.5108,
+      "step": 19000
+    },
+    {
+      "epoch": 5.53293727066224,
+      "eval_accuracy": 0.36167172146690857,
+      "eval_loss": 3.62229061126709,
+      "eval_runtime": 181.6251,
+      "eval_samples_per_second": 91.65,
+      "eval_steps_per_second": 5.732,
+      "step": 19000
+    },
+    {
+      "epoch": 5.5474983982759625,
+      "grad_norm": 0.2991660237312317,
+      "learning_rate": 0.0005337447552447552,
+      "loss": 3.5153,
+      "step": 19050
+    },
+    {
+      "epoch": 5.562059525889685,
+      "grad_norm": 0.33114874362945557,
+      "learning_rate": 0.00053356993006993,
+      "loss": 3.5165,
+      "step": 19100
+    },
+    {
+      "epoch": 5.576620653503407,
+      "grad_norm": 0.3147285282611847,
+      "learning_rate": 0.0005333951048951048,
+      "loss": 3.5263,
+      "step": 19150
+    },
+    {
+      "epoch": 5.59118178111713,
+      "grad_norm": 0.32433706521987915,
+      "learning_rate": 0.0005332202797202796,
+      "loss": 3.5213,
+      "step": 19200
+    },
+    {
+      "epoch": 5.605742908730852,
+      "grad_norm": 0.386737197637558,
+      "learning_rate": 0.0005330454545454546,
+      "loss": 3.5211,
+      "step": 19250
+    },
+    {
+      "epoch": 5.620304036344574,
+      "grad_norm": 0.32082608342170715,
+      "learning_rate": 0.0005328706293706292,
+      "loss": 3.53,
+      "step": 19300
+    },
+    {
+      "epoch": 5.634865163958297,
+      "grad_norm": 0.2935941219329834,
+      "learning_rate": 0.0005326958041958042,
+      "loss": 3.5197,
+      "step": 19350
+    },
+    {
+      "epoch": 5.649426291572019,
+      "grad_norm": 0.31667831540107727,
+      "learning_rate": 0.000532520979020979,
+      "loss": 3.5095,
+      "step": 19400
+    },
+    {
+      "epoch": 5.663987419185742,
+      "grad_norm": 0.306999534368515,
+      "learning_rate": 0.0005323461538461538,
+      "loss": 3.5187,
+      "step": 19450
+    },
+    {
+      "epoch": 5.6785485467994645,
+      "grad_norm": 0.30406883358955383,
+      "learning_rate": 0.0005321713286713287,
+      "loss": 3.5199,
+      "step": 19500
+    },
+    {
+      "epoch": 5.693109674413186,
+      "grad_norm": 0.3210119903087616,
+      "learning_rate": 0.0005319965034965035,
+      "loss": 3.527,
+      "step": 19550
+    },
+    {
+      "epoch": 5.707670802026909,
+      "grad_norm": 0.33628612756729126,
+      "learning_rate": 0.0005318216783216783,
+      "loss": 3.5168,
+      "step": 19600
+    },
+    {
+      "epoch": 5.722231929640631,
+      "grad_norm": 0.3189808428287506,
+      "learning_rate": 0.0005316468531468531,
+      "loss": 3.5246,
+      "step": 19650
+    },
+    {
+      "epoch": 5.736793057254354,
+      "grad_norm": 0.3225274682044983,
+      "learning_rate": 0.000531472027972028,
+      "loss": 3.5168,
+      "step": 19700
+    },
+    {
+      "epoch": 5.7513541848680765,
+      "grad_norm": 0.32404860854148865,
+      "learning_rate": 0.0005312972027972028,
+      "loss": 3.5305,
+      "step": 19750
+    },
+    {
+      "epoch": 5.765915312481798,
+      "grad_norm": 0.33228975534439087,
+      "learning_rate": 0.0005311223776223776,
+      "loss": 3.5191,
+      "step": 19800
+    },
+    {
+      "epoch": 5.780476440095521,
+      "grad_norm": 0.3346024453639984,
+      "learning_rate": 0.0005309475524475524,
+      "loss": 3.5233,
+      "step": 19850
+    },
+    {
+      "epoch": 5.795037567709244,
+      "grad_norm": 0.30757418274879456,
+      "learning_rate": 0.0005307727272727273,
+      "loss": 3.5203,
+      "step": 19900
+    },
+    {
+      "epoch": 5.809598695322966,
+      "grad_norm": 0.33240988850593567,
+      "learning_rate": 0.0005305979020979021,
+      "loss": 3.5232,
+      "step": 19950
+    },
+    {
+      "epoch": 5.824159822936688,
+      "grad_norm": 0.3051111698150635,
+      "learning_rate": 0.0005304230769230769,
+      "loss": 3.5355,
+      "step": 20000
+    },
+    {
+      "epoch": 5.824159822936688,
+      "eval_accuracy": 0.3623371258246723,
+      "eval_loss": 3.6103367805480957,
+      "eval_runtime": 181.4963,
+      "eval_samples_per_second": 91.715,
+      "eval_steps_per_second": 5.736,
+      "step": 20000
+    },
+    {
+      "epoch": 5.83872095055041,
+      "grad_norm": 0.3242921233177185,
+      "learning_rate": 0.0005302482517482517,
+      "loss": 3.5295,
+      "step": 20050
+    },
+    {
+      "epoch": 5.853282078164133,
+      "grad_norm": 0.3176933526992798,
+      "learning_rate": 0.0005300734265734265,
+      "loss": 3.5191,
+      "step": 20100
+    },
+    {
+      "epoch": 5.867843205777856,
+      "grad_norm": 0.34109026193618774,
+      "learning_rate": 0.0005298986013986013,
+      "loss": 3.5357,
+      "step": 20150
+    },
+    {
+      "epoch": 5.882404333391578,
+      "grad_norm": 0.2950345575809479,
+      "learning_rate": 0.0005297237762237762,
+      "loss": 3.5309,
+      "step": 20200
+    },
+    {
+      "epoch": 5.8969654610053,
+      "grad_norm": 0.30492424964904785,
+      "learning_rate": 0.000529548951048951,
+      "loss": 3.5426,
+      "step": 20250
+    },
+    {
+      "epoch": 5.911526588619022,
+      "grad_norm": 0.3427123725414276,
+      "learning_rate": 0.0005293741258741258,
+      "loss": 3.5293,
+      "step": 20300
+    },
+    {
+      "epoch": 5.926087716232745,
+      "grad_norm": 0.30794087052345276,
+      "learning_rate": 0.0005291993006993007,
+      "loss": 3.5297,
+      "step": 20350
+    },
+    {
+      "epoch": 5.940648843846468,
+      "grad_norm": 0.30005186796188354,
+      "learning_rate": 0.0005290244755244755,
+      "loss": 3.5312,
+      "step": 20400
+    },
+    {
+      "epoch": 5.95520997146019,
+      "grad_norm": 0.3218655586242676,
+      "learning_rate": 0.0005288496503496503,
+      "loss": 3.5224,
+      "step": 20450
+    },
+    {
+      "epoch": 5.969771099073912,
+      "grad_norm": 0.3385027348995209,
+      "learning_rate": 0.0005286748251748251,
+      "loss": 3.5319,
+      "step": 20500
+    },
+    {
+      "epoch": 5.984332226687634,
+      "grad_norm": 0.33277666568756104,
+      "learning_rate": 0.0005285,
+      "loss": 3.5229,
+      "step": 20550
+    },
+    {
+      "epoch": 5.998893354301357,
+      "grad_norm": 0.2984437346458435,
+      "learning_rate": 0.0005283251748251748,
+      "loss": 3.5246,
+      "step": 20600
+    },
+    {
+      "epoch": 6.013396237404625,
+      "grad_norm": 0.3273489773273468,
+      "learning_rate": 0.0005281503496503496,
+      "loss": 3.4096,
+      "step": 20650
+    },
+    {
+      "epoch": 6.027957365018347,
+      "grad_norm": 0.32948926091194153,
+      "learning_rate": 0.0005279755244755244,
+      "loss": 3.4094,
+      "step": 20700
+    },
+    {
+      "epoch": 6.04251849263207,
+      "grad_norm": 0.3348131775856018,
+      "learning_rate": 0.0005278006993006993,
+      "loss": 3.4196,
+      "step": 20750
+    },
+    {
+      "epoch": 6.0570796202457915,
+      "grad_norm": 0.3126414716243744,
+      "learning_rate": 0.000527625874125874,
+      "loss": 3.4201,
+      "step": 20800
+    },
+    {
+      "epoch": 6.071640747859514,
+      "grad_norm": 0.33334097266197205,
+      "learning_rate": 0.0005274510489510489,
+      "loss": 3.4269,
+      "step": 20850
+    },
+    {
+      "epoch": 6.086201875473237,
+      "grad_norm": 0.32628950476646423,
+      "learning_rate": 0.0005272762237762238,
+      "loss": 3.432,
+      "step": 20900
+    },
+    {
+      "epoch": 6.100763003086959,
+      "grad_norm": 0.3498513996601105,
+      "learning_rate": 0.0005271013986013985,
+      "loss": 3.425,
+      "step": 20950
+    },
+    {
+      "epoch": 6.115324130700682,
+      "grad_norm": 0.29833361506462097,
+      "learning_rate": 0.0005269265734265734,
+      "loss": 3.4252,
+      "step": 21000
+    },
+    {
+      "epoch": 6.115324130700682,
+      "eval_accuracy": 0.36289260914453686,
+      "eval_loss": 3.613590717315674,
+      "eval_runtime": 181.366,
+      "eval_samples_per_second": 91.781,
+      "eval_steps_per_second": 5.74,
+      "step": 21000
+    },
+    {
+      "epoch": 6.1298852583144035,
+      "grad_norm": 0.3302992880344391,
+      "learning_rate": 0.0005267517482517482,
+      "loss": 3.4361,
+      "step": 21050
+    },
+    {
+      "epoch": 6.144446385928126,
+      "grad_norm": 0.3029468059539795,
+      "learning_rate": 0.000526576923076923,
+      "loss": 3.4468,
+      "step": 21100
+    },
+    {
+      "epoch": 6.159007513541849,
+      "grad_norm": 0.3242226243019104,
+      "learning_rate": 0.0005264020979020978,
+      "loss": 3.4352,
+      "step": 21150
+    },
+    {
+      "epoch": 6.173568641155571,
+      "grad_norm": 0.31961268186569214,
+      "learning_rate": 0.0005262272727272727,
+      "loss": 3.4358,
+      "step": 21200
+    },
+    {
+      "epoch": 6.1881297687692935,
+      "grad_norm": 0.33194205164909363,
+      "learning_rate": 0.0005260524475524475,
+      "loss": 3.4442,
+      "step": 21250
+    },
+    {
+      "epoch": 6.202690896383016,
+      "grad_norm": 0.32972991466522217,
+      "learning_rate": 0.0005258776223776223,
+      "loss": 3.4521,
+      "step": 21300
+    },
+    {
+      "epoch": 6.217252023996738,
+      "grad_norm": 0.31378456950187683,
+      "learning_rate": 0.0005257027972027971,
+      "loss": 3.4469,
+      "step": 21350
+    },
+    {
+      "epoch": 6.231813151610461,
+      "grad_norm": 0.3384927809238434,
+      "learning_rate": 0.000525527972027972,
+      "loss": 3.4621,
+      "step": 21400
+    },
+    {
+      "epoch": 6.246374279224183,
+      "grad_norm": 0.3361877202987671,
+      "learning_rate": 0.0005253531468531468,
+      "loss": 3.4574,
+      "step": 21450
+    },
+    {
+      "epoch": 6.2609354068379055,
+      "grad_norm": 0.3300999402999878,
+      "learning_rate": 0.0005251783216783216,
+      "loss": 3.4494,
+      "step": 21500
+    },
+    {
+      "epoch": 6.275496534451628,
+      "grad_norm": 0.31691470742225647,
+      "learning_rate": 0.0005250034965034965,
+      "loss": 3.4443,
+      "step": 21550
+    },
+    {
+      "epoch": 6.29005766206535,
+      "grad_norm": 0.30609896779060364,
+      "learning_rate": 0.0005248286713286712,
+      "loss": 3.465,
+      "step": 21600
+    },
+    {
+      "epoch": 6.304618789679073,
+      "grad_norm": 0.3142344653606415,
+      "learning_rate": 0.0005246538461538461,
+      "loss": 3.4489,
+      "step": 21650
+    },
+    {
+      "epoch": 6.319179917292795,
+      "grad_norm": 0.3219060003757477,
+      "learning_rate": 0.0005244790209790209,
+      "loss": 3.4663,
+      "step": 21700
+    },
+    {
+      "epoch": 6.3337410449065175,
+      "grad_norm": 0.3099410831928253,
+      "learning_rate": 0.0005243041958041957,
+      "loss": 3.472,
+      "step": 21750
+    },
+    {
+      "epoch": 6.34830217252024,
+      "grad_norm": 0.31502556800842285,
+      "learning_rate": 0.0005241293706293705,
+      "loss": 3.4804,
+      "step": 21800
+    },
+    {
+      "epoch": 6.362863300133962,
+      "grad_norm": 0.30814129114151,
+      "learning_rate": 0.0005239545454545454,
+      "loss": 3.4601,
+      "step": 21850
+    },
+    {
+      "epoch": 6.377424427747685,
+      "grad_norm": 0.3281274735927582,
+      "learning_rate": 0.0005237797202797202,
+      "loss": 3.4743,
+      "step": 21900
+    },
+    {
+      "epoch": 6.391985555361408,
+      "grad_norm": 0.33183300495147705,
+      "learning_rate": 0.000523604895104895,
+      "loss": 3.4548,
+      "step": 21950
+    },
+    {
+      "epoch": 6.406546682975129,
+      "grad_norm": 0.33608192205429077,
+      "learning_rate": 0.0005234300699300698,
+      "loss": 3.4717,
+      "step": 22000
+    },
+    {
+      "epoch": 6.406546682975129,
+      "eval_accuracy": 0.3633190087214996,
+      "eval_loss": 3.6065382957458496,
+      "eval_runtime": 181.0468,
+      "eval_samples_per_second": 91.943,
+      "eval_steps_per_second": 5.75,
+      "step": 22000
+    },
+    {
+      "epoch": 6.421107810588852,
+      "grad_norm": 0.3260321319103241,
+      "learning_rate": 0.0005232552447552447,
+      "loss": 3.4682,
+      "step": 22050
+    },
+    {
+      "epoch": 6.435668938202574,
+      "grad_norm": 0.33026236295700073,
+      "learning_rate": 0.0005230804195804195,
+      "loss": 3.4802,
+      "step": 22100
+    },
+    {
+      "epoch": 6.450230065816297,
+      "grad_norm": 0.30417993664741516,
+      "learning_rate": 0.0005229055944055943,
+      "loss": 3.476,
+      "step": 22150
+    },
+    {
+      "epoch": 6.4647911934300195,
+      "grad_norm": 0.3218960464000702,
+      "learning_rate": 0.0005227307692307691,
+      "loss": 3.4754,
+      "step": 22200
+    },
+    {
+      "epoch": 6.479352321043741,
+      "grad_norm": 0.31791952252388,
+      "learning_rate": 0.0005225559440559441,
+      "loss": 3.4712,
+      "step": 22250
+    },
+    {
+      "epoch": 6.493913448657464,
+      "grad_norm": 0.31013575196266174,
+      "learning_rate": 0.0005223811188811189,
+      "loss": 3.4725,
+      "step": 22300
+    },
+    {
+      "epoch": 6.508474576271187,
+      "grad_norm": 0.3217204213142395,
+      "learning_rate": 0.0005222062937062937,
+      "loss": 3.4725,
+      "step": 22350
+    },
+    {
+      "epoch": 6.523035703884909,
+      "grad_norm": 0.31372180581092834,
+      "learning_rate": 0.0005220314685314686,
+      "loss": 3.4723,
+      "step": 22400
+    },
+    {
+      "epoch": 6.5375968314986315,
+      "grad_norm": 0.3291754126548767,
+      "learning_rate": 0.0005218566433566433,
+      "loss": 3.4768,
+      "step": 22450
+    },
+    {
+      "epoch": 6.552157959112353,
+      "grad_norm": 0.3256649672985077,
+      "learning_rate": 0.0005216818181818182,
+      "loss": 3.4629,
+      "step": 22500
+    },
+    {
+      "epoch": 6.566719086726076,
+      "grad_norm": 0.323404461145401,
+      "learning_rate": 0.000521506993006993,
+      "loss": 3.4747,
+      "step": 22550
+    },
+    {
+      "epoch": 6.581280214339799,
+      "grad_norm": 0.33667072653770447,
+      "learning_rate": 0.0005213321678321678,
+      "loss": 3.4652,
+      "step": 22600
+    },
+    {
+      "epoch": 6.595841341953521,
+      "grad_norm": 0.3154466450214386,
+      "learning_rate": 0.0005211573426573426,
+      "loss": 3.4836,
+      "step": 22650
+    },
+    {
+      "epoch": 6.610402469567243,
+      "grad_norm": 0.34323811531066895,
+      "learning_rate": 0.0005209825174825175,
+      "loss": 3.4798,
+      "step": 22700
+    },
+    {
+      "epoch": 6.624963597180965,
+      "grad_norm": 0.3353569209575653,
+      "learning_rate": 0.0005208076923076923,
+      "loss": 3.4756,
+      "step": 22750
+    },
+    {
+      "epoch": 6.639524724794688,
+      "grad_norm": 0.3173069655895233,
+      "learning_rate": 0.0005206328671328671,
+      "loss": 3.4838,
+      "step": 22800
+    },
+    {
+      "epoch": 6.654085852408411,
+      "grad_norm": 0.3082946836948395,
+      "learning_rate": 0.0005204580419580419,
+      "loss": 3.4859,
+      "step": 22850
+    },
+    {
+      "epoch": 6.668646980022133,
+      "grad_norm": 0.32030490040779114,
+      "learning_rate": 0.0005202832167832168,
+      "loss": 3.4902,
+      "step": 22900
+    },
+    {
+      "epoch": 6.683208107635855,
+      "grad_norm": 0.3350687623023987,
+      "learning_rate": 0.0005201083916083916,
+      "loss": 3.4842,
+      "step": 22950
+    },
+    {
+      "epoch": 6.697769235249577,
+      "grad_norm": 0.324341356754303,
+      "learning_rate": 0.0005199335664335664,
+      "loss": 3.4676,
+      "step": 23000
+    },
+    {
+      "epoch": 6.697769235249577,
+      "eval_accuracy": 0.3642195382940208,
+      "eval_loss": 3.5976805686950684,
+      "eval_runtime": 181.2142,
+      "eval_samples_per_second": 91.858,
+      "eval_steps_per_second": 5.745,
+      "step": 23000
+    },
+    {
+      "epoch": 6.7123303628633,
+      "grad_norm": 0.31654438376426697,
+      "learning_rate": 0.0005197587412587413,
+      "loss": 3.487,
+      "step": 23050
+    },
+    {
+      "epoch": 6.726891490477023,
+      "grad_norm": 0.324179083108902,
+      "learning_rate": 0.0005195839160839161,
+      "loss": 3.4772,
+      "step": 23100
+    },
+    {
+      "epoch": 6.741452618090745,
+      "grad_norm": 0.3194001019001007,
+      "learning_rate": 0.0005194090909090909,
+      "loss": 3.4873,
+      "step": 23150
+    },
+    {
+      "epoch": 6.756013745704467,
+      "grad_norm": 0.3092760741710663,
+      "learning_rate": 0.0005192342657342657,
+      "loss": 3.4903,
+      "step": 23200
+    },
+    {
+      "epoch": 6.77057487331819,
+      "grad_norm": 0.3114040195941925,
+      "learning_rate": 0.0005190594405594405,
+      "loss": 3.4924,
+      "step": 23250
+    },
+    {
+      "epoch": 6.785136000931912,
+      "grad_norm": 0.3235475420951843,
+      "learning_rate": 0.0005188846153846153,
+      "loss": 3.4826,
+      "step": 23300
+    },
+    {
+      "epoch": 6.799697128545635,
+      "grad_norm": 0.3458871841430664,
+      "learning_rate": 0.0005187097902097902,
+      "loss": 3.4872,
+      "step": 23350
+    },
+    {
+      "epoch": 6.814258256159357,
+      "grad_norm": 0.34217914938926697,
+      "learning_rate": 0.000518534965034965,
+      "loss": 3.4947,
+      "step": 23400
+    },
+    {
+      "epoch": 6.828819383773079,
+      "grad_norm": 0.3139799237251282,
+      "learning_rate": 0.0005183601398601398,
+      "loss": 3.4925,
+      "step": 23450
+    },
+    {
+      "epoch": 6.843380511386802,
+      "grad_norm": 0.33358052372932434,
+      "learning_rate": 0.0005181853146853146,
+      "loss": 3.4904,
+      "step": 23500
+    },
+    {
+      "epoch": 6.857941639000524,
+      "grad_norm": 0.32350531220436096,
+      "learning_rate": 0.0005180104895104895,
+      "loss": 3.4951,
+      "step": 23550
+    },
+    {
+      "epoch": 6.872502766614247,
+      "grad_norm": 0.32974010705947876,
+      "learning_rate": 0.0005178356643356643,
+      "loss": 3.493,
+      "step": 23600
+    },
+    {
+      "epoch": 6.887063894227969,
+      "grad_norm": 0.33600473403930664,
+      "learning_rate": 0.0005176608391608391,
+      "loss": 3.4893,
+      "step": 23650
+    },
+    {
+      "epoch": 6.901625021841691,
+      "grad_norm": 0.32952529191970825,
+      "learning_rate": 0.000517486013986014,
+      "loss": 3.4971,
+      "step": 23700
+    },
+    {
+      "epoch": 6.916186149455414,
+      "grad_norm": 0.325702041387558,
+      "learning_rate": 0.0005173111888111888,
+      "loss": 3.4979,
+      "step": 23750
+    },
+    {
+      "epoch": 6.930747277069136,
+      "grad_norm": 0.32848599553108215,
+      "learning_rate": 0.0005171363636363636,
+      "loss": 3.4868,
+      "step": 23800
+    },
+    {
+      "epoch": 6.945308404682859,
+      "grad_norm": 0.3094465732574463,
+      "learning_rate": 0.0005169615384615384,
+      "loss": 3.4919,
+      "step": 23850
+    },
+    {
+      "epoch": 6.959869532296581,
+      "grad_norm": 0.3336765468120575,
+      "learning_rate": 0.0005167867132867133,
+      "loss": 3.4885,
+      "step": 23900
+    },
+    {
+      "epoch": 6.974430659910303,
+      "grad_norm": 0.32281845808029175,
+      "learning_rate": 0.000516611888111888,
+      "loss": 3.4788,
+      "step": 23950
+    },
+    {
+      "epoch": 6.988991787524026,
+      "grad_norm": 0.37186485528945923,
+      "learning_rate": 0.0005164370629370629,
+      "loss": 3.4867,
+      "step": 24000
+    },
+    {
+      "epoch": 6.988991787524026,
+      "eval_accuracy": 0.3649564207170708,
+      "eval_loss": 3.5883593559265137,
+      "eval_runtime": 181.1707,
+      "eval_samples_per_second": 91.88,
+      "eval_steps_per_second": 5.746,
+      "step": 24000
+    },
+    {
+      "epoch": 7.003494670627293,
+      "grad_norm": 0.31783783435821533,
+      "learning_rate": 0.0005162622377622377,
+      "loss": 3.459,
+      "step": 24050
+    },
+    {
+      "epoch": 7.018055798241016,
+      "grad_norm": 0.34084847569465637,
+      "learning_rate": 0.0005160874125874125,
+      "loss": 3.379,
+      "step": 24100
+    },
+    {
+      "epoch": 7.032616925854738,
+      "grad_norm": 0.3169894218444824,
+      "learning_rate": 0.0005159125874125873,
+      "loss": 3.3842,
+      "step": 24150
+    },
+    {
+      "epoch": 7.0471780534684605,
+      "grad_norm": 0.32522594928741455,
+      "learning_rate": 0.0005157377622377622,
+      "loss": 3.3877,
+      "step": 24200
+    },
+    {
+      "epoch": 7.061739181082183,
+      "grad_norm": 0.3461341857910156,
+      "learning_rate": 0.000515562937062937,
+      "loss": 3.3905,
+      "step": 24250
+    },
+    {
+      "epoch": 7.076300308695905,
+      "grad_norm": 0.328916996717453,
+      "learning_rate": 0.0005153881118881118,
+      "loss": 3.3949,
+      "step": 24300
+    },
+    {
+      "epoch": 7.090861436309628,
+      "grad_norm": 0.34981271624565125,
+      "learning_rate": 0.0005152132867132867,
+      "loss": 3.399,
+      "step": 24350
+    },
+    {
+      "epoch": 7.105422563923351,
+      "grad_norm": 0.3182472884654999,
+      "learning_rate": 0.0005150384615384615,
+      "loss": 3.4032,
+      "step": 24400
+    },
+    {
+      "epoch": 7.1199836915370724,
+      "grad_norm": 0.3407321870326996,
+      "learning_rate": 0.0005148636363636363,
+      "loss": 3.4098,
+      "step": 24450
+    },
+    {
+      "epoch": 7.134544819150795,
+      "grad_norm": 0.31166911125183105,
+      "learning_rate": 0.0005146888111888111,
+      "loss": 3.3845,
+      "step": 24500
+    },
+    {
+      "epoch": 7.149105946764517,
+      "grad_norm": 0.3245963454246521,
+      "learning_rate": 0.000514513986013986,
+      "loss": 3.4064,
+      "step": 24550
+    },
+    {
+      "epoch": 7.16366707437824,
+      "grad_norm": 0.35094335675239563,
+      "learning_rate": 0.0005143391608391608,
+      "loss": 3.4088,
+      "step": 24600
+    },
+    {
+      "epoch": 7.1782282019919625,
+      "grad_norm": 0.31868040561676025,
+      "learning_rate": 0.0005141643356643356,
+      "loss": 3.4047,
+      "step": 24650
+    },
+    {
+      "epoch": 7.192789329605684,
+      "grad_norm": 0.32330843806266785,
+      "learning_rate": 0.0005139895104895104,
+      "loss": 3.4007,
+      "step": 24700
+    },
+    {
+      "epoch": 7.207350457219407,
+      "grad_norm": 0.34358319640159607,
+      "learning_rate": 0.0005138146853146852,
+      "loss": 3.4181,
+      "step": 24750
+    },
+    {
+      "epoch": 7.22191158483313,
+      "grad_norm": 0.3291915953159332,
+      "learning_rate": 0.00051363986013986,
+      "loss": 3.4165,
+      "step": 24800
+    },
+    {
+      "epoch": 7.236472712446852,
+      "grad_norm": 0.31977158784866333,
+      "learning_rate": 0.0005134650349650349,
+      "loss": 3.4246,
+      "step": 24850
+    },
+    {
+      "epoch": 7.2510338400605745,
+      "grad_norm": 0.33271709084510803,
+      "learning_rate": 0.0005132902097902097,
+      "loss": 3.4101,
+      "step": 24900
+    },
+    {
+      "epoch": 7.265594967674296,
+      "grad_norm": 0.3385375738143921,
+      "learning_rate": 0.0005131153846153845,
+      "loss": 3.4185,
+      "step": 24950
+    },
+    {
+      "epoch": 7.280156095288019,
+      "grad_norm": 0.359745055437088,
+      "learning_rate": 0.0005129405594405594,
+      "loss": 3.4248,
+      "step": 25000
+    },
+    {
+      "epoch": 7.280156095288019,
+      "eval_accuracy": 0.36494066732768204,
+      "eval_loss": 3.5982604026794434,
+      "eval_runtime": 181.1034,
+      "eval_samples_per_second": 91.914,
+      "eval_steps_per_second": 5.748,
+      "step": 25000
+    },
+    {
+      "epoch": 7.294717222901742,
+      "grad_norm": 0.32137957215309143,
+      "learning_rate": 0.0005127657342657342,
+      "loss": 3.4147,
+      "step": 25050
+    },
+    {
+      "epoch": 7.309278350515464,
+      "grad_norm": 0.3147618770599365,
+      "learning_rate": 0.000512590909090909,
+      "loss": 3.4202,
+      "step": 25100
+    },
+    {
+      "epoch": 7.3238394781291865,
+      "grad_norm": 0.3232477605342865,
+      "learning_rate": 0.0005124160839160838,
+      "loss": 3.4304,
+      "step": 25150
+    },
+    {
+      "epoch": 7.338400605742908,
+      "grad_norm": 0.3342421352863312,
+      "learning_rate": 0.0005122412587412588,
+      "loss": 3.4261,
+      "step": 25200
+    },
+    {
+      "epoch": 7.352961733356631,
+      "grad_norm": 0.3140034079551697,
+      "learning_rate": 0.0005120664335664336,
+      "loss": 3.435,
+      "step": 25250
+    },
+    {
+      "epoch": 7.367522860970354,
+      "grad_norm": 0.33676761388778687,
+      "learning_rate": 0.0005118916083916084,
+      "loss": 3.4313,
+      "step": 25300
+    },
+    {
+      "epoch": 7.382083988584076,
+      "grad_norm": 0.33039015531539917,
+      "learning_rate": 0.0005117167832167832,
+      "loss": 3.435,
+      "step": 25350
+    },
+    {
+      "epoch": 7.396645116197798,
+      "grad_norm": 0.34258249402046204,
+      "learning_rate": 0.0005115419580419581,
+      "loss": 3.4223,
+      "step": 25400
+    },
+    {
+      "epoch": 7.411206243811521,
+      "grad_norm": 0.3384631276130676,
+      "learning_rate": 0.0005113671328671328,
+      "loss": 3.4243,
+      "step": 25450
+    },
+    {
+      "epoch": 7.425767371425243,
+      "grad_norm": 0.32956787943840027,
+      "learning_rate": 0.0005111923076923077,
+      "loss": 3.4416,
+      "step": 25500
+    },
+    {
+      "epoch": 7.440328499038966,
+      "grad_norm": 0.3352907598018646,
+      "learning_rate": 0.0005110174825174825,
+      "loss": 3.4284,
+      "step": 25550
+    },
+    {
+      "epoch": 7.454889626652688,
+      "grad_norm": 0.3281717002391815,
+      "learning_rate": 0.0005108426573426573,
+      "loss": 3.4443,
+      "step": 25600
+    },
+    {
+      "epoch": 7.46945075426641,
+      "grad_norm": 0.34179285168647766,
+      "learning_rate": 0.0005106678321678321,
+      "loss": 3.435,
+      "step": 25650
+    },
+    {
+      "epoch": 7.484011881880133,
+      "grad_norm": 0.35076916217803955,
+      "learning_rate": 0.000510493006993007,
+      "loss": 3.4447,
+      "step": 25700
+    },
+    {
+      "epoch": 7.498573009493855,
+      "grad_norm": 0.3248736262321472,
+      "learning_rate": 0.0005103181818181818,
+      "loss": 3.4386,
+      "step": 25750
+    },
+    {
+      "epoch": 7.513134137107578,
+      "grad_norm": 0.3269752860069275,
+      "learning_rate": 0.0005101433566433566,
+      "loss": 3.4369,
+      "step": 25800
+    },
+    {
+      "epoch": 7.5276952647213005,
+      "grad_norm": 0.343089759349823,
+      "learning_rate": 0.0005099685314685315,
+      "loss": 3.447,
+      "step": 25850
+    },
+    {
+      "epoch": 7.542256392335022,
+      "grad_norm": 0.3274691104888916,
+      "learning_rate": 0.0005097937062937063,
+      "loss": 3.4378,
+      "step": 25900
+    },
+    {
+      "epoch": 7.556817519948745,
+      "grad_norm": 0.31769490242004395,
+      "learning_rate": 0.0005096188811188811,
+      "loss": 3.4429,
+      "step": 25950
+    },
+    {
+      "epoch": 7.571378647562467,
+      "grad_norm": 0.30322396755218506,
+      "learning_rate": 0.0005094440559440559,
+      "loss": 3.4455,
+      "step": 26000
+    },
+    {
+      "epoch": 7.571378647562467,
+      "eval_accuracy": 0.3655577534538131,
+      "eval_loss": 3.586430072784424,
+      "eval_runtime": 181.088,
+      "eval_samples_per_second": 91.922,
+      "eval_steps_per_second": 5.749,
+      "step": 26000
+    },
+    {
+      "epoch": 7.58593977517619,
+      "grad_norm": 0.3594802916049957,
+      "learning_rate": 0.0005092692307692308,
+      "loss": 3.4449,
+      "step": 26050
+    },
+    {
+      "epoch": 7.600500902789912,
+      "grad_norm": 0.34348103404045105,
+      "learning_rate": 0.0005090944055944056,
+      "loss": 3.4474,
+      "step": 26100
+    },
+    {
+      "epoch": 7.615062030403634,
+      "grad_norm": 0.32391390204429626,
+      "learning_rate": 0.0005089195804195804,
+      "loss": 3.445,
+      "step": 26150
+    },
+    {
+      "epoch": 7.629623158017357,
+      "grad_norm": 0.3242975175380707,
+      "learning_rate": 0.0005087447552447552,
+      "loss": 3.4404,
+      "step": 26200
+    },
+    {
+      "epoch": 7.644184285631079,
+      "grad_norm": 0.35450035333633423,
+      "learning_rate": 0.00050856993006993,
+      "loss": 3.4475,
+      "step": 26250
+    },
+    {
+      "epoch": 7.658745413244802,
+      "grad_norm": 0.32055407762527466,
+      "learning_rate": 0.0005083951048951048,
+      "loss": 3.4406,
+      "step": 26300
+    },
+    {
+      "epoch": 7.673306540858524,
+      "grad_norm": 0.31129223108291626,
+      "learning_rate": 0.0005082202797202797,
+      "loss": 3.4532,
+      "step": 26350
+    },
+    {
+      "epoch": 7.687867668472246,
+      "grad_norm": 0.33475667238235474,
+      "learning_rate": 0.0005080454545454545,
+      "loss": 3.4582,
+      "step": 26400
+    },
+    {
+      "epoch": 7.702428796085969,
+      "grad_norm": 0.3636000454425812,
+      "learning_rate": 0.0005078706293706293,
+      "loss": 3.442,
+      "step": 26450
+    },
+    {
+      "epoch": 7.716989923699691,
+      "grad_norm": 0.3021737039089203,
+      "learning_rate": 0.0005076958041958042,
+      "loss": 3.4586,
+      "step": 26500
+    },
+    {
+      "epoch": 7.731551051313414,
+      "grad_norm": 0.32579872012138367,
+      "learning_rate": 0.000507520979020979,
+      "loss": 3.4468,
+      "step": 26550
+    },
+    {
+      "epoch": 7.746112178927136,
+      "grad_norm": 0.3306429982185364,
+      "learning_rate": 0.0005073461538461538,
+      "loss": 3.4508,
+      "step": 26600
+    },
+    {
+      "epoch": 7.760673306540858,
+      "grad_norm": 0.33379238843917847,
+      "learning_rate": 0.0005071713286713286,
+      "loss": 3.4554,
+      "step": 26650
+    },
+    {
+      "epoch": 7.775234434154581,
+      "grad_norm": 0.3489669859409332,
+      "learning_rate": 0.0005069965034965035,
+      "loss": 3.4479,
+      "step": 26700
+    },
+    {
+      "epoch": 7.789795561768304,
+      "grad_norm": 0.3256928324699402,
+      "learning_rate": 0.0005068216783216783,
+      "loss": 3.4562,
+      "step": 26750
+    },
+    {
+      "epoch": 7.8043566893820255,
+      "grad_norm": 0.30848199129104614,
+      "learning_rate": 0.0005066468531468531,
+      "loss": 3.4483,
+      "step": 26800
+    },
+    {
+      "epoch": 7.818917816995748,
+      "grad_norm": 0.3393913507461548,
+      "learning_rate": 0.0005064720279720279,
+      "loss": 3.4709,
+      "step": 26850
+    },
+    {
+      "epoch": 7.833478944609471,
+      "grad_norm": 0.31153106689453125,
+      "learning_rate": 0.0005062972027972028,
+      "loss": 3.4658,
+      "step": 26900
+    },
+    {
+      "epoch": 7.848040072223193,
+      "grad_norm": 0.328171044588089,
+      "learning_rate": 0.0005061223776223775,
+      "loss": 3.4517,
+      "step": 26950
+    },
+    {
+      "epoch": 7.862601199836916,
+      "grad_norm": 0.32433098554611206,
+      "learning_rate": 0.0005059475524475524,
+      "loss": 3.4606,
+      "step": 27000
+    },
+    {
+      "epoch": 7.862601199836916,
+      "eval_accuracy": 0.366670836220475,
+      "eval_loss": 3.5772430896759033,
+      "eval_runtime": 181.1731,
+      "eval_samples_per_second": 91.879,
+      "eval_steps_per_second": 5.746,
+      "step": 27000
+    },
+    {
+      "epoch": 7.8771623274506375,
+      "grad_norm": 0.3007674515247345,
+      "learning_rate": 0.0005057727272727272,
+      "loss": 3.4629,
+      "step": 27050
+    },
+    {
+      "epoch": 7.89172345506436,
+      "grad_norm": 0.3290211856365204,
+      "learning_rate": 0.000505597902097902,
+      "loss": 3.4581,
+      "step": 27100
+    },
+    {
+      "epoch": 7.906284582678083,
+      "grad_norm": 0.3350273072719574,
+      "learning_rate": 0.0005054230769230769,
+      "loss": 3.4522,
+      "step": 27150
+    },
+    {
+      "epoch": 7.920845710291805,
+      "grad_norm": 0.3044837713241577,
+      "learning_rate": 0.0005052482517482517,
+      "loss": 3.4548,
+      "step": 27200
+    },
+    {
+      "epoch": 7.935406837905528,
+      "grad_norm": 0.3208082914352417,
+      "learning_rate": 0.0005050734265734265,
+      "loss": 3.4407,
+      "step": 27250
+    },
+    {
+      "epoch": 7.9499679655192494,
+      "grad_norm": 0.33409732580184937,
+      "learning_rate": 0.0005048986013986013,
+      "loss": 3.4498,
+      "step": 27300
+    },
+    {
+      "epoch": 7.964529093132972,
+      "grad_norm": 0.33270469307899475,
+      "learning_rate": 0.0005047237762237762,
+      "loss": 3.4592,
+      "step": 27350
+    },
+    {
+      "epoch": 7.979090220746695,
+      "grad_norm": 0.336601585149765,
+      "learning_rate": 0.000504548951048951,
+      "loss": 3.4603,
+      "step": 27400
+    },
+    {
+      "epoch": 7.993651348360417,
+      "grad_norm": 0.32460638880729675,
+      "learning_rate": 0.0005043741258741258,
+      "loss": 3.4516,
+      "step": 27450
+    },
+    {
+      "epoch": 8.008154231463685,
+      "grad_norm": 0.33659544587135315,
+      "learning_rate": 0.0005041993006993006,
+      "loss": 3.404,
+      "step": 27500
+    },
+    {
+      "epoch": 8.022715359077408,
+      "grad_norm": 0.3334440588951111,
+      "learning_rate": 0.0005040244755244755,
+      "loss": 3.3413,
+      "step": 27550
+    },
+    {
+      "epoch": 8.037276486691129,
+      "grad_norm": 0.32501035928726196,
+      "learning_rate": 0.0005038496503496503,
+      "loss": 3.3455,
+      "step": 27600
+    },
+    {
+      "epoch": 8.051837614304851,
+      "grad_norm": 0.33923956751823425,
+      "learning_rate": 0.0005036748251748251,
+      "loss": 3.3632,
+      "step": 27650
+    },
+    {
+      "epoch": 8.066398741918574,
+      "grad_norm": 0.3274213671684265,
+      "learning_rate": 0.0005034999999999999,
+      "loss": 3.3568,
+      "step": 27700
+    },
+    {
+      "epoch": 8.080959869532297,
+      "grad_norm": 0.3114986717700958,
+      "learning_rate": 0.0005033251748251747,
+      "loss": 3.3671,
+      "step": 27750
+    },
+    {
+      "epoch": 8.09552099714602,
+      "grad_norm": 0.33176517486572266,
+      "learning_rate": 0.0005031503496503496,
+      "loss": 3.3705,
+      "step": 27800
+    },
+    {
+      "epoch": 8.11008212475974,
+      "grad_norm": 0.3320193290710449,
+      "learning_rate": 0.0005029755244755244,
+      "loss": 3.365,
+      "step": 27850
+    },
+    {
+      "epoch": 8.124643252373463,
+      "grad_norm": 0.3689771890640259,
+      "learning_rate": 0.0005028006993006992,
+      "loss": 3.3782,
+      "step": 27900
+    },
+    {
+      "epoch": 8.139204379987186,
+      "grad_norm": 0.32468920946121216,
+      "learning_rate": 0.000502625874125874,
+      "loss": 3.372,
+      "step": 27950
+    },
+    {
+      "epoch": 8.153765507600909,
+      "grad_norm": 0.3300785422325134,
+      "learning_rate": 0.000502451048951049,
+      "loss": 3.3815,
+      "step": 28000
+    },
+    {
+      "epoch": 8.153765507600909,
+      "eval_accuracy": 0.3664210156797952,
+      "eval_loss": 3.589651584625244,
+      "eval_runtime": 181.0562,
+      "eval_samples_per_second": 91.938,
+      "eval_steps_per_second": 5.75,
+      "step": 28000
+    },
+    {
+      "epoch": 8.168326635214632,
+      "grad_norm": 0.3502894639968872,
+      "learning_rate": 0.0005022762237762237,
+      "loss": 3.3722,
+      "step": 28050
+    },
+    {
+      "epoch": 8.182887762828354,
+      "grad_norm": 0.32582199573516846,
+      "learning_rate": 0.0005021013986013985,
+      "loss": 3.3825,
+      "step": 28100
+    },
+    {
+      "epoch": 8.197448890442075,
+      "grad_norm": 0.3195355534553528,
+      "learning_rate": 0.0005019265734265733,
+      "loss": 3.3769,
+      "step": 28150
+    },
+    {
+      "epoch": 8.212010018055798,
+      "grad_norm": 0.3466898500919342,
+      "learning_rate": 0.0005017517482517483,
+      "loss": 3.3755,
+      "step": 28200
+    },
+    {
+      "epoch": 8.22657114566952,
+      "grad_norm": 0.31910231709480286,
+      "learning_rate": 0.0005015769230769231,
+      "loss": 3.3877,
+      "step": 28250
+    },
+    {
+      "epoch": 8.241132273283243,
+      "grad_norm": 0.3181063234806061,
+      "learning_rate": 0.0005014020979020979,
+      "loss": 3.3976,
+      "step": 28300
+    },
+    {
+      "epoch": 8.255693400896966,
+      "grad_norm": 0.3086499869823456,
+      "learning_rate": 0.0005012272727272727,
+      "loss": 3.3798,
+      "step": 28350
+    },
+    {
+      "epoch": 8.270254528510687,
+      "grad_norm": 0.32086050510406494,
+      "learning_rate": 0.0005010524475524476,
+      "loss": 3.3847,
+      "step": 28400
+    },
+    {
+      "epoch": 8.28481565612441,
+      "grad_norm": 0.33809781074523926,
+      "learning_rate": 0.0005008776223776223,
+      "loss": 3.4049,
+      "step": 28450
+    },
+    {
+      "epoch": 8.299376783738133,
+      "grad_norm": 0.365345299243927,
+      "learning_rate": 0.0005007027972027972,
+      "loss": 3.3976,
+      "step": 28500
+    },
+    {
+      "epoch": 8.313937911351855,
+      "grad_norm": 0.31952720880508423,
+      "learning_rate": 0.000500527972027972,
+      "loss": 3.4071,
+      "step": 28550
+    },
+    {
+      "epoch": 8.328499038965578,
+      "grad_norm": 0.3320535123348236,
+      "learning_rate": 0.0005003531468531468,
+      "loss": 3.4125,
+      "step": 28600
+    },
+    {
+      "epoch": 8.3430601665793,
+      "grad_norm": 0.3217061758041382,
+      "learning_rate": 0.0005001783216783217,
+      "loss": 3.3984,
+      "step": 28650
+    },
+    {
+      "epoch": 8.357621294193022,
+      "grad_norm": 0.3326950967311859,
+      "learning_rate": 0.0005000034965034965,
+      "loss": 3.3986,
+      "step": 28700
+    },
+    {
+      "epoch": 8.372182421806745,
+      "grad_norm": 0.33330121636390686,
+      "learning_rate": 0.0004998286713286713,
+      "loss": 3.3976,
+      "step": 28750
+    },
+    {
+      "epoch": 8.386743549420467,
+      "grad_norm": 0.3166747987270355,
+      "learning_rate": 0.0004996538461538461,
+      "loss": 3.4002,
+      "step": 28800
+    },
+    {
+      "epoch": 8.40130467703419,
+      "grad_norm": 0.34696197509765625,
+      "learning_rate": 0.000499479020979021,
+      "loss": 3.4108,
+      "step": 28850
+    },
+    {
+      "epoch": 8.415865804647911,
+      "grad_norm": 0.3111904263496399,
+      "learning_rate": 0.0004993041958041958,
+      "loss": 3.3988,
+      "step": 28900
+    },
+    {
+      "epoch": 8.430426932261634,
+      "grad_norm": 0.3558425307273865,
+      "learning_rate": 0.0004991293706293706,
+      "loss": 3.4024,
+      "step": 28950
+    },
+    {
+      "epoch": 8.444988059875357,
+      "grad_norm": 0.37141433358192444,
+      "learning_rate": 0.0004989545454545454,
+      "loss": 3.4063,
+      "step": 29000
+    },
+    {
+      "epoch": 8.444988059875357,
+      "eval_accuracy": 0.3664435877004119,
+      "eval_loss": 3.581801414489746,
+      "eval_runtime": 181.2088,
+      "eval_samples_per_second": 91.861,
+      "eval_steps_per_second": 5.745,
+      "step": 29000
+    },
+    {
+      "epoch": 8.45954918748908,
+      "grad_norm": 0.31924572587013245,
+      "learning_rate": 0.0004987797202797203,
+      "loss": 3.4258,
+      "step": 29050
+    },
+    {
+      "epoch": 8.474110315102802,
+      "grad_norm": 0.3425278961658478,
+      "learning_rate": 0.0004986048951048951,
+      "loss": 3.4101,
+      "step": 29100
+    },
+    {
+      "epoch": 8.488671442716523,
+      "grad_norm": 0.3478877544403076,
+      "learning_rate": 0.0004984300699300699,
+      "loss": 3.4145,
+      "step": 29150
+    },
+    {
+      "epoch": 8.503232570330246,
+      "grad_norm": 0.35354796051979065,
+      "learning_rate": 0.0004982552447552448,
+      "loss": 3.4072,
+      "step": 29200
+    },
+    {
+      "epoch": 8.517793697943969,
+      "grad_norm": 0.32307496666908264,
+      "learning_rate": 0.0004980804195804195,
+      "loss": 3.4305,
+      "step": 29250
+    },
+    {
+      "epoch": 8.532354825557691,
+      "grad_norm": 0.3157086670398712,
+      "learning_rate": 0.0004979055944055944,
+      "loss": 3.4064,
+      "step": 29300
+    },
+    {
+      "epoch": 8.546915953171414,
+      "grad_norm": 0.3342621326446533,
+      "learning_rate": 0.0004977307692307692,
+      "loss": 3.3926,
+      "step": 29350
+    },
+    {
+      "epoch": 8.561477080785137,
+      "grad_norm": 0.31849896907806396,
+      "learning_rate": 0.000497555944055944,
+      "loss": 3.4163,
+      "step": 29400
+    },
+    {
+      "epoch": 8.576038208398858,
+      "grad_norm": 0.31333833932876587,
+      "learning_rate": 0.0004973811188811188,
+      "loss": 3.4039,
+      "step": 29450
+    },
+    {
+      "epoch": 8.59059933601258,
+      "grad_norm": 0.3372572660446167,
+      "learning_rate": 0.0004972062937062937,
+      "loss": 3.4262,
+      "step": 29500
+    },
+    {
+      "epoch": 8.605160463626303,
+      "grad_norm": 0.3162132799625397,
+      "learning_rate": 0.0004970314685314685,
+      "loss": 3.4172,
+      "step": 29550
+    },
+    {
+      "epoch": 8.619721591240026,
+      "grad_norm": 0.3454541862010956,
+      "learning_rate": 0.0004968566433566433,
+      "loss": 3.4144,
+      "step": 29600
+    },
+    {
+      "epoch": 8.634282718853749,
+      "grad_norm": 0.3138660192489624,
+      "learning_rate": 0.0004966818181818181,
+      "loss": 3.4081,
+      "step": 29650
+    },
+    {
+      "epoch": 8.64884384646747,
+      "grad_norm": 0.3220912516117096,
+      "learning_rate": 0.000496506993006993,
+      "loss": 3.417,
+      "step": 29700
+    },
+    {
+      "epoch": 8.663404974081192,
+      "grad_norm": 0.32789018750190735,
+      "learning_rate": 0.0004963321678321678,
+      "loss": 3.4131,
+      "step": 29750
+    },
+    {
+      "epoch": 8.677966101694915,
+      "grad_norm": 0.3420283794403076,
+      "learning_rate": 0.0004961573426573426,
+      "loss": 3.4202,
+      "step": 29800
+    },
+    {
+      "epoch": 8.692527229308638,
+      "grad_norm": 0.35813525319099426,
+      "learning_rate": 0.0004959825174825175,
+      "loss": 3.4203,
+      "step": 29850
+    },
+    {
+      "epoch": 8.70708835692236,
+      "grad_norm": 0.33372020721435547,
+      "learning_rate": 0.0004958076923076923,
+      "loss": 3.4255,
+      "step": 29900
+    },
+    {
+      "epoch": 8.721649484536082,
+      "grad_norm": 0.3299264907836914,
+      "learning_rate": 0.0004956328671328671,
+      "loss": 3.4189,
+      "step": 29950
+    },
+    {
+      "epoch": 8.736210612149804,
+      "grad_norm": 0.34100714325904846,
+      "learning_rate": 0.0004954580419580419,
+      "loss": 3.4294,
+      "step": 30000
+    },
+    {
+      "epoch": 8.736210612149804,
+      "eval_accuracy": 0.367117809253729,
+      "eval_loss": 3.57382869720459,
+      "eval_runtime": 181.1298,
+      "eval_samples_per_second": 91.901,
+      "eval_steps_per_second": 5.747,
+      "step": 30000
+    },
+    {
+      "epoch": 8.750771739763527,
+      "grad_norm": 0.32569196820259094,
+      "learning_rate": 0.0004952832167832167,
+      "loss": 3.4342,
+      "step": 30050
+    },
+    {
+      "epoch": 8.76533286737725,
+      "grad_norm": 0.3259648084640503,
+      "learning_rate": 0.0004951083916083915,
+      "loss": 3.4277,
+      "step": 30100
+    },
+    {
+      "epoch": 8.779893994990973,
+      "grad_norm": 0.3290475904941559,
+      "learning_rate": 0.0004949335664335664,
+      "loss": 3.422,
+      "step": 30150
+    },
+    {
+      "epoch": 8.794455122604695,
+      "grad_norm": 0.3241061270236969,
+      "learning_rate": 0.0004947587412587412,
+      "loss": 3.44,
+      "step": 30200
+    },
+    {
+      "epoch": 8.809016250218416,
+      "grad_norm": 0.3573022782802582,
+      "learning_rate": 0.000494583916083916,
+      "loss": 3.4274,
+      "step": 30250
+    },
+    {
+      "epoch": 8.82357737783214,
+      "grad_norm": 0.341971218585968,
+      "learning_rate": 0.0004944090909090908,
+      "loss": 3.4207,
+      "step": 30300
+    },
+    {
+      "epoch": 8.838138505445862,
+      "grad_norm": 0.3399566113948822,
+      "learning_rate": 0.0004942342657342657,
+      "loss": 3.4246,
+      "step": 30350
+    },
+    {
+      "epoch": 8.852699633059585,
+      "grad_norm": 0.3449283540248871,
+      "learning_rate": 0.0004940594405594405,
+      "loss": 3.4276,
+      "step": 30400
+    },
+    {
+      "epoch": 8.867260760673307,
+      "grad_norm": 0.3288925886154175,
+      "learning_rate": 0.0004938846153846153,
+      "loss": 3.4198,
+      "step": 30450
+    },
+    {
+      "epoch": 8.881821888287028,
+      "grad_norm": 0.36345532536506653,
+      "learning_rate": 0.0004937097902097901,
+      "loss": 3.4235,
+      "step": 30500
+    },
+    {
+      "epoch": 8.896383015900751,
+      "grad_norm": 0.34723007678985596,
+      "learning_rate": 0.000493534965034965,
+      "loss": 3.4248,
+      "step": 30550
+    },
+    {
+      "epoch": 8.910944143514474,
+      "grad_norm": 0.33478838205337524,
+      "learning_rate": 0.0004933601398601398,
+      "loss": 3.4302,
+      "step": 30600
+    },
+    {
+      "epoch": 8.925505271128197,
+      "grad_norm": 0.35558992624282837,
+      "learning_rate": 0.0004931853146853146,
+      "loss": 3.4245,
+      "step": 30650
+    },
+    {
+      "epoch": 8.94006639874192,
+      "grad_norm": 0.3179035484790802,
+      "learning_rate": 0.0004930104895104895,
+      "loss": 3.4157,
+      "step": 30700
+    },
+    {
+      "epoch": 8.95462752635564,
+      "grad_norm": 0.32329732179641724,
+      "learning_rate": 0.0004928356643356642,
+      "loss": 3.4231,
+      "step": 30750
+    },
+    {
+      "epoch": 8.969188653969363,
+      "grad_norm": 0.3081290125846863,
+      "learning_rate": 0.0004926608391608391,
+      "loss": 3.4235,
+      "step": 30800
+    },
+    {
+      "epoch": 8.983749781583086,
+      "grad_norm": 0.3562612533569336,
+      "learning_rate": 0.0004924860139860139,
+      "loss": 3.4373,
+      "step": 30850
+    },
+    {
+      "epoch": 8.998310909196809,
+      "grad_norm": 0.3363591432571411,
+      "learning_rate": 0.0004923111888111887,
+      "loss": 3.4291,
+      "step": 30900
+    },
+    {
+      "epoch": 9.012813792300076,
+      "grad_norm": 0.33998963236808777,
+      "learning_rate": 0.0004921363636363635,
+      "loss": 3.333,
+      "step": 30950
+    },
+    {
+      "epoch": 9.027374919913798,
+      "grad_norm": 0.32185572385787964,
+      "learning_rate": 0.0004919615384615384,
+      "loss": 3.3145,
+      "step": 31000
+    },
+    {
+      "epoch": 9.027374919913798,
+      "eval_accuracy": 0.3672698177050697,
+      "eval_loss": 3.5787909030914307,
+      "eval_runtime": 181.1954,
+      "eval_samples_per_second": 91.868,
+      "eval_steps_per_second": 5.745,
+      "step": 31000
+    },
+    {
+      "epoch": 9.041936047527521,
+      "grad_norm": 0.33225589990615845,
+      "learning_rate": 0.0004917867132867132,
+      "loss": 3.3119,
+      "step": 31050
+    },
+    {
+      "epoch": 9.056497175141242,
+      "grad_norm": 0.34493061900138855,
+      "learning_rate": 0.000491611888111888,
+      "loss": 3.3234,
+      "step": 31100
+    },
+    {
+      "epoch": 9.071058302754965,
+      "grad_norm": 0.330806165933609,
+      "learning_rate": 0.0004914370629370628,
+      "loss": 3.339,
+      "step": 31150
+    },
+    {
+      "epoch": 9.085619430368688,
+      "grad_norm": 0.3284892737865448,
+      "learning_rate": 0.0004912622377622378,
+      "loss": 3.3447,
+      "step": 31200
+    },
+    {
+      "epoch": 9.10018055798241,
+      "grad_norm": 0.3395969271659851,
+      "learning_rate": 0.0004910874125874126,
+      "loss": 3.3239,
+      "step": 31250
+    },
+    {
+      "epoch": 9.114741685596133,
+      "grad_norm": 0.3392273783683777,
+      "learning_rate": 0.0004909125874125874,
+      "loss": 3.3447,
+      "step": 31300
+    },
+    {
+      "epoch": 9.129302813209854,
+      "grad_norm": 0.33892425894737244,
+      "learning_rate": 0.0004907377622377623,
+      "loss": 3.347,
+      "step": 31350
+    },
+    {
+      "epoch": 9.143863940823577,
+      "grad_norm": 0.3378540277481079,
+      "learning_rate": 0.0004905629370629371,
+      "loss": 3.3475,
+      "step": 31400
+    },
+    {
+      "epoch": 9.1584250684373,
+      "grad_norm": 0.3581297993659973,
+      "learning_rate": 0.0004903881118881119,
+      "loss": 3.3494,
+      "step": 31450
+    },
+    {
+      "epoch": 9.172986196051022,
+      "grad_norm": 0.34648507833480835,
+      "learning_rate": 0.0004902132867132867,
+      "loss": 3.3472,
+      "step": 31500
+    },
+    {
+      "epoch": 9.187547323664745,
+      "grad_norm": 0.32370853424072266,
+      "learning_rate": 0.0004900384615384615,
+      "loss": 3.3487,
+      "step": 31550
+    },
+    {
+      "epoch": 9.202108451278466,
+      "grad_norm": 0.3414316475391388,
+      "learning_rate": 0.0004898636363636363,
+      "loss": 3.3649,
+      "step": 31600
+    },
+    {
+      "epoch": 9.216669578892189,
+      "grad_norm": 0.3349761664867401,
+      "learning_rate": 0.0004896888111888112,
+      "loss": 3.374,
+      "step": 31650
+    },
+    {
+      "epoch": 9.231230706505912,
+      "grad_norm": 0.33960890769958496,
+      "learning_rate": 0.000489513986013986,
+      "loss": 3.3574,
+      "step": 31700
+    },
+    {
+      "epoch": 9.245791834119634,
+      "grad_norm": 0.3169540762901306,
+      "learning_rate": 0.0004893391608391608,
+      "loss": 3.3552,
+      "step": 31750
+    },
+    {
+      "epoch": 9.260352961733357,
+      "grad_norm": 0.3403364419937134,
+      "learning_rate": 0.0004891643356643356,
+      "loss": 3.3565,
+      "step": 31800
+    },
+    {
+      "epoch": 9.27491408934708,
+      "grad_norm": 0.32378724217414856,
+      "learning_rate": 0.0004889895104895105,
+      "loss": 3.3724,
+      "step": 31850
+    },
+    {
+      "epoch": 9.2894752169608,
+      "grad_norm": 0.32773780822753906,
+      "learning_rate": 0.0004888146853146853,
+      "loss": 3.3674,
+      "step": 31900
+    },
+    {
+      "epoch": 9.304036344574524,
+      "grad_norm": 0.3442809581756592,
+      "learning_rate": 0.0004886398601398601,
+      "loss": 3.3699,
+      "step": 31950
+    },
+    {
+      "epoch": 9.318597472188246,
+      "grad_norm": 0.35078513622283936,
+      "learning_rate": 0.000488465034965035,
+      "loss": 3.3916,
+      "step": 32000
+    },
+    {
+      "epoch": 9.318597472188246,
+      "eval_accuracy": 0.3675079995476191,
+      "eval_loss": 3.576542854309082,
+      "eval_runtime": 181.1266,
+      "eval_samples_per_second": 91.903,
+      "eval_steps_per_second": 5.747,
+      "step": 32000
+    },
+    {
+      "epoch": 9.333158599801969,
+      "grad_norm": 0.3307846188545227,
+      "learning_rate": 0.0004882902097902098,
+      "loss": 3.3732,
+      "step": 32050
+    },
+    {
+      "epoch": 9.347719727415692,
+      "grad_norm": 0.33339476585388184,
+      "learning_rate": 0.0004881153846153846,
+      "loss": 3.3788,
+      "step": 32100
+    },
+    {
+      "epoch": 9.362280855029413,
+      "grad_norm": 0.3615858852863312,
+      "learning_rate": 0.0004879405594405594,
+      "loss": 3.3832,
+      "step": 32150
+    },
+    {
+      "epoch": 9.376841982643136,
+      "grad_norm": 0.3240891993045807,
+      "learning_rate": 0.00048776573426573424,
+      "loss": 3.3818,
+      "step": 32200
+    },
+    {
+      "epoch": 9.391403110256858,
+      "grad_norm": 0.3376239538192749,
+      "learning_rate": 0.00048759090909090904,
+      "loss": 3.3769,
+      "step": 32250
+    },
+    {
+      "epoch": 9.405964237870581,
+      "grad_norm": 0.33899468183517456,
+      "learning_rate": 0.0004874160839160839,
+      "loss": 3.3806,
+      "step": 32300
+    },
+    {
+      "epoch": 9.420525365484304,
+      "grad_norm": 0.3659190237522125,
+      "learning_rate": 0.0004872412587412587,
+      "loss": 3.3799,
+      "step": 32350
+    },
+    {
+      "epoch": 9.435086493098025,
+      "grad_norm": 0.32162418961524963,
+      "learning_rate": 0.00048706643356643354,
+      "loss": 3.3742,
+      "step": 32400
+    },
+    {
+      "epoch": 9.449647620711747,
+      "grad_norm": 0.3324882686138153,
+      "learning_rate": 0.00048689160839160834,
+      "loss": 3.3826,
+      "step": 32450
+    },
+    {
+      "epoch": 9.46420874832547,
+      "grad_norm": 0.36081743240356445,
+      "learning_rate": 0.0004867167832167832,
+      "loss": 3.3747,
+      "step": 32500
+    },
+    {
+      "epoch": 9.478769875939193,
+      "grad_norm": 0.3334961533546448,
+      "learning_rate": 0.00048654195804195794,
+      "loss": 3.3889,
+      "step": 32550
+    },
+    {
+      "epoch": 9.493331003552916,
+      "grad_norm": 0.34522441029548645,
+      "learning_rate": 0.00048636713286713285,
+      "loss": 3.3826,
+      "step": 32600
+    },
+    {
+      "epoch": 9.507892131166638,
+      "grad_norm": 0.32337555289268494,
+      "learning_rate": 0.0004861923076923077,
+      "loss": 3.3825,
+      "step": 32650
+    },
+    {
+      "epoch": 9.52245325878036,
+      "grad_norm": 0.34286314249038696,
+      "learning_rate": 0.00048601748251748245,
+      "loss": 3.3901,
+      "step": 32700
+    },
+    {
+      "epoch": 9.537014386394082,
+      "grad_norm": 0.35533496737480164,
+      "learning_rate": 0.0004858426573426573,
+      "loss": 3.3923,
+      "step": 32750
+    },
+    {
+      "epoch": 9.551575514007805,
+      "grad_norm": 0.3230282962322235,
+      "learning_rate": 0.0004856678321678321,
+      "loss": 3.3914,
+      "step": 32800
+    },
+    {
+      "epoch": 9.566136641621528,
+      "grad_norm": 0.35662367939949036,
+      "learning_rate": 0.00048549300699300696,
+      "loss": 3.3924,
+      "step": 32850
+    },
+    {
+      "epoch": 9.58069776923525,
+      "grad_norm": 0.3356196880340576,
+      "learning_rate": 0.00048531818181818176,
+      "loss": 3.3961,
+      "step": 32900
+    },
+    {
+      "epoch": 9.595258896848971,
+      "grad_norm": 0.34684255719184875,
+      "learning_rate": 0.0004851433566433566,
+      "loss": 3.3887,
+      "step": 32950
+    },
+    {
+      "epoch": 9.609820024462694,
+      "grad_norm": 0.3561384081840515,
+      "learning_rate": 0.0004849685314685314,
+      "loss": 3.3941,
+      "step": 33000
+    },
+    {
+      "epoch": 9.609820024462694,
+      "eval_accuracy": 0.3682031472450496,
+      "eval_loss": 3.5684826374053955,
+      "eval_runtime": 181.122,
+      "eval_samples_per_second": 91.905,
+      "eval_steps_per_second": 5.748,
+      "step": 33000
+    },
+    {
+      "epoch": 9.624381152076417,
+      "grad_norm": 0.3471704125404358,
+      "learning_rate": 0.00048479370629370627,
+      "loss": 3.394,
+      "step": 33050
+    },
+    {
+      "epoch": 9.63894227969014,
+      "grad_norm": 0.3371427059173584,
+      "learning_rate": 0.00048461888111888106,
+      "loss": 3.3987,
+      "step": 33100
+    },
+    {
+      "epoch": 9.653503407303862,
+      "grad_norm": 0.3287416398525238,
+      "learning_rate": 0.0004844440559440559,
+      "loss": 3.4021,
+      "step": 33150
+    },
+    {
+      "epoch": 9.668064534917583,
+      "grad_norm": 0.3420673608779907,
+      "learning_rate": 0.0004842692307692307,
+      "loss": 3.4041,
+      "step": 33200
+    },
+    {
+      "epoch": 9.682625662531306,
+      "grad_norm": 0.36185222864151,
+      "learning_rate": 0.00048409440559440557,
+      "loss": 3.4019,
+      "step": 33250
+    },
+    {
+      "epoch": 9.697186790145029,
+      "grad_norm": 0.336234450340271,
+      "learning_rate": 0.0004839195804195803,
+      "loss": 3.3962,
+      "step": 33300
+    },
+    {
+      "epoch": 9.711747917758752,
+      "grad_norm": 0.30870339274406433,
+      "learning_rate": 0.0004837447552447552,
+      "loss": 3.3934,
+      "step": 33350
+    },
+    {
+      "epoch": 9.726309045372474,
+      "grad_norm": 0.3311539888381958,
+      "learning_rate": 0.0004835699300699301,
+      "loss": 3.3974,
+      "step": 33400
+    },
+    {
+      "epoch": 9.740870172986195,
+      "grad_norm": 0.32065773010253906,
+      "learning_rate": 0.0004833951048951048,
+      "loss": 3.4009,
+      "step": 33450
+    },
+    {
+      "epoch": 9.755431300599918,
+      "grad_norm": 0.31630703806877136,
+      "learning_rate": 0.0004832202797202797,
+      "loss": 3.3958,
+      "step": 33500
+    },
+    {
+      "epoch": 9.76999242821364,
+      "grad_norm": 0.34896165132522583,
+      "learning_rate": 0.0004830454545454545,
+      "loss": 3.3964,
+      "step": 33550
+    },
+    {
+      "epoch": 9.784553555827364,
+      "grad_norm": 0.350842148065567,
+      "learning_rate": 0.00048287062937062933,
+      "loss": 3.3972,
+      "step": 33600
+    },
+    {
+      "epoch": 9.799114683441086,
+      "grad_norm": 0.33389005064964294,
+      "learning_rate": 0.00048269580419580413,
+      "loss": 3.3946,
+      "step": 33650
+    },
+    {
+      "epoch": 9.813675811054807,
+      "grad_norm": 0.3580504059791565,
+      "learning_rate": 0.000482520979020979,
+      "loss": 3.3936,
+      "step": 33700
+    },
+    {
+      "epoch": 9.82823693866853,
+      "grad_norm": 0.3484356701374054,
+      "learning_rate": 0.0004823461538461538,
+      "loss": 3.4076,
+      "step": 33750
+    },
+    {
+      "epoch": 9.842798066282253,
+      "grad_norm": 0.32637032866477966,
+      "learning_rate": 0.00048217132867132864,
+      "loss": 3.4109,
+      "step": 33800
+    },
+    {
+      "epoch": 9.857359193895975,
+      "grad_norm": 0.3298273980617523,
+      "learning_rate": 0.00048199650349650344,
+      "loss": 3.3958,
+      "step": 33850
+    },
+    {
+      "epoch": 9.871920321509698,
+      "grad_norm": 0.33490678668022156,
+      "learning_rate": 0.0004818216783216783,
+      "loss": 3.4054,
+      "step": 33900
+    },
+    {
+      "epoch": 9.88648144912342,
+      "grad_norm": 0.33199331164360046,
+      "learning_rate": 0.0004816468531468531,
+      "loss": 3.3937,
+      "step": 33950
+    },
+    {
+      "epoch": 9.901042576737142,
+      "grad_norm": 0.3390347957611084,
+      "learning_rate": 0.00048147202797202795,
+      "loss": 3.3895,
+      "step": 34000
+    },
+    {
+      "epoch": 9.901042576737142,
+      "eval_accuracy": 0.36896742175561886,
+      "eval_loss": 3.5596354007720947,
+      "eval_runtime": 181.1726,
+      "eval_samples_per_second": 91.879,
+      "eval_steps_per_second": 5.746,
+      "step": 34000
+    },
+    {
+      "epoch": 9.915603704350865,
+      "grad_norm": 0.33487066626548767,
+      "learning_rate": 0.0004812972027972028,
+      "loss": 3.3921,
+      "step": 34050
+    },
+    {
+      "epoch": 9.930164831964587,
+      "grad_norm": 0.3349563479423523,
+      "learning_rate": 0.0004811223776223776,
+      "loss": 3.4017,
+      "step": 34100
+    },
+    {
+      "epoch": 9.94472595957831,
+      "grad_norm": 0.33379611372947693,
+      "learning_rate": 0.00048094755244755245,
+      "loss": 3.4057,
+      "step": 34150
+    },
+    {
+      "epoch": 9.959287087192033,
+      "grad_norm": 0.30914467573165894,
+      "learning_rate": 0.0004807727272727272,
+      "loss": 3.4112,
+      "step": 34200
+    },
+    {
+      "epoch": 9.973848214805754,
+      "grad_norm": 0.3371839225292206,
+      "learning_rate": 0.00048059790209790205,
+      "loss": 3.4083,
+      "step": 34250
+    },
+    {
+      "epoch": 9.988409342419477,
+      "grad_norm": 0.33769628405570984,
+      "learning_rate": 0.00048042307692307685,
+      "loss": 3.4224,
+      "step": 34300
+    },
+    {
+      "epoch": 10.002912225522744,
+      "grad_norm": 0.33109721541404724,
+      "learning_rate": 0.0004802482517482517,
+      "loss": 3.3967,
+      "step": 34350
+    },
+    {
+      "epoch": 10.017473353136467,
+      "grad_norm": 0.344075083732605,
+      "learning_rate": 0.0004800734265734265,
+      "loss": 3.3015,
+      "step": 34400
+    },
+    {
+      "epoch": 10.03203448075019,
+      "grad_norm": 0.3365168273448944,
+      "learning_rate": 0.00047989860139860136,
+      "loss": 3.295,
+      "step": 34450
+    },
+    {
+      "epoch": 10.046595608363912,
+      "grad_norm": 0.336918443441391,
+      "learning_rate": 0.00047972377622377616,
+      "loss": 3.3091,
+      "step": 34500
+    },
+    {
+      "epoch": 10.061156735977635,
+      "grad_norm": 0.3636518716812134,
+      "learning_rate": 0.000479548951048951,
+      "loss": 3.3087,
+      "step": 34550
+    },
+    {
+      "epoch": 10.075717863591356,
+      "grad_norm": 0.35915517807006836,
+      "learning_rate": 0.0004793741258741258,
+      "loss": 3.2959,
+      "step": 34600
+    },
+    {
+      "epoch": 10.090278991205079,
+      "grad_norm": 0.35610800981521606,
+      "learning_rate": 0.00047919930069930067,
+      "loss": 3.3225,
+      "step": 34650
+    },
+    {
+      "epoch": 10.104840118818801,
+      "grad_norm": 0.34759587049484253,
+      "learning_rate": 0.0004790244755244755,
+      "loss": 3.3141,
+      "step": 34700
+    },
+    {
+      "epoch": 10.119401246432524,
+      "grad_norm": 0.3453555405139923,
+      "learning_rate": 0.0004788496503496503,
+      "loss": 3.3289,
+      "step": 34750
+    },
+    {
+      "epoch": 10.133962374046247,
+      "grad_norm": 0.3237037658691406,
+      "learning_rate": 0.0004786748251748252,
+      "loss": 3.3033,
+      "step": 34800
+    },
+    {
+      "epoch": 10.148523501659968,
+      "grad_norm": 0.3217065632343292,
+      "learning_rate": 0.0004785,
+      "loss": 3.3222,
+      "step": 34850
+    },
+    {
+      "epoch": 10.16308462927369,
+      "grad_norm": 0.328022837638855,
+      "learning_rate": 0.00047832517482517483,
+      "loss": 3.3252,
+      "step": 34900
+    },
+    {
+      "epoch": 10.177645756887413,
+      "grad_norm": 0.3419390916824341,
+      "learning_rate": 0.0004781503496503496,
+      "loss": 3.3357,
+      "step": 34950
+    },
+    {
+      "epoch": 10.192206884501136,
+      "grad_norm": 0.35278406739234924,
+      "learning_rate": 0.00047797552447552443,
+      "loss": 3.3348,
+      "step": 35000
+    },
+    {
+      "epoch": 10.192206884501136,
+      "eval_accuracy": 0.3681650569602589,
+      "eval_loss": 3.5723941326141357,
+      "eval_runtime": 204.5444,
+      "eval_samples_per_second": 81.381,
+      "eval_steps_per_second": 5.089,
+      "step": 35000
+    },
+    {
+      "epoch": 10.206768012114859,
+      "grad_norm": 0.3506302833557129,
+      "learning_rate": 0.00047780069930069923,
+      "loss": 3.3334,
+      "step": 35050
+    },
+    {
+      "epoch": 10.221329139728581,
+      "grad_norm": 0.3560106158256531,
+      "learning_rate": 0.0004776258741258741,
+      "loss": 3.328,
+      "step": 35100
+    },
+    {
+      "epoch": 10.235890267342302,
+      "grad_norm": 0.3484821319580078,
+      "learning_rate": 0.0004774510489510489,
+      "loss": 3.3401,
+      "step": 35150
+    },
+    {
+      "epoch": 10.250451394956025,
+      "grad_norm": 0.3474741280078888,
+      "learning_rate": 0.00047727622377622374,
+      "loss": 3.3448,
+      "step": 35200
+    },
+    {
+      "epoch": 10.265012522569748,
+      "grad_norm": 0.3430100083351135,
+      "learning_rate": 0.00047710139860139854,
+      "loss": 3.3393,
+      "step": 35250
+    },
+    {
+      "epoch": 10.27957365018347,
+      "grad_norm": 0.3785252571105957,
+      "learning_rate": 0.0004769265734265734,
+      "loss": 3.3382,
+      "step": 35300
+    },
+    {
+      "epoch": 10.294134777797193,
+      "grad_norm": 0.3831712007522583,
+      "learning_rate": 0.0004767517482517482,
+      "loss": 3.3491,
+      "step": 35350
+    },
+    {
+      "epoch": 10.308695905410914,
+      "grad_norm": 0.3325560390949249,
+      "learning_rate": 0.00047657692307692304,
+      "loss": 3.3533,
+      "step": 35400
+    },
+    {
+      "epoch": 10.323257033024637,
+      "grad_norm": 0.3680790066719055,
+      "learning_rate": 0.0004764020979020979,
+      "loss": 3.3613,
+      "step": 35450
+    },
+    {
+      "epoch": 10.33781816063836,
+      "grad_norm": 0.34670522809028625,
+      "learning_rate": 0.0004762272727272727,
+      "loss": 3.3505,
+      "step": 35500
+    },
+    {
+      "epoch": 10.352379288252083,
+      "grad_norm": 0.34935176372528076,
+      "learning_rate": 0.00047605244755244755,
+      "loss": 3.3521,
+      "step": 35550
+    },
+    {
+      "epoch": 10.366940415865805,
+      "grad_norm": 0.35227638483047485,
+      "learning_rate": 0.00047587762237762235,
+      "loss": 3.3542,
+      "step": 35600
+    },
+    {
+      "epoch": 10.381501543479526,
+      "grad_norm": 0.3471412658691406,
+      "learning_rate": 0.0004757027972027972,
+      "loss": 3.3576,
+      "step": 35650
+    },
+    {
+      "epoch": 10.396062671093249,
+      "grad_norm": 0.3685144782066345,
+      "learning_rate": 0.00047552797202797195,
+      "loss": 3.355,
+      "step": 35700
+    },
+    {
+      "epoch": 10.410623798706972,
+      "grad_norm": 0.3381234109401703,
+      "learning_rate": 0.0004753531468531468,
+      "loss": 3.3566,
+      "step": 35750
+    },
+    {
+      "epoch": 10.425184926320695,
+      "grad_norm": 0.3421459197998047,
+      "learning_rate": 0.0004751783216783216,
+      "loss": 3.3541,
+      "step": 35800
+    },
+    {
+      "epoch": 10.439746053934417,
+      "grad_norm": 0.344374418258667,
+      "learning_rate": 0.00047500349650349646,
+      "loss": 3.3754,
+      "step": 35850
+    },
+    {
+      "epoch": 10.454307181548138,
+      "grad_norm": 0.3318965435028076,
+      "learning_rate": 0.00047482867132867126,
+      "loss": 3.3728,
+      "step": 35900
+    },
+    {
+      "epoch": 10.468868309161861,
+      "grad_norm": 0.3651151657104492,
+      "learning_rate": 0.0004746538461538461,
+      "loss": 3.3506,
+      "step": 35950
+    },
+    {
+      "epoch": 10.483429436775584,
+      "grad_norm": 0.407438725233078,
+      "learning_rate": 0.0004744790209790209,
+      "loss": 3.3676,
+      "step": 36000
+    },
+    {
+      "epoch": 10.483429436775584,
+      "eval_accuracy": 0.3688570304672902,
+      "eval_loss": 3.5666871070861816,
+      "eval_runtime": 181.1422,
+      "eval_samples_per_second": 91.895,
+      "eval_steps_per_second": 5.747,
+      "step": 36000
+    },
+    {
+      "epoch": 10.497990564389307,
+      "grad_norm": 0.3370654881000519,
+      "learning_rate": 0.00047430419580419576,
+      "loss": 3.3636,
+      "step": 36050
+    },
+    {
+      "epoch": 10.51255169200303,
+      "grad_norm": 0.3517342805862427,
+      "learning_rate": 0.0004741293706293706,
+      "loss": 3.3626,
+      "step": 36100
+    },
+    {
+      "epoch": 10.52711281961675,
+      "grad_norm": 0.3267398476600647,
+      "learning_rate": 0.0004739545454545454,
+      "loss": 3.3724,
+      "step": 36150
+    },
+    {
+      "epoch": 10.541673947230473,
+      "grad_norm": 0.3496518135070801,
+      "learning_rate": 0.00047377972027972027,
+      "loss": 3.3583,
+      "step": 36200
+    },
+    {
+      "epoch": 10.556235074844196,
+      "grad_norm": 0.3922159671783447,
+      "learning_rate": 0.00047360489510489507,
+      "loss": 3.3742,
+      "step": 36250
+    },
+    {
+      "epoch": 10.570796202457919,
+      "grad_norm": 0.32678449153900146,
+      "learning_rate": 0.0004734300699300699,
+      "loss": 3.3647,
+      "step": 36300
+    },
+    {
+      "epoch": 10.585357330071641,
+      "grad_norm": 0.3430236577987671,
+      "learning_rate": 0.0004732552447552447,
+      "loss": 3.3716,
+      "step": 36350
+    },
+    {
+      "epoch": 10.599918457685362,
+      "grad_norm": 0.3338622450828552,
+      "learning_rate": 0.0004730804195804196,
+      "loss": 3.3735,
+      "step": 36400
+    },
+    {
+      "epoch": 10.614479585299085,
+      "grad_norm": 0.37562769651412964,
+      "learning_rate": 0.0004729055944055943,
+      "loss": 3.3813,
+      "step": 36450
+    },
+    {
+      "epoch": 10.629040712912808,
+      "grad_norm": 0.32546374201774597,
+      "learning_rate": 0.0004727307692307692,
+      "loss": 3.3721,
+      "step": 36500
+    },
+    {
+      "epoch": 10.64360184052653,
+      "grad_norm": 0.34546393156051636,
+      "learning_rate": 0.000472555944055944,
+      "loss": 3.3797,
+      "step": 36550
+    },
+    {
+      "epoch": 10.658162968140253,
+      "grad_norm": 0.3651503026485443,
+      "learning_rate": 0.00047238111888111883,
+      "loss": 3.3764,
+      "step": 36600
+    },
+    {
+      "epoch": 10.672724095753976,
+      "grad_norm": 0.3390817642211914,
+      "learning_rate": 0.00047220629370629363,
+      "loss": 3.3635,
+      "step": 36650
+    },
+    {
+      "epoch": 10.687285223367697,
+      "grad_norm": 0.35961946845054626,
+      "learning_rate": 0.0004720314685314685,
+      "loss": 3.379,
+      "step": 36700
+    },
+    {
+      "epoch": 10.70184635098142,
+      "grad_norm": 0.3362646698951721,
+      "learning_rate": 0.0004718566433566433,
+      "loss": 3.3717,
+      "step": 36750
+    },
+    {
+      "epoch": 10.716407478595142,
+      "grad_norm": 0.34462296962738037,
+      "learning_rate": 0.00047168181818181814,
+      "loss": 3.3879,
+      "step": 36800
+    },
+    {
+      "epoch": 10.730968606208865,
+      "grad_norm": 0.35153433680534363,
+      "learning_rate": 0.000471506993006993,
+      "loss": 3.3792,
+      "step": 36850
+    },
+    {
+      "epoch": 10.745529733822588,
+      "grad_norm": 0.3449765741825104,
+      "learning_rate": 0.0004713321678321678,
+      "loss": 3.3798,
+      "step": 36900
+    },
+    {
+      "epoch": 10.760090861436309,
+      "grad_norm": 0.3359948992729187,
+      "learning_rate": 0.00047115734265734265,
+      "loss": 3.3809,
+      "step": 36950
+    },
+    {
+      "epoch": 10.774651989050032,
+      "grad_norm": 0.35020774602890015,
+      "learning_rate": 0.00047098251748251745,
+      "loss": 3.3776,
+      "step": 37000
+    },
+    {
+      "epoch": 10.774651989050032,
+      "eval_accuracy": 0.369682319971089,
+      "eval_loss": 3.555363416671753,
+      "eval_runtime": 181.073,
+      "eval_samples_per_second": 91.93,
+      "eval_steps_per_second": 5.749,
+      "step": 37000
+    },
+    {
+      "epoch": 10.789213116663754,
+      "grad_norm": 0.33577054738998413,
+      "learning_rate": 0.0004708076923076923,
+      "loss": 3.3843,
+      "step": 37050
+    },
+    {
+      "epoch": 10.803774244277477,
+      "grad_norm": 0.3409827649593353,
+      "learning_rate": 0.0004706328671328671,
+      "loss": 3.3745,
+      "step": 37100
+    },
+    {
+      "epoch": 10.8183353718912,
+      "grad_norm": 0.3596167266368866,
+      "learning_rate": 0.00047045804195804195,
+      "loss": 3.3734,
+      "step": 37150
+    },
+    {
+      "epoch": 10.83289649950492,
+      "grad_norm": 0.3197990953922272,
+      "learning_rate": 0.0004702832167832167,
+      "loss": 3.383,
+      "step": 37200
+    },
+    {
+      "epoch": 10.847457627118644,
+      "grad_norm": 0.31918731331825256,
+      "learning_rate": 0.00047010839160839155,
+      "loss": 3.3748,
+      "step": 37250
+    },
+    {
+      "epoch": 10.862018754732366,
+      "grad_norm": 0.3445984721183777,
+      "learning_rate": 0.00046993356643356635,
+      "loss": 3.3725,
+      "step": 37300
+    },
+    {
+      "epoch": 10.876579882346089,
+      "grad_norm": 0.37393128871917725,
+      "learning_rate": 0.0004697587412587412,
+      "loss": 3.3707,
+      "step": 37350
+    },
+    {
+      "epoch": 10.891141009959812,
+      "grad_norm": 0.34462884068489075,
+      "learning_rate": 0.000469583916083916,
+      "loss": 3.3879,
+      "step": 37400
+    },
+    {
+      "epoch": 10.905702137573535,
+      "grad_norm": 0.3394433259963989,
+      "learning_rate": 0.00046940909090909086,
+      "loss": 3.391,
+      "step": 37450
+    },
+    {
+      "epoch": 10.920263265187256,
+      "grad_norm": 0.34529218077659607,
+      "learning_rate": 0.0004692342657342657,
+      "loss": 3.3895,
+      "step": 37500
+    },
+    {
+      "epoch": 10.934824392800978,
+      "grad_norm": 0.3486793637275696,
+      "learning_rate": 0.0004690594405594405,
+      "loss": 3.392,
+      "step": 37550
+    },
+    {
+      "epoch": 10.949385520414701,
+      "grad_norm": 0.33855584263801575,
+      "learning_rate": 0.00046888461538461537,
+      "loss": 3.3722,
+      "step": 37600
+    },
+    {
+      "epoch": 10.963946648028424,
+      "grad_norm": 0.3652046322822571,
+      "learning_rate": 0.00046870979020979017,
+      "loss": 3.3803,
+      "step": 37650
+    },
+    {
+      "epoch": 10.978507775642147,
+      "grad_norm": 0.3586157262325287,
+      "learning_rate": 0.000468534965034965,
+      "loss": 3.3837,
+      "step": 37700
+    },
+    {
+      "epoch": 10.993068903255867,
+      "grad_norm": 0.333840548992157,
+      "learning_rate": 0.0004683601398601398,
+      "loss": 3.3817,
+      "step": 37750
+    },
+    {
+      "epoch": 11.007571786359136,
+      "grad_norm": 0.3377648890018463,
+      "learning_rate": 0.0004681853146853147,
+      "loss": 3.3336,
+      "step": 37800
+    },
+    {
+      "epoch": 11.022132913972857,
+      "grad_norm": 0.335593581199646,
+      "learning_rate": 0.0004680104895104895,
+      "loss": 3.2756,
+      "step": 37850
+    },
+    {
+      "epoch": 11.03669404158658,
+      "grad_norm": 0.35685721039772034,
+      "learning_rate": 0.00046783566433566433,
+      "loss": 3.2872,
+      "step": 37900
+    },
+    {
+      "epoch": 11.051255169200303,
+      "grad_norm": 0.3342174291610718,
+      "learning_rate": 0.0004676608391608391,
+      "loss": 3.2909,
+      "step": 37950
+    },
+    {
+      "epoch": 11.065816296814026,
+      "grad_norm": 0.33387836813926697,
+      "learning_rate": 0.00046748601398601393,
+      "loss": 3.2947,
+      "step": 38000
+    },
+    {
+      "epoch": 11.065816296814026,
+      "eval_accuracy": 0.3693142314473861,
+      "eval_loss": 3.567906618118286,
+      "eval_runtime": 181.2168,
+      "eval_samples_per_second": 91.857,
+      "eval_steps_per_second": 5.744,
+      "step": 38000
+    },
+    {
+      "epoch": 11.080377424427748,
+      "grad_norm": 0.35043972730636597,
+      "learning_rate": 0.00046731118881118873,
+      "loss": 3.2817,
+      "step": 38050
+    },
+    {
+      "epoch": 11.09493855204147,
+      "grad_norm": 0.34657034277915955,
+      "learning_rate": 0.0004671363636363636,
+      "loss": 3.2913,
+      "step": 38100
+    },
+    {
+      "epoch": 11.109499679655192,
+      "grad_norm": 0.3644031882286072,
+      "learning_rate": 0.00046696153846153844,
+      "loss": 3.3013,
+      "step": 38150
+    },
+    {
+      "epoch": 11.124060807268915,
+      "grad_norm": 0.35239455103874207,
+      "learning_rate": 0.00046678671328671324,
+      "loss": 3.304,
+      "step": 38200
+    },
+    {
+      "epoch": 11.138621934882638,
+      "grad_norm": 0.3717592656612396,
+      "learning_rate": 0.0004666118881118881,
+      "loss": 3.302,
+      "step": 38250
+    },
+    {
+      "epoch": 11.15318306249636,
+      "grad_norm": 0.34851017594337463,
+      "learning_rate": 0.0004664370629370629,
+      "loss": 3.3065,
+      "step": 38300
+    },
+    {
+      "epoch": 11.167744190110081,
+      "grad_norm": 0.35993218421936035,
+      "learning_rate": 0.00046626223776223774,
+      "loss": 3.3046,
+      "step": 38350
+    },
+    {
+      "epoch": 11.182305317723804,
+      "grad_norm": 0.35800787806510925,
+      "learning_rate": 0.00046608741258741254,
+      "loss": 3.3082,
+      "step": 38400
+    },
+    {
+      "epoch": 11.196866445337527,
+      "grad_norm": 0.3414912223815918,
+      "learning_rate": 0.0004659125874125874,
+      "loss": 3.2973,
+      "step": 38450
+    },
+    {
+      "epoch": 11.21142757295125,
+      "grad_norm": 0.34826183319091797,
+      "learning_rate": 0.0004657377622377622,
+      "loss": 3.3153,
+      "step": 38500
+    },
+    {
+      "epoch": 11.225988700564972,
+      "grad_norm": 0.3334352374076843,
+      "learning_rate": 0.00046556293706293705,
+      "loss": 3.3123,
+      "step": 38550
+    },
+    {
+      "epoch": 11.240549828178693,
+      "grad_norm": 0.3609830439090729,
+      "learning_rate": 0.00046538811188811185,
+      "loss": 3.323,
+      "step": 38600
+    },
+    {
+      "epoch": 11.255110955792416,
+      "grad_norm": 0.36535823345184326,
+      "learning_rate": 0.0004652132867132867,
+      "loss": 3.3181,
+      "step": 38650
+    },
+    {
+      "epoch": 11.269672083406139,
+      "grad_norm": 0.35265401005744934,
+      "learning_rate": 0.00046503846153846145,
+      "loss": 3.3222,
+      "step": 38700
+    },
+    {
+      "epoch": 11.284233211019862,
+      "grad_norm": 0.34238654375076294,
+      "learning_rate": 0.0004648636363636363,
+      "loss": 3.3289,
+      "step": 38750
+    },
+    {
+      "epoch": 11.298794338633584,
+      "grad_norm": 0.38800933957099915,
+      "learning_rate": 0.0004646888111888111,
+      "loss": 3.3386,
+      "step": 38800
+    },
+    {
+      "epoch": 11.313355466247307,
+      "grad_norm": 0.36785754561424255,
+      "learning_rate": 0.00046451398601398596,
+      "loss": 3.3245,
+      "step": 38850
+    },
+    {
+      "epoch": 11.327916593861028,
+      "grad_norm": 0.34194216132164,
+      "learning_rate": 0.0004643391608391608,
+      "loss": 3.3234,
+      "step": 38900
+    },
+    {
+      "epoch": 11.34247772147475,
+      "grad_norm": 0.3501145541667938,
+      "learning_rate": 0.0004641643356643356,
+      "loss": 3.3333,
+      "step": 38950
+    },
+    {
+      "epoch": 11.357038849088473,
+      "grad_norm": 0.3359377384185791,
+      "learning_rate": 0.00046398951048951046,
+      "loss": 3.3428,
+      "step": 39000
+    },
+    {
+      "epoch": 11.357038849088473,
+      "eval_accuracy": 0.3693486772913481,
+      "eval_loss": 3.5647213459014893,
+      "eval_runtime": 181.321,
+      "eval_samples_per_second": 91.804,
+      "eval_steps_per_second": 5.741,
+      "step": 39000
+    },
+    {
+      "epoch": 11.371599976702196,
+      "grad_norm": 0.3317386507987976,
+      "learning_rate": 0.00046381468531468526,
+      "loss": 3.3287,
+      "step": 39050
+    },
+    {
+      "epoch": 11.386161104315919,
+      "grad_norm": 0.35291844606399536,
+      "learning_rate": 0.0004636398601398601,
+      "loss": 3.3348,
+      "step": 39100
+    },
+    {
+      "epoch": 11.40072223192964,
+      "grad_norm": 0.3461180627346039,
+      "learning_rate": 0.0004634650349650349,
+      "loss": 3.3244,
+      "step": 39150
+    },
+    {
+      "epoch": 11.415283359543363,
+      "grad_norm": 0.3356523811817169,
+      "learning_rate": 0.00046329020979020977,
+      "loss": 3.3411,
+      "step": 39200
+    },
+    {
+      "epoch": 11.429844487157085,
+      "grad_norm": 0.3385235369205475,
+      "learning_rate": 0.00046311538461538457,
+      "loss": 3.3413,
+      "step": 39250
+    },
+    {
+      "epoch": 11.444405614770808,
+      "grad_norm": 0.34884148836135864,
+      "learning_rate": 0.0004629405594405594,
+      "loss": 3.3485,
+      "step": 39300
+    },
+    {
+      "epoch": 11.458966742384531,
+      "grad_norm": 0.36972281336784363,
+      "learning_rate": 0.0004627657342657342,
+      "loss": 3.3496,
+      "step": 39350
+    },
+    {
+      "epoch": 11.473527869998252,
+      "grad_norm": 0.3516969084739685,
+      "learning_rate": 0.0004625909090909091,
+      "loss": 3.3302,
+      "step": 39400
+    },
+    {
+      "epoch": 11.488088997611975,
+      "grad_norm": 0.3294467329978943,
+      "learning_rate": 0.0004624160839160838,
+      "loss": 3.3499,
+      "step": 39450
+    },
+    {
+      "epoch": 11.502650125225697,
+      "grad_norm": 0.3424534201622009,
+      "learning_rate": 0.0004622412587412587,
+      "loss": 3.3462,
+      "step": 39500
+    },
+    {
+      "epoch": 11.51721125283942,
+      "grad_norm": 0.363540917634964,
+      "learning_rate": 0.00046206643356643353,
+      "loss": 3.3483,
+      "step": 39550
+    },
+    {
+      "epoch": 11.531772380453143,
+      "grad_norm": 0.37824833393096924,
+      "learning_rate": 0.00046189160839160833,
+      "loss": 3.3442,
+      "step": 39600
+    },
+    {
+      "epoch": 11.546333508066864,
+      "grad_norm": 0.34854546189308167,
+      "learning_rate": 0.0004617167832167832,
+      "loss": 3.3494,
+      "step": 39650
+    },
+    {
+      "epoch": 11.560894635680587,
+      "grad_norm": 0.3524269461631775,
+      "learning_rate": 0.000461541958041958,
+      "loss": 3.3448,
+      "step": 39700
+    },
+    {
+      "epoch": 11.57545576329431,
+      "grad_norm": 0.3473973870277405,
+      "learning_rate": 0.00046136713286713284,
+      "loss": 3.3549,
+      "step": 39750
+    },
+    {
+      "epoch": 11.590016890908032,
+      "grad_norm": 0.35165148973464966,
+      "learning_rate": 0.00046119230769230764,
+      "loss": 3.3442,
+      "step": 39800
+    },
+    {
+      "epoch": 11.604578018521755,
+      "grad_norm": 0.3505858778953552,
+      "learning_rate": 0.0004610174825174825,
+      "loss": 3.3533,
+      "step": 39850
+    },
+    {
+      "epoch": 11.619139146135478,
+      "grad_norm": 0.3432624638080597,
+      "learning_rate": 0.0004608426573426573,
+      "loss": 3.3463,
+      "step": 39900
+    },
+    {
+      "epoch": 11.633700273749199,
+      "grad_norm": 0.35243409872055054,
+      "learning_rate": 0.00046066783216783215,
+      "loss": 3.3491,
+      "step": 39950
+    },
+    {
+      "epoch": 11.648261401362921,
+      "grad_norm": 0.32807716727256775,
+      "learning_rate": 0.00046049300699300695,
+      "loss": 3.3673,
+      "step": 40000
+    },
+    {
+      "epoch": 11.648261401362921,
+      "eval_accuracy": 0.3696696232094921,
+      "eval_loss": 3.55596661567688,
+      "eval_runtime": 181.2437,
+      "eval_samples_per_second": 91.843,
+      "eval_steps_per_second": 5.744,
+      "step": 40000
+    },
+    {
+      "epoch": 11.662822528976644,
+      "grad_norm": 0.3393252193927765,
+      "learning_rate": 0.0004603181818181818,
+      "loss": 3.3652,
+      "step": 40050
+    },
+    {
+      "epoch": 11.677383656590367,
+      "grad_norm": 0.34770268201828003,
+      "learning_rate": 0.0004601433566433566,
+      "loss": 3.3548,
+      "step": 40100
+    },
+    {
+      "epoch": 11.69194478420409,
+      "grad_norm": 0.35043734312057495,
+      "learning_rate": 0.00045996853146853145,
+      "loss": 3.3571,
+      "step": 40150
+    },
+    {
+      "epoch": 11.70650591181781,
+      "grad_norm": 0.3447571098804474,
+      "learning_rate": 0.0004597937062937062,
+      "loss": 3.3577,
+      "step": 40200
+    },
+    {
+      "epoch": 11.721067039431533,
+      "grad_norm": 0.35654211044311523,
+      "learning_rate": 0.00045961888111888105,
+      "loss": 3.36,
+      "step": 40250
+    },
+    {
+      "epoch": 11.735628167045256,
+      "grad_norm": 0.36046501994132996,
+      "learning_rate": 0.0004594440559440559,
+      "loss": 3.3633,
+      "step": 40300
+    },
+    {
+      "epoch": 11.750189294658979,
+      "grad_norm": 0.32962411642074585,
+      "learning_rate": 0.0004592692307692307,
+      "loss": 3.362,
+      "step": 40350
+    },
+    {
+      "epoch": 11.764750422272702,
+      "grad_norm": 0.34335944056510925,
+      "learning_rate": 0.00045909440559440556,
+      "loss": 3.3376,
+      "step": 40400
+    },
+    {
+      "epoch": 11.779311549886422,
+      "grad_norm": 0.3403267562389374,
+      "learning_rate": 0.00045891958041958036,
+      "loss": 3.3643,
+      "step": 40450
+    },
+    {
+      "epoch": 11.793872677500145,
+      "grad_norm": 0.3420957326889038,
+      "learning_rate": 0.0004587447552447552,
+      "loss": 3.3683,
+      "step": 40500
+    },
+    {
+      "epoch": 11.808433805113868,
+      "grad_norm": 0.3415950834751129,
+      "learning_rate": 0.00045856993006993,
+      "loss": 3.3641,
+      "step": 40550
+    },
+    {
+      "epoch": 11.82299493272759,
+      "grad_norm": 0.3702256679534912,
+      "learning_rate": 0.00045839510489510487,
+      "loss": 3.3513,
+      "step": 40600
+    },
+    {
+      "epoch": 11.837556060341313,
+      "grad_norm": 0.3508910536766052,
+      "learning_rate": 0.00045822027972027967,
+      "loss": 3.368,
+      "step": 40650
+    },
+    {
+      "epoch": 11.852117187955034,
+      "grad_norm": 0.35359275341033936,
+      "learning_rate": 0.0004580454545454545,
+      "loss": 3.3677,
+      "step": 40700
+    },
+    {
+      "epoch": 11.866678315568757,
+      "grad_norm": 0.33525702357292175,
+      "learning_rate": 0.0004578706293706293,
+      "loss": 3.3598,
+      "step": 40750
+    },
+    {
+      "epoch": 11.88123944318248,
+      "grad_norm": 0.3661658763885498,
+      "learning_rate": 0.0004576958041958042,
+      "loss": 3.3576,
+      "step": 40800
+    },
+    {
+      "epoch": 11.895800570796203,
+      "grad_norm": 0.35114672780036926,
+      "learning_rate": 0.000457520979020979,
+      "loss": 3.3659,
+      "step": 40850
+    },
+    {
+      "epoch": 11.910361698409925,
+      "grad_norm": 0.3382083475589752,
+      "learning_rate": 0.00045734615384615383,
+      "loss": 3.3659,
+      "step": 40900
+    },
+    {
+      "epoch": 11.924922826023646,
+      "grad_norm": 0.3518007695674896,
+      "learning_rate": 0.0004571713286713287,
+      "loss": 3.3552,
+      "step": 40950
+    },
+    {
+      "epoch": 11.93948395363737,
+      "grad_norm": 0.3505655825138092,
+      "learning_rate": 0.00045699650349650343,
+      "loss": 3.3775,
+      "step": 41000
+    },
+    {
+      "epoch": 11.93948395363737,
+      "eval_accuracy": 0.3705243033651356,
+      "eval_loss": 3.5475268363952637,
+      "eval_runtime": 181.3685,
+      "eval_samples_per_second": 91.78,
+      "eval_steps_per_second": 5.74,
+      "step": 41000
+    },
+    {
+      "epoch": 11.954045081251092,
+      "grad_norm": 0.31989786028862,
+      "learning_rate": 0.0004568216783216783,
+      "loss": 3.371,
+      "step": 41050
+    },
+    {
+      "epoch": 11.968606208864815,
+      "grad_norm": 0.3351122736930847,
+      "learning_rate": 0.0004566468531468531,
+      "loss": 3.3533,
+      "step": 41100
+    },
+    {
+      "epoch": 11.983167336478537,
+      "grad_norm": 0.33592069149017334,
+      "learning_rate": 0.00045647202797202794,
+      "loss": 3.3761,
+      "step": 41150
+    },
+    {
+      "epoch": 11.99772846409226,
+      "grad_norm": 0.3345586657524109,
+      "learning_rate": 0.00045629720279720274,
+      "loss": 3.3698,
+      "step": 41200
+    },
+    {
+      "epoch": 12.012231347195527,
+      "grad_norm": 0.3332180380821228,
+      "learning_rate": 0.0004561223776223776,
+      "loss": 3.272,
+      "step": 41250
+    },
+    {
+      "epoch": 12.02679247480925,
+      "grad_norm": 0.33513668179512024,
+      "learning_rate": 0.0004559475524475524,
+      "loss": 3.2583,
+      "step": 41300
+    },
+    {
+      "epoch": 12.041353602422971,
+      "grad_norm": 0.3612571060657501,
+      "learning_rate": 0.00045577272727272724,
+      "loss": 3.264,
+      "step": 41350
+    },
+    {
+      "epoch": 12.055914730036694,
+      "grad_norm": 0.3716042935848236,
+      "learning_rate": 0.00045559790209790204,
+      "loss": 3.2606,
+      "step": 41400
+    },
+    {
+      "epoch": 12.070475857650417,
+      "grad_norm": 0.3391723036766052,
+      "learning_rate": 0.0004554230769230769,
+      "loss": 3.2722,
+      "step": 41450
+    },
+    {
+      "epoch": 12.08503698526414,
+      "grad_norm": 0.35828056931495667,
+      "learning_rate": 0.0004552482517482517,
+      "loss": 3.2589,
+      "step": 41500
+    },
+    {
+      "epoch": 12.099598112877862,
+      "grad_norm": 0.3321901261806488,
+      "learning_rate": 0.00045507342657342655,
+      "loss": 3.276,
+      "step": 41550
+    },
+    {
+      "epoch": 12.114159240491583,
+      "grad_norm": 0.3395446538925171,
+      "learning_rate": 0.00045489860139860135,
+      "loss": 3.2639,
+      "step": 41600
+    },
+    {
+      "epoch": 12.128720368105306,
+      "grad_norm": 0.3980446457862854,
+      "learning_rate": 0.0004547237762237762,
+      "loss": 3.2869,
+      "step": 41650
+    },
+    {
+      "epoch": 12.143281495719028,
+      "grad_norm": 0.36757737398147583,
+      "learning_rate": 0.00045454895104895106,
+      "loss": 3.2742,
+      "step": 41700
+    },
+    {
+      "epoch": 12.157842623332751,
+      "grad_norm": 0.34751492738723755,
+      "learning_rate": 0.0004543741258741258,
+      "loss": 3.2939,
+      "step": 41750
+    },
+    {
+      "epoch": 12.172403750946474,
+      "grad_norm": 0.3655042350292206,
+      "learning_rate": 0.00045419930069930066,
+      "loss": 3.2931,
+      "step": 41800
+    },
+    {
+      "epoch": 12.186964878560195,
+      "grad_norm": 0.35502806305885315,
+      "learning_rate": 0.00045402447552447546,
+      "loss": 3.2992,
+      "step": 41850
+    },
+    {
+      "epoch": 12.201526006173918,
+      "grad_norm": 0.35130417346954346,
+      "learning_rate": 0.0004538496503496503,
+      "loss": 3.3008,
+      "step": 41900
+    },
+    {
+      "epoch": 12.21608713378764,
+      "grad_norm": 0.39589038491249084,
+      "learning_rate": 0.0004536748251748251,
+      "loss": 3.291,
+      "step": 41950
+    },
+    {
+      "epoch": 12.230648261401363,
+      "grad_norm": 0.37069767713546753,
+      "learning_rate": 0.00045349999999999996,
+      "loss": 3.3046,
+      "step": 42000
+    },
+    {
+      "epoch": 12.230648261401363,
+      "eval_accuracy": 0.36991568174673584,
+      "eval_loss": 3.559845209121704,
+      "eval_runtime": 181.1401,
+      "eval_samples_per_second": 91.896,
+      "eval_steps_per_second": 5.747,
+      "step": 42000
+    },
+    {
+      "epoch": 12.245209389015086,
+      "grad_norm": 0.34463024139404297,
+      "learning_rate": 0.00045332517482517476,
+      "loss": 3.2938,
+      "step": 42050
+    },
+    {
+      "epoch": 12.259770516628807,
+      "grad_norm": 0.3618188202381134,
+      "learning_rate": 0.0004531503496503496,
+      "loss": 3.313,
+      "step": 42100
+    },
+    {
+      "epoch": 12.27433164424253,
+      "grad_norm": 0.3733763098716736,
+      "learning_rate": 0.0004529755244755244,
+      "loss": 3.3022,
+      "step": 42150
+    },
+    {
+      "epoch": 12.288892771856252,
+      "grad_norm": 0.3802759647369385,
+      "learning_rate": 0.00045280069930069927,
+      "loss": 3.2983,
+      "step": 42200
+    },
+    {
+      "epoch": 12.303453899469975,
+      "grad_norm": 0.3399848937988281,
+      "learning_rate": 0.00045262587412587407,
+      "loss": 3.302,
+      "step": 42250
+    },
+    {
+      "epoch": 12.318015027083698,
+      "grad_norm": 0.37302467226982117,
+      "learning_rate": 0.0004524510489510489,
+      "loss": 3.3111,
+      "step": 42300
+    },
+    {
+      "epoch": 12.33257615469742,
+      "grad_norm": 0.38620615005493164,
+      "learning_rate": 0.0004522762237762238,
+      "loss": 3.3157,
+      "step": 42350
+    },
+    {
+      "epoch": 12.347137282311142,
+      "grad_norm": 0.3746669292449951,
+      "learning_rate": 0.0004521013986013986,
+      "loss": 3.3092,
+      "step": 42400
+    },
+    {
+      "epoch": 12.361698409924864,
+      "grad_norm": 0.35703712701797485,
+      "learning_rate": 0.00045192657342657343,
+      "loss": 3.3064,
+      "step": 42450
+    },
+    {
+      "epoch": 12.376259537538587,
+      "grad_norm": 0.3509250581264496,
+      "learning_rate": 0.0004517517482517482,
+      "loss": 3.3034,
+      "step": 42500
+    },
+    {
+      "epoch": 12.39082066515231,
+      "grad_norm": 0.3635379672050476,
+      "learning_rate": 0.00045157692307692303,
+      "loss": 3.3163,
+      "step": 42550
+    },
+    {
+      "epoch": 12.405381792766033,
+      "grad_norm": 0.34037286043167114,
+      "learning_rate": 0.00045140209790209783,
+      "loss": 3.3285,
+      "step": 42600
+    },
+    {
+      "epoch": 12.419942920379754,
+      "grad_norm": 0.3448360860347748,
+      "learning_rate": 0.0004512272727272727,
+      "loss": 3.3246,
+      "step": 42650
+    },
+    {
+      "epoch": 12.434504047993476,
+      "grad_norm": 0.35941165685653687,
+      "learning_rate": 0.0004510524475524475,
+      "loss": 3.3246,
+      "step": 42700
+    },
+    {
+      "epoch": 12.449065175607199,
+      "grad_norm": 0.3781452178955078,
+      "learning_rate": 0.00045087762237762234,
+      "loss": 3.3292,
+      "step": 42750
+    },
+    {
+      "epoch": 12.463626303220922,
+      "grad_norm": 0.36656999588012695,
+      "learning_rate": 0.00045070279720279714,
+      "loss": 3.3335,
+      "step": 42800
+    },
+    {
+      "epoch": 12.478187430834645,
+      "grad_norm": 0.3548363745212555,
+      "learning_rate": 0.000450527972027972,
+      "loss": 3.323,
+      "step": 42850
+    },
+    {
+      "epoch": 12.492748558448366,
+      "grad_norm": 0.3620319366455078,
+      "learning_rate": 0.0004503531468531468,
+      "loss": 3.3329,
+      "step": 42900
+    },
+    {
+      "epoch": 12.507309686062088,
+      "grad_norm": 0.353424072265625,
+      "learning_rate": 0.00045017832167832165,
+      "loss": 3.3285,
+      "step": 42950
+    },
+    {
+      "epoch": 12.521870813675811,
+      "grad_norm": 0.3987106382846832,
+      "learning_rate": 0.0004500034965034965,
+      "loss": 3.3397,
+      "step": 43000
+    },
+    {
+      "epoch": 12.521870813675811,
+      "eval_accuracy": 0.37050572847316976,
+      "eval_loss": 3.554832935333252,
+      "eval_runtime": 181.1131,
+      "eval_samples_per_second": 91.909,
+      "eval_steps_per_second": 5.748,
+      "step": 43000
+    },
+    {
+      "epoch": 12.536431941289534,
+      "grad_norm": 0.35408225655555725,
+      "learning_rate": 0.0004498286713286713,
+      "loss": 3.3272,
+      "step": 43050
+    },
+    {
+      "epoch": 12.550993068903256,
+      "grad_norm": 0.36602988839149475,
+      "learning_rate": 0.00044965384615384615,
+      "loss": 3.3313,
+      "step": 43100
+    },
+    {
+      "epoch": 12.565554196516977,
+      "grad_norm": 0.34285175800323486,
+      "learning_rate": 0.00044947902097902095,
+      "loss": 3.337,
+      "step": 43150
+    },
+    {
+      "epoch": 12.5801153241307,
+      "grad_norm": 0.32985830307006836,
+      "learning_rate": 0.0004493041958041958,
+      "loss": 3.3361,
+      "step": 43200
+    },
+    {
+      "epoch": 12.594676451744423,
+      "grad_norm": 0.3447939455509186,
+      "learning_rate": 0.00044912937062937055,
+      "loss": 3.3375,
+      "step": 43250
+    },
+    {
+      "epoch": 12.609237579358146,
+      "grad_norm": 0.3655174970626831,
+      "learning_rate": 0.0004489545454545454,
+      "loss": 3.3341,
+      "step": 43300
+    },
+    {
+      "epoch": 12.623798706971868,
+      "grad_norm": 0.36965784430503845,
+      "learning_rate": 0.0004487797202797202,
+      "loss": 3.3322,
+      "step": 43350
+    },
+    {
+      "epoch": 12.63835983458559,
+      "grad_norm": 0.3464057147502899,
+      "learning_rate": 0.00044860489510489506,
+      "loss": 3.3377,
+      "step": 43400
+    },
+    {
+      "epoch": 12.652920962199312,
+      "grad_norm": 0.3564925789833069,
+      "learning_rate": 0.00044843006993006986,
+      "loss": 3.3461,
+      "step": 43450
+    },
+    {
+      "epoch": 12.667482089813035,
+      "grad_norm": 0.3518389165401459,
+      "learning_rate": 0.0004482552447552447,
+      "loss": 3.3374,
+      "step": 43500
+    },
+    {
+      "epoch": 12.682043217426758,
+      "grad_norm": 0.3491464853286743,
+      "learning_rate": 0.0004480804195804195,
+      "loss": 3.3344,
+      "step": 43550
+    },
+    {
+      "epoch": 12.69660434504048,
+      "grad_norm": 0.3686160147190094,
+      "learning_rate": 0.00044790559440559437,
+      "loss": 3.3378,
+      "step": 43600
+    },
+    {
+      "epoch": 12.711165472654203,
+      "grad_norm": 0.36158478260040283,
+      "learning_rate": 0.00044773076923076917,
+      "loss": 3.3407,
+      "step": 43650
+    },
+    {
+      "epoch": 12.725726600267924,
+      "grad_norm": 0.34943732619285583,
+      "learning_rate": 0.000447555944055944,
+      "loss": 3.3453,
+      "step": 43700
+    },
+    {
+      "epoch": 12.740287727881647,
+      "grad_norm": 0.35694846510887146,
+      "learning_rate": 0.0004473811188811189,
+      "loss": 3.3455,
+      "step": 43750
+    },
+    {
+      "epoch": 12.75484885549537,
+      "grad_norm": 0.33400988578796387,
+      "learning_rate": 0.0004472062937062937,
+      "loss": 3.3478,
+      "step": 43800
+    },
+    {
+      "epoch": 12.769409983109092,
+      "grad_norm": 0.3382965326309204,
+      "learning_rate": 0.00044703146853146853,
+      "loss": 3.3439,
+      "step": 43850
+    },
+    {
+      "epoch": 12.783971110722815,
+      "grad_norm": 0.35423627495765686,
+      "learning_rate": 0.00044685664335664333,
+      "loss": 3.353,
+      "step": 43900
+    },
+    {
+      "epoch": 12.798532238336536,
+      "grad_norm": 0.38060978055000305,
+      "learning_rate": 0.0004466818181818182,
+      "loss": 3.3379,
+      "step": 43950
+    },
+    {
+      "epoch": 12.813093365950259,
+      "grad_norm": 0.3597307801246643,
+      "learning_rate": 0.00044650699300699293,
+      "loss": 3.3404,
+      "step": 44000
+    },
+    {
+      "epoch": 12.813093365950259,
+      "eval_accuracy": 0.3710835486884363,
+      "eval_loss": 3.544434070587158,
+      "eval_runtime": 181.0811,
+      "eval_samples_per_second": 91.926,
+      "eval_steps_per_second": 5.749,
+      "step": 44000
+    },
+    {
+      "epoch": 12.827654493563982,
+      "grad_norm": 0.35751357674598694,
+      "learning_rate": 0.0004463321678321678,
+      "loss": 3.3484,
+      "step": 44050
+    },
+    {
+      "epoch": 12.842215621177704,
+      "grad_norm": 0.37856724858283997,
+      "learning_rate": 0.0004461573426573426,
+      "loss": 3.3484,
+      "step": 44100
+    },
+    {
+      "epoch": 12.856776748791427,
+      "grad_norm": 0.3474246561527252,
+      "learning_rate": 0.00044598251748251744,
+      "loss": 3.338,
+      "step": 44150
+    },
+    {
+      "epoch": 12.871337876405148,
+      "grad_norm": 0.37576624751091003,
+      "learning_rate": 0.00044580769230769224,
+      "loss": 3.344,
+      "step": 44200
+    },
+    {
+      "epoch": 12.88589900401887,
+      "grad_norm": 0.37470582127571106,
+      "learning_rate": 0.0004456328671328671,
+      "loss": 3.337,
+      "step": 44250
+    },
+    {
+      "epoch": 12.900460131632594,
+      "grad_norm": 0.3437161147594452,
+      "learning_rate": 0.0004454580419580419,
+      "loss": 3.3501,
+      "step": 44300
+    },
+    {
+      "epoch": 12.915021259246316,
+      "grad_norm": 0.37224942445755005,
+      "learning_rate": 0.00044528321678321674,
+      "loss": 3.3511,
+      "step": 44350
+    },
+    {
+      "epoch": 12.929582386860039,
+      "grad_norm": 0.3465173542499542,
+      "learning_rate": 0.0004451083916083916,
+      "loss": 3.3397,
+      "step": 44400
+    },
+    {
+      "epoch": 12.944143514473762,
+      "grad_norm": 0.3810703456401825,
+      "learning_rate": 0.0004449335664335664,
+      "loss": 3.3479,
+      "step": 44450
+    },
+    {
+      "epoch": 12.958704642087483,
+      "grad_norm": 0.36394184827804565,
+      "learning_rate": 0.00044475874125874125,
+      "loss": 3.3696,
+      "step": 44500
+    },
+    {
+      "epoch": 12.973265769701205,
+      "grad_norm": 0.32505106925964355,
+      "learning_rate": 0.00044458391608391605,
+      "loss": 3.341,
+      "step": 44550
+    },
+    {
+      "epoch": 12.987826897314928,
+      "grad_norm": 0.33128371834754944,
+      "learning_rate": 0.0004444090909090909,
+      "loss": 3.347,
+      "step": 44600
+    },
+    {
+      "epoch": 13.002329780418195,
+      "grad_norm": 0.36666157841682434,
+      "learning_rate": 0.0004442342657342657,
+      "loss": 3.3406,
+      "step": 44650
+    },
+    {
+      "epoch": 13.016890908031918,
+      "grad_norm": 0.33953845500946045,
+      "learning_rate": 0.00044405944055944056,
+      "loss": 3.2525,
+      "step": 44700
+    },
+    {
+      "epoch": 13.031452035645641,
+      "grad_norm": 0.38074326515197754,
+      "learning_rate": 0.0004438846153846153,
+      "loss": 3.2372,
+      "step": 44750
+    },
+    {
+      "epoch": 13.046013163259364,
+      "grad_norm": 0.4044519066810608,
+      "learning_rate": 0.00044370979020979016,
+      "loss": 3.2448,
+      "step": 44800
+    },
+    {
+      "epoch": 13.060574290873085,
+      "grad_norm": 0.3917773365974426,
+      "learning_rate": 0.00044353496503496496,
+      "loss": 3.246,
+      "step": 44850
+    },
+    {
+      "epoch": 13.075135418486807,
+      "grad_norm": 0.3733667731285095,
+      "learning_rate": 0.0004433601398601398,
+      "loss": 3.2477,
+      "step": 44900
+    },
+    {
+      "epoch": 13.08969654610053,
+      "grad_norm": 0.35852932929992676,
+      "learning_rate": 0.0004431853146853146,
+      "loss": 3.2545,
+      "step": 44950
+    },
+    {
+      "epoch": 13.104257673714253,
+      "grad_norm": 0.3601677417755127,
+      "learning_rate": 0.00044301048951048946,
+      "loss": 3.2597,
+      "step": 45000
+    },
+    {
+      "epoch": 13.104257673714253,
+      "eval_accuracy": 0.37051289979221985,
+      "eval_loss": 3.557664394378662,
+      "eval_runtime": 181.1955,
+      "eval_samples_per_second": 91.868,
+      "eval_steps_per_second": 5.745,
+      "step": 45000
+    },
+    {
+      "epoch": 13.118818801327976,
+      "grad_norm": 0.3457536995410919,
+      "learning_rate": 0.00044283566433566426,
+      "loss": 3.265,
+      "step": 45050
+    },
+    {
+      "epoch": 13.133379928941697,
+      "grad_norm": 0.3704872131347656,
+      "learning_rate": 0.0004426608391608391,
+      "loss": 3.2645,
+      "step": 45100
+    },
+    {
+      "epoch": 13.14794105655542,
+      "grad_norm": 0.37310388684272766,
+      "learning_rate": 0.00044248601398601397,
+      "loss": 3.2746,
+      "step": 45150
+    },
+    {
+      "epoch": 13.162502184169142,
+      "grad_norm": 0.3809090852737427,
+      "learning_rate": 0.00044231118881118877,
+      "loss": 3.2804,
+      "step": 45200
+    },
+    {
+      "epoch": 13.177063311782865,
+      "grad_norm": 0.36043018102645874,
+      "learning_rate": 0.0004421363636363636,
+      "loss": 3.2662,
+      "step": 45250
+    },
+    {
+      "epoch": 13.191624439396588,
+      "grad_norm": 0.3653396666049957,
+      "learning_rate": 0.0004419615384615384,
+      "loss": 3.265,
+      "step": 45300
+    },
+    {
+      "epoch": 13.206185567010309,
+      "grad_norm": 0.3747062385082245,
+      "learning_rate": 0.0004417867132867133,
+      "loss": 3.2763,
+      "step": 45350
+    },
+    {
+      "epoch": 13.220746694624031,
+      "grad_norm": 0.3532830476760864,
+      "learning_rate": 0.0004416118881118881,
+      "loss": 3.2838,
+      "step": 45400
+    },
+    {
+      "epoch": 13.235307822237754,
+      "grad_norm": 0.3714682459831238,
+      "learning_rate": 0.00044143706293706293,
+      "loss": 3.2829,
+      "step": 45450
+    },
+    {
+      "epoch": 13.249868949851477,
+      "grad_norm": 0.3347223103046417,
+      "learning_rate": 0.0004412622377622377,
+      "loss": 3.2799,
+      "step": 45500
+    },
+    {
+      "epoch": 13.2644300774652,
+      "grad_norm": 0.37033647298812866,
+      "learning_rate": 0.00044108741258741253,
+      "loss": 3.2957,
+      "step": 45550
+    },
+    {
+      "epoch": 13.27899120507892,
+      "grad_norm": 0.39102688431739807,
+      "learning_rate": 0.00044091258741258733,
+      "loss": 3.2914,
+      "step": 45600
+    },
+    {
+      "epoch": 13.293552332692643,
+      "grad_norm": 0.37037813663482666,
+      "learning_rate": 0.0004407377622377622,
+      "loss": 3.2995,
+      "step": 45650
+    },
+    {
+      "epoch": 13.308113460306366,
+      "grad_norm": 0.37284520268440247,
+      "learning_rate": 0.000440562937062937,
+      "loss": 3.2843,
+      "step": 45700
+    },
+    {
+      "epoch": 13.322674587920089,
+      "grad_norm": 0.37667202949523926,
+      "learning_rate": 0.00044038811188811184,
+      "loss": 3.2849,
+      "step": 45750
+    },
+    {
+      "epoch": 13.337235715533811,
+      "grad_norm": 0.3747294843196869,
+      "learning_rate": 0.0004402132867132867,
+      "loss": 3.3116,
+      "step": 45800
+    },
+    {
+      "epoch": 13.351796843147532,
+      "grad_norm": 0.3424604535102844,
+      "learning_rate": 0.0004400384615384615,
+      "loss": 3.2948,
+      "step": 45850
+    },
+    {
+      "epoch": 13.366357970761255,
+      "grad_norm": 0.36772021651268005,
+      "learning_rate": 0.00043986363636363635,
+      "loss": 3.2919,
+      "step": 45900
+    },
+    {
+      "epoch": 13.380919098374978,
+      "grad_norm": 0.3693206310272217,
+      "learning_rate": 0.00043968881118881115,
+      "loss": 3.3231,
+      "step": 45950
+    },
+    {
+      "epoch": 13.3954802259887,
+      "grad_norm": 0.385343074798584,
+      "learning_rate": 0.000439513986013986,
+      "loss": 3.3136,
+      "step": 46000
+    },
+    {
+      "epoch": 13.3954802259887,
+      "eval_accuracy": 0.3708942728905565,
+      "eval_loss": 3.553351402282715,
+      "eval_runtime": 181.4338,
+      "eval_samples_per_second": 91.747,
+      "eval_steps_per_second": 5.738,
+      "step": 46000
+    },
+    {
+      "epoch": 13.410041353602423,
+      "grad_norm": 0.3651343584060669,
+      "learning_rate": 0.0004393391608391608,
+      "loss": 3.3064,
+      "step": 46050
+    },
+    {
+      "epoch": 13.424602481216146,
+      "grad_norm": 0.3632429540157318,
+      "learning_rate": 0.00043916433566433565,
+      "loss": 3.3071,
+      "step": 46100
+    },
+    {
+      "epoch": 13.439163608829867,
+      "grad_norm": 0.3616783320903778,
+      "learning_rate": 0.00043898951048951045,
+      "loss": 3.3066,
+      "step": 46150
+    },
+    {
+      "epoch": 13.45372473644359,
+      "grad_norm": 0.36047831177711487,
+      "learning_rate": 0.0004388146853146853,
+      "loss": 3.3063,
+      "step": 46200
+    },
+    {
+      "epoch": 13.468285864057313,
+      "grad_norm": 0.38732126355171204,
+      "learning_rate": 0.00043863986013986005,
+      "loss": 3.3196,
+      "step": 46250
+    },
+    {
+      "epoch": 13.482846991671035,
+      "grad_norm": 0.3714425563812256,
+      "learning_rate": 0.0004384650349650349,
+      "loss": 3.3089,
+      "step": 46300
+    },
+    {
+      "epoch": 13.497408119284758,
+      "grad_norm": 0.3905268609523773,
+      "learning_rate": 0.0004382902097902097,
+      "loss": 3.3034,
+      "step": 46350
+    },
+    {
+      "epoch": 13.51196924689848,
+      "grad_norm": 0.36184632778167725,
+      "learning_rate": 0.00043811538461538456,
+      "loss": 3.3161,
+      "step": 46400
+    },
+    {
+      "epoch": 13.526530374512202,
+      "grad_norm": 0.37481918931007385,
+      "learning_rate": 0.0004379405594405594,
+      "loss": 3.3152,
+      "step": 46450
+    },
+    {
+      "epoch": 13.541091502125925,
+      "grad_norm": 0.35389721393585205,
+      "learning_rate": 0.0004377657342657342,
+      "loss": 3.3079,
+      "step": 46500
+    },
+    {
+      "epoch": 13.555652629739647,
+      "grad_norm": 0.3536003530025482,
+      "learning_rate": 0.00043759090909090907,
+      "loss": 3.3226,
+      "step": 46550
+    },
+    {
+      "epoch": 13.57021375735337,
+      "grad_norm": 0.4076172709465027,
+      "learning_rate": 0.00043741608391608387,
+      "loss": 3.3193,
+      "step": 46600
+    },
+    {
+      "epoch": 13.584774884967091,
+      "grad_norm": 0.38209009170532227,
+      "learning_rate": 0.0004372412587412587,
+      "loss": 3.3166,
+      "step": 46650
+    },
+    {
+      "epoch": 13.599336012580814,
+      "grad_norm": 0.37584787607192993,
+      "learning_rate": 0.0004370664335664335,
+      "loss": 3.3176,
+      "step": 46700
+    },
+    {
+      "epoch": 13.613897140194537,
+      "grad_norm": 0.3730042278766632,
+      "learning_rate": 0.0004368916083916084,
+      "loss": 3.329,
+      "step": 46750
+    },
+    {
+      "epoch": 13.62845826780826,
+      "grad_norm": 0.37600407004356384,
+      "learning_rate": 0.0004367167832167832,
+      "loss": 3.3182,
+      "step": 46800
+    },
+    {
+      "epoch": 13.643019395421982,
+      "grad_norm": 0.3546999990940094,
+      "learning_rate": 0.00043654195804195803,
+      "loss": 3.3099,
+      "step": 46850
+    },
+    {
+      "epoch": 13.657580523035705,
+      "grad_norm": 0.41130438446998596,
+      "learning_rate": 0.00043636713286713283,
+      "loss": 3.3174,
+      "step": 46900
+    },
+    {
+      "epoch": 13.672141650649426,
+      "grad_norm": 0.380054235458374,
+      "learning_rate": 0.0004361923076923077,
+      "loss": 3.3115,
+      "step": 46950
+    },
+    {
+      "epoch": 13.686702778263149,
+      "grad_norm": 0.3687915503978729,
+      "learning_rate": 0.00043601748251748243,
+      "loss": 3.326,
+      "step": 47000
+    },
+    {
+      "epoch": 13.686702778263149,
+      "eval_accuracy": 0.3714850249926347,
+      "eval_loss": 3.5448427200317383,
+      "eval_runtime": 181.0835,
+      "eval_samples_per_second": 91.924,
+      "eval_steps_per_second": 5.749,
+      "step": 47000
+    },
+    {
+      "epoch": 13.701263905876871,
+      "grad_norm": 0.3640954792499542,
+      "learning_rate": 0.00043584265734265734,
+      "loss": 3.3238,
+      "step": 47050
+    },
+    {
+      "epoch": 13.715825033490594,
+      "grad_norm": 0.33664220571517944,
+      "learning_rate": 0.0004356678321678321,
+      "loss": 3.3274,
+      "step": 47100
+    },
+    {
+      "epoch": 13.730386161104317,
+      "grad_norm": 0.36196938157081604,
+      "learning_rate": 0.00043549300699300694,
+      "loss": 3.3151,
+      "step": 47150
+    },
+    {
+      "epoch": 13.744947288718038,
+      "grad_norm": 0.35833385586738586,
+      "learning_rate": 0.0004353181818181818,
+      "loss": 3.3317,
+      "step": 47200
+    },
+    {
+      "epoch": 13.75950841633176,
+      "grad_norm": 0.367707222700119,
+      "learning_rate": 0.0004351433566433566,
+      "loss": 3.3222,
+      "step": 47250
+    },
+    {
+      "epoch": 13.774069543945483,
+      "grad_norm": 0.34781572222709656,
+      "learning_rate": 0.00043496853146853144,
+      "loss": 3.3146,
+      "step": 47300
+    },
+    {
+      "epoch": 13.788630671559206,
+      "grad_norm": 0.37640494108200073,
+      "learning_rate": 0.00043479370629370624,
+      "loss": 3.3311,
+      "step": 47350
+    },
+    {
+      "epoch": 13.803191799172929,
+      "grad_norm": 0.3690575361251831,
+      "learning_rate": 0.0004346188811188811,
+      "loss": 3.328,
+      "step": 47400
+    },
+    {
+      "epoch": 13.81775292678665,
+      "grad_norm": 0.3566841185092926,
+      "learning_rate": 0.0004344440559440559,
+      "loss": 3.3264,
+      "step": 47450
+    },
+    {
+      "epoch": 13.832314054400372,
+      "grad_norm": 0.3551981449127197,
+      "learning_rate": 0.00043426923076923075,
+      "loss": 3.3376,
+      "step": 47500
+    },
+    {
+      "epoch": 13.846875182014095,
+      "grad_norm": 0.3538823425769806,
+      "learning_rate": 0.00043409440559440555,
+      "loss": 3.3526,
+      "step": 47550
+    },
+    {
+      "epoch": 13.861436309627818,
+      "grad_norm": 0.374664306640625,
+      "learning_rate": 0.0004339195804195804,
+      "loss": 3.3266,
+      "step": 47600
+    },
+    {
+      "epoch": 13.87599743724154,
+      "grad_norm": 0.36062973737716675,
+      "learning_rate": 0.0004337447552447552,
+      "loss": 3.3177,
+      "step": 47650
+    },
+    {
+      "epoch": 13.890558564855262,
+      "grad_norm": 0.34848466515541077,
+      "learning_rate": 0.00043356993006993006,
+      "loss": 3.3315,
+      "step": 47700
+    },
+    {
+      "epoch": 13.905119692468984,
+      "grad_norm": 0.3789747953414917,
+      "learning_rate": 0.0004333951048951048,
+      "loss": 3.3319,
+      "step": 47750
+    },
+    {
+      "epoch": 13.919680820082707,
+      "grad_norm": 0.3523997366428375,
+      "learning_rate": 0.0004332202797202797,
+      "loss": 3.3265,
+      "step": 47800
+    },
+    {
+      "epoch": 13.93424194769643,
+      "grad_norm": 0.38667649030685425,
+      "learning_rate": 0.00043304545454545456,
+      "loss": 3.3346,
+      "step": 47850
+    },
+    {
+      "epoch": 13.948803075310153,
+      "grad_norm": 0.3615758717060089,
+      "learning_rate": 0.0004328706293706293,
+      "loss": 3.3312,
+      "step": 47900
+    },
+    {
+      "epoch": 13.963364202923874,
+      "grad_norm": 0.3342243731021881,
+      "learning_rate": 0.00043269580419580416,
+      "loss": 3.3312,
+      "step": 47950
+    },
+    {
+      "epoch": 13.977925330537596,
+      "grad_norm": 0.34050750732421875,
+      "learning_rate": 0.00043252097902097896,
+      "loss": 3.3423,
+      "step": 48000
+    },
+    {
+      "epoch": 13.977925330537596,
+      "eval_accuracy": 0.3718588740840991,
+      "eval_loss": 3.540470838546753,
+      "eval_runtime": 181.0794,
+      "eval_samples_per_second": 91.927,
+      "eval_steps_per_second": 5.749,
+      "step": 48000
+    },
+    {
+      "epoch": 13.992486458151319,
+      "grad_norm": 0.3358824551105499,
+      "learning_rate": 0.0004323461538461538,
+      "loss": 3.3435,
+      "step": 48050
+    },
+    {
+      "epoch": 14.006989341254586,
+      "grad_norm": 0.3420553505420685,
+      "learning_rate": 0.0004321713286713286,
+      "loss": 3.277,
+      "step": 48100
+    },
+    {
+      "epoch": 14.021550468868309,
+      "grad_norm": 0.3606970012187958,
+      "learning_rate": 0.00043199650349650347,
+      "loss": 3.2118,
+      "step": 48150
+    },
+    {
+      "epoch": 14.036111596482032,
+      "grad_norm": 0.4001006484031677,
+      "learning_rate": 0.00043182167832167827,
+      "loss": 3.2243,
+      "step": 48200
+    },
+    {
+      "epoch": 14.050672724095755,
+      "grad_norm": 0.3614968955516815,
+      "learning_rate": 0.0004316468531468531,
+      "loss": 3.2362,
+      "step": 48250
+    },
+    {
+      "epoch": 14.065233851709475,
+      "grad_norm": 0.360830694437027,
+      "learning_rate": 0.0004314720279720279,
+      "loss": 3.2397,
+      "step": 48300
+    },
+    {
+      "epoch": 14.079794979323198,
+      "grad_norm": 0.3907882869243622,
+      "learning_rate": 0.0004312972027972028,
+      "loss": 3.2394,
+      "step": 48350
+    },
+    {
+      "epoch": 14.094356106936921,
+      "grad_norm": 0.35445448756217957,
+      "learning_rate": 0.0004311223776223776,
+      "loss": 3.2431,
+      "step": 48400
+    },
+    {
+      "epoch": 14.108917234550644,
+      "grad_norm": 0.3781708776950836,
+      "learning_rate": 0.00043094755244755243,
+      "loss": 3.2474,
+      "step": 48450
+    },
+    {
+      "epoch": 14.123478362164366,
+      "grad_norm": 0.34598371386528015,
+      "learning_rate": 0.0004307727272727272,
+      "loss": 3.2527,
+      "step": 48500
+    },
+    {
+      "epoch": 14.13803948977809,
+      "grad_norm": 0.36756083369255066,
+      "learning_rate": 0.0004305979020979021,
+      "loss": 3.2582,
+      "step": 48550
+    },
+    {
+      "epoch": 14.15260061739181,
+      "grad_norm": 0.383299857378006,
+      "learning_rate": 0.00043042307692307694,
+      "loss": 3.2613,
+      "step": 48600
+    },
+    {
+      "epoch": 14.167161745005533,
+      "grad_norm": 0.37315934896469116,
+      "learning_rate": 0.0004302482517482517,
+      "loss": 3.2581,
+      "step": 48650
+    },
+    {
+      "epoch": 14.181722872619256,
+      "grad_norm": 0.3663264214992523,
+      "learning_rate": 0.00043007342657342654,
+      "loss": 3.2604,
+      "step": 48700
+    },
+    {
+      "epoch": 14.196284000232978,
+      "grad_norm": 0.368505597114563,
+      "learning_rate": 0.00042989860139860134,
+      "loss": 3.263,
+      "step": 48750
+    },
+    {
+      "epoch": 14.210845127846701,
+      "grad_norm": 0.3589208126068115,
+      "learning_rate": 0.0004297237762237762,
+      "loss": 3.25,
+      "step": 48800
+    },
+    {
+      "epoch": 14.225406255460422,
+      "grad_norm": 0.35528650879859924,
+      "learning_rate": 0.000429548951048951,
+      "loss": 3.2608,
+      "step": 48850
+    },
+    {
+      "epoch": 14.239967383074145,
+      "grad_norm": 0.3896448314189911,
+      "learning_rate": 0.00042937412587412585,
+      "loss": 3.2699,
+      "step": 48900
+    },
+    {
+      "epoch": 14.254528510687868,
+      "grad_norm": 0.389419823884964,
+      "learning_rate": 0.00042919930069930065,
+      "loss": 3.2815,
+      "step": 48950
+    },
+    {
+      "epoch": 14.26908963830159,
+      "grad_norm": 0.36435467004776,
+      "learning_rate": 0.0004290244755244755,
+      "loss": 3.2747,
+      "step": 49000
+    },
+    {
+      "epoch": 14.26908963830159,
+      "eval_accuracy": 0.371175952897836,
+      "eval_loss": 3.5533642768859863,
+      "eval_runtime": 188.2618,
+      "eval_samples_per_second": 88.419,
+      "eval_steps_per_second": 5.53,
+      "step": 49000
+    },
+    {
+      "epoch": 14.283650765915313,
+      "grad_norm": 0.3509522080421448,
+      "learning_rate": 0.0004288496503496503,
+      "loss": 3.28,
+      "step": 49050
+    },
+    {
+      "epoch": 14.298211893529034,
+      "grad_norm": 0.334859162569046,
+      "learning_rate": 0.00042867482517482515,
+      "loss": 3.2796,
+      "step": 49100
+    },
+    {
+      "epoch": 14.312773021142757,
+      "grad_norm": 0.3892660439014435,
+      "learning_rate": 0.00042849999999999995,
+      "loss": 3.2847,
+      "step": 49150
+    },
+    {
+      "epoch": 14.32733414875648,
+      "grad_norm": 0.35437408089637756,
+      "learning_rate": 0.0004283251748251748,
+      "loss": 3.2628,
+      "step": 49200
+    },
+    {
+      "epoch": 14.341895276370202,
+      "grad_norm": 0.37474045157432556,
+      "learning_rate": 0.00042815034965034966,
+      "loss": 3.2988,
+      "step": 49250
+    },
+    {
+      "epoch": 14.356456403983925,
+      "grad_norm": 0.4226810336112976,
+      "learning_rate": 0.00042797552447552446,
+      "loss": 3.2899,
+      "step": 49300
+    },
+    {
+      "epoch": 14.371017531597648,
+      "grad_norm": 0.3750922977924347,
+      "learning_rate": 0.0004278006993006993,
+      "loss": 3.2828,
+      "step": 49350
+    },
+    {
+      "epoch": 14.385578659211369,
+      "grad_norm": 0.35923951864242554,
+      "learning_rate": 0.00042762587412587406,
+      "loss": 3.286,
+      "step": 49400
+    },
+    {
+      "epoch": 14.400139786825092,
+      "grad_norm": 0.35671406984329224,
+      "learning_rate": 0.0004274510489510489,
+      "loss": 3.2892,
+      "step": 49450
+    },
+    {
+      "epoch": 14.414700914438814,
+      "grad_norm": 0.35129812359809875,
+      "learning_rate": 0.0004272762237762237,
+      "loss": 3.2945,
+      "step": 49500
+    },
+    {
+      "epoch": 14.429262042052537,
+      "grad_norm": 0.37475308775901794,
+      "learning_rate": 0.00042710139860139857,
+      "loss": 3.2978,
+      "step": 49550
+    },
+    {
+      "epoch": 14.44382316966626,
+      "grad_norm": 0.38268592953681946,
+      "learning_rate": 0.00042692657342657337,
+      "loss": 3.2959,
+      "step": 49600
+    },
+    {
+      "epoch": 14.45838429727998,
+      "grad_norm": 0.3419938385486603,
+      "learning_rate": 0.0004267517482517482,
+      "loss": 3.2877,
+      "step": 49650
+    },
+    {
+      "epoch": 14.472945424893704,
+      "grad_norm": 0.35204702615737915,
+      "learning_rate": 0.000426576923076923,
+      "loss": 3.2877,
+      "step": 49700
+    },
+    {
+      "epoch": 14.487506552507426,
+      "grad_norm": 0.36226049065589905,
+      "learning_rate": 0.0004264020979020979,
+      "loss": 3.2871,
+      "step": 49750
+    },
+    {
+      "epoch": 14.502067680121149,
+      "grad_norm": 0.382007896900177,
+      "learning_rate": 0.0004262272727272727,
+      "loss": 3.2989,
+      "step": 49800
+    },
+    {
+      "epoch": 14.516628807734872,
+      "grad_norm": 0.34697645902633667,
+      "learning_rate": 0.00042605244755244753,
+      "loss": 3.2957,
+      "step": 49850
+    },
+    {
+      "epoch": 14.531189935348593,
+      "grad_norm": 0.3803808391094208,
+      "learning_rate": 0.00042587762237762233,
+      "loss": 3.302,
+      "step": 49900
+    },
+    {
+      "epoch": 14.545751062962315,
+      "grad_norm": 0.3486500084400177,
+      "learning_rate": 0.0004257027972027972,
+      "loss": 3.3026,
+      "step": 49950
+    },
+    {
+      "epoch": 14.560312190576038,
+      "grad_norm": 0.3419564366340637,
+      "learning_rate": 0.00042552797202797204,
+      "loss": 3.287,
+      "step": 50000
+    },
+    {
+      "epoch": 14.560312190576038,
+      "eval_accuracy": 0.37168276529824573,
+      "eval_loss": 3.546523094177246,
+      "eval_runtime": 181.058,
+      "eval_samples_per_second": 91.937,
+      "eval_steps_per_second": 5.75,
+      "step": 50000
+    },
+    {
+      "epoch": 14.574873318189761,
+      "grad_norm": 0.3464997410774231,
+      "learning_rate": 0.00042535314685314684,
+      "loss": 3.2981,
+      "step": 50050
+    },
+    {
+      "epoch": 14.589434445803484,
+      "grad_norm": 0.34635862708091736,
+      "learning_rate": 0.0004251783216783217,
+      "loss": 3.3036,
+      "step": 50100
+    },
+    {
+      "epoch": 14.603995573417205,
+      "grad_norm": 0.3743698298931122,
+      "learning_rate": 0.00042500349650349643,
+      "loss": 3.3031,
+      "step": 50150
+    },
+    {
+      "epoch": 14.618556701030927,
+      "grad_norm": 0.3436216711997986,
+      "learning_rate": 0.0004248286713286713,
+      "loss": 3.3117,
+      "step": 50200
+    },
+    {
+      "epoch": 14.63311782864465,
+      "grad_norm": 0.3991907238960266,
+      "learning_rate": 0.0004246538461538461,
+      "loss": 3.3056,
+      "step": 50250
+    },
+    {
+      "epoch": 14.647678956258373,
+      "grad_norm": 0.39424341917037964,
+      "learning_rate": 0.00042447902097902094,
+      "loss": 3.3109,
+      "step": 50300
+    },
+    {
+      "epoch": 14.662240083872096,
+      "grad_norm": 0.3627629280090332,
+      "learning_rate": 0.00042430419580419574,
+      "loss": 3.3071,
+      "step": 50350
+    },
+    {
+      "epoch": 14.676801211485817,
+      "grad_norm": 0.3603801727294922,
+      "learning_rate": 0.0004241293706293706,
+      "loss": 3.2999,
+      "step": 50400
+    },
+    {
+      "epoch": 14.69136233909954,
+      "grad_norm": 0.3873232305049896,
+      "learning_rate": 0.0004239545454545454,
+      "loss": 3.3156,
+      "step": 50450
+    },
+    {
+      "epoch": 14.705923466713262,
+      "grad_norm": 0.35634586215019226,
+      "learning_rate": 0.00042377972027972025,
+      "loss": 3.3007,
+      "step": 50500
+    },
+    {
+      "epoch": 14.720484594326985,
+      "grad_norm": 0.383419007062912,
+      "learning_rate": 0.00042360489510489505,
+      "loss": 3.3109,
+      "step": 50550
+    },
+    {
+      "epoch": 14.735045721940708,
+      "grad_norm": 0.3834097981452942,
+      "learning_rate": 0.0004234300699300699,
+      "loss": 3.3024,
+      "step": 50600
+    },
+    {
+      "epoch": 14.749606849554429,
+      "grad_norm": 0.3689540922641754,
+      "learning_rate": 0.00042325524475524476,
+      "loss": 3.3118,
+      "step": 50650
+    },
+    {
+      "epoch": 14.764167977168151,
+      "grad_norm": 0.3960738778114319,
+      "learning_rate": 0.00042308041958041956,
+      "loss": 3.3045,
+      "step": 50700
+    },
+    {
+      "epoch": 14.778729104781874,
+      "grad_norm": 0.3666040301322937,
+      "learning_rate": 0.0004229055944055944,
+      "loss": 3.3053,
+      "step": 50750
+    },
+    {
+      "epoch": 14.793290232395597,
+      "grad_norm": 0.3593136668205261,
+      "learning_rate": 0.0004227307692307692,
+      "loss": 3.3182,
+      "step": 50800
+    },
+    {
+      "epoch": 14.80785136000932,
+      "grad_norm": 0.36340007185935974,
+      "learning_rate": 0.00042255594405594406,
+      "loss": 3.3169,
+      "step": 50850
+    },
+    {
+      "epoch": 14.822412487623042,
+      "grad_norm": 0.3798671066761017,
+      "learning_rate": 0.0004223811188811188,
+      "loss": 3.3162,
+      "step": 50900
+    },
+    {
+      "epoch": 14.836973615236763,
+      "grad_norm": 0.3715970814228058,
+      "learning_rate": 0.00042220629370629366,
+      "loss": 3.3201,
+      "step": 50950
+    },
+    {
+      "epoch": 14.851534742850486,
+      "grad_norm": 0.39670634269714355,
+      "learning_rate": 0.00042203146853146846,
+      "loss": 3.3153,
+      "step": 51000
+    },
+    {
+      "epoch": 14.851534742850486,
+      "eval_accuracy": 0.37229244498011194,
+      "eval_loss": 3.5407721996307373,
+      "eval_runtime": 230.4734,
+      "eval_samples_per_second": 72.225,
+      "eval_steps_per_second": 4.517,
+      "step": 51000
+    },
+    {
+      "epoch": 14.866095870464209,
+      "grad_norm": 0.3576284945011139,
+      "learning_rate": 0.0004218566433566433,
+      "loss": 3.3094,
+      "step": 51050
+    },
+    {
+      "epoch": 14.880656998077932,
+      "grad_norm": 0.36275455355644226,
+      "learning_rate": 0.0004216818181818181,
+      "loss": 3.3142,
+      "step": 51100
+    },
+    {
+      "epoch": 14.895218125691654,
+      "grad_norm": 0.376268595457077,
+      "learning_rate": 0.00042150699300699297,
+      "loss": 3.3316,
+      "step": 51150
+    },
+    {
+      "epoch": 14.909779253305375,
+      "grad_norm": 0.3608756363391876,
+      "learning_rate": 0.00042133216783216777,
+      "loss": 3.3067,
+      "step": 51200
+    },
+    {
+      "epoch": 14.924340380919098,
+      "grad_norm": 0.41920217871665955,
+      "learning_rate": 0.0004211573426573426,
+      "loss": 3.3175,
+      "step": 51250
+    },
+    {
+      "epoch": 14.93890150853282,
+      "grad_norm": 0.35982200503349304,
+      "learning_rate": 0.0004209825174825175,
+      "loss": 3.3127,
+      "step": 51300
+    },
+    {
+      "epoch": 14.953462636146543,
+      "grad_norm": 0.3523630201816559,
+      "learning_rate": 0.0004208076923076923,
+      "loss": 3.3256,
+      "step": 51350
+    },
+    {
+      "epoch": 14.968023763760266,
+      "grad_norm": 0.358729749917984,
+      "learning_rate": 0.00042063286713286713,
+      "loss": 3.3293,
+      "step": 51400
+    },
+    {
+      "epoch": 14.982584891373987,
+      "grad_norm": 0.3890109658241272,
+      "learning_rate": 0.00042045804195804193,
+      "loss": 3.3267,
+      "step": 51450
+    },
+    {
+      "epoch": 14.99714601898771,
+      "grad_norm": 0.335075318813324,
+      "learning_rate": 0.0004202832167832168,
+      "loss": 3.3264,
+      "step": 51500
+    },
+    {
+      "epoch": 15.011648902090977,
+      "grad_norm": 0.40458938479423523,
+      "learning_rate": 0.0004201083916083916,
+      "loss": 3.2301,
+      "step": 51550
+    },
+    {
+      "epoch": 15.0262100297047,
+      "grad_norm": 0.38281872868537903,
+      "learning_rate": 0.00041993356643356644,
+      "loss": 3.2213,
+      "step": 51600
+    },
+    {
+      "epoch": 15.040771157318423,
+      "grad_norm": 0.3873460590839386,
+      "learning_rate": 0.0004197587412587412,
+      "loss": 3.2218,
+      "step": 51650
+    },
+    {
+      "epoch": 15.055332284932145,
+      "grad_norm": 0.3797636330127716,
+      "learning_rate": 0.00041958391608391604,
+      "loss": 3.2326,
+      "step": 51700
+    },
+    {
+      "epoch": 15.069893412545868,
+      "grad_norm": 0.3885267674922943,
+      "learning_rate": 0.00041940909090909084,
+      "loss": 3.2326,
+      "step": 51750
+    },
+    {
+      "epoch": 15.084454540159589,
+      "grad_norm": 0.3854827284812927,
+      "learning_rate": 0.0004192342657342657,
+      "loss": 3.2266,
+      "step": 51800
+    },
+    {
+      "epoch": 15.099015667773312,
+      "grad_norm": 0.35731711983680725,
+      "learning_rate": 0.0004190594405594405,
+      "loss": 3.2336,
+      "step": 51850
+    },
+    {
+      "epoch": 15.113576795387035,
+      "grad_norm": 0.3774850070476532,
+      "learning_rate": 0.00041888461538461535,
+      "loss": 3.236,
+      "step": 51900
+    },
+    {
+      "epoch": 15.128137923000757,
+      "grad_norm": 0.3611111342906952,
+      "learning_rate": 0.00041870979020979015,
+      "loss": 3.2463,
+      "step": 51950
+    },
+    {
+      "epoch": 15.14269905061448,
+      "grad_norm": 0.35102859139442444,
+      "learning_rate": 0.000418534965034965,
+      "loss": 3.2414,
+      "step": 52000
+    },
+    {
+      "epoch": 15.14269905061448,
+      "eval_accuracy": 0.3716459682021362,
+      "eval_loss": 3.556779384613037,
+      "eval_runtime": 181.0814,
+      "eval_samples_per_second": 91.926,
+      "eval_steps_per_second": 5.749,
+      "step": 52000
+    },
+    {
+      "epoch": 15.157260178228203,
+      "grad_norm": 0.38801994919776917,
+      "learning_rate": 0.00041836013986013985,
+      "loss": 3.2494,
+      "step": 52050
+    },
+    {
+      "epoch": 15.171821305841924,
+      "grad_norm": 0.37931784987449646,
+      "learning_rate": 0.00041818531468531465,
+      "loss": 3.2417,
+      "step": 52100
+    },
+    {
+      "epoch": 15.186382433455647,
+      "grad_norm": 0.3645617663860321,
+      "learning_rate": 0.0004180104895104895,
+      "loss": 3.2525,
+      "step": 52150
+    },
+    {
+      "epoch": 15.20094356106937,
+      "grad_norm": 0.42287391424179077,
+      "learning_rate": 0.0004178356643356643,
+      "loss": 3.2569,
+      "step": 52200
+    },
+    {
+      "epoch": 15.215504688683092,
+      "grad_norm": 0.390807569026947,
+      "learning_rate": 0.00041766083916083916,
+      "loss": 3.2519,
+      "step": 52250
+    },
+    {
+      "epoch": 15.230065816296815,
+      "grad_norm": 0.38645055890083313,
+      "learning_rate": 0.00041748601398601396,
+      "loss": 3.259,
+      "step": 52300
+    },
+    {
+      "epoch": 15.244626943910536,
+      "grad_norm": 0.3853817284107208,
+      "learning_rate": 0.0004173111888111888,
+      "loss": 3.2575,
+      "step": 52350
+    },
+    {
+      "epoch": 15.259188071524258,
+      "grad_norm": 0.3735731840133667,
+      "learning_rate": 0.00041713636363636356,
+      "loss": 3.2641,
+      "step": 52400
+    },
+    {
+      "epoch": 15.273749199137981,
+      "grad_norm": 0.3664243817329407,
+      "learning_rate": 0.0004169615384615384,
+      "loss": 3.2684,
+      "step": 52450
+    },
+    {
+      "epoch": 15.288310326751704,
+      "grad_norm": 0.36439111828804016,
+      "learning_rate": 0.0004167867132867132,
+      "loss": 3.2573,
+      "step": 52500
+    },
+    {
+      "epoch": 15.302871454365427,
+      "grad_norm": 0.37758076190948486,
+      "learning_rate": 0.00041661188811188807,
+      "loss": 3.2668,
+      "step": 52550
+    },
+    {
+      "epoch": 15.317432581979148,
+      "grad_norm": 0.37438371777534485,
+      "learning_rate": 0.00041643706293706287,
+      "loss": 3.2654,
+      "step": 52600
+    },
+    {
+      "epoch": 15.33199370959287,
+      "grad_norm": 0.3982774019241333,
+      "learning_rate": 0.0004162622377622377,
+      "loss": 3.26,
+      "step": 52650
+    },
+    {
+      "epoch": 15.346554837206593,
+      "grad_norm": 0.3701336979866028,
+      "learning_rate": 0.0004160874125874126,
+      "loss": 3.2697,
+      "step": 52700
+    },
+    {
+      "epoch": 15.361115964820316,
+      "grad_norm": 0.3646653890609741,
+      "learning_rate": 0.0004159125874125874,
+      "loss": 3.2706,
+      "step": 52750
+    },
+    {
+      "epoch": 15.375677092434039,
+      "grad_norm": 0.34608614444732666,
+      "learning_rate": 0.00041573776223776223,
+      "loss": 3.2784,
+      "step": 52800
+    },
+    {
+      "epoch": 15.39023822004776,
+      "grad_norm": 0.3782461881637573,
+      "learning_rate": 0.00041556293706293703,
+      "loss": 3.2585,
+      "step": 52850
+    },
+    {
+      "epoch": 15.404799347661482,
+      "grad_norm": 0.36291107535362244,
+      "learning_rate": 0.0004153881118881119,
+      "loss": 3.269,
+      "step": 52900
+    },
+    {
+      "epoch": 15.419360475275205,
+      "grad_norm": 0.36434653401374817,
+      "learning_rate": 0.0004152132867132867,
+      "loss": 3.2665,
+      "step": 52950
+    },
+    {
+      "epoch": 15.433921602888928,
+      "grad_norm": 0.35678714513778687,
+      "learning_rate": 0.00041503846153846154,
+      "loss": 3.2794,
+      "step": 53000
+    },
+    {
+      "epoch": 15.433921602888928,
+      "eval_accuracy": 0.371921534953832,
+      "eval_loss": 3.546081781387329,
+      "eval_runtime": 181.0963,
+      "eval_samples_per_second": 91.918,
+      "eval_steps_per_second": 5.748,
+      "step": 53000
+    },
+    {
+      "epoch": 15.44848273050265,
+      "grad_norm": 0.39062851667404175,
+      "learning_rate": 0.00041486363636363634,
+      "loss": 3.2884,
+      "step": 53050
+    },
+    {
+      "epoch": 15.463043858116373,
+      "grad_norm": 0.3518424332141876,
+      "learning_rate": 0.0004146888111888112,
+      "loss": 3.2721,
+      "step": 53100
+    },
+    {
+      "epoch": 15.477604985730094,
+      "grad_norm": 0.3690439462661743,
+      "learning_rate": 0.00041451398601398593,
+      "loss": 3.2735,
+      "step": 53150
+    },
+    {
+      "epoch": 15.492166113343817,
+      "grad_norm": 0.3729609251022339,
+      "learning_rate": 0.0004143391608391608,
+      "loss": 3.2771,
+      "step": 53200
+    },
+    {
+      "epoch": 15.50672724095754,
+      "grad_norm": 0.40663325786590576,
+      "learning_rate": 0.0004141643356643356,
+      "loss": 3.2676,
+      "step": 53250
+    },
+    {
+      "epoch": 15.521288368571263,
+      "grad_norm": 0.36350148916244507,
+      "learning_rate": 0.00041398951048951044,
+      "loss": 3.2801,
+      "step": 53300
+    },
+    {
+      "epoch": 15.535849496184985,
+      "grad_norm": 0.37049153447151184,
+      "learning_rate": 0.00041381468531468524,
+      "loss": 3.2764,
+      "step": 53350
+    },
+    {
+      "epoch": 15.550410623798706,
+      "grad_norm": 0.42307063937187195,
+      "learning_rate": 0.0004136398601398601,
+      "loss": 3.2896,
+      "step": 53400
+    },
+    {
+      "epoch": 15.564971751412429,
+      "grad_norm": 0.40026119351387024,
+      "learning_rate": 0.00041346503496503495,
+      "loss": 3.2919,
+      "step": 53450
+    },
+    {
+      "epoch": 15.579532879026152,
+      "grad_norm": 0.37621283531188965,
+      "learning_rate": 0.00041329020979020975,
+      "loss": 3.291,
+      "step": 53500
+    },
+    {
+      "epoch": 15.594094006639875,
+      "grad_norm": 0.3707675635814667,
+      "learning_rate": 0.0004131153846153846,
+      "loss": 3.2875,
+      "step": 53550
+    },
+    {
+      "epoch": 15.608655134253597,
+      "grad_norm": 0.39700794219970703,
+      "learning_rate": 0.0004129405594405594,
+      "loss": 3.2752,
+      "step": 53600
+    },
+    {
+      "epoch": 15.623216261867318,
+      "grad_norm": 0.38050559163093567,
+      "learning_rate": 0.00041276573426573426,
+      "loss": 3.297,
+      "step": 53650
+    },
+    {
+      "epoch": 15.637777389481041,
+      "grad_norm": 0.3697868883609772,
+      "learning_rate": 0.00041259090909090906,
+      "loss": 3.2839,
+      "step": 53700
+    },
+    {
+      "epoch": 15.652338517094764,
+      "grad_norm": 0.3799356520175934,
+      "learning_rate": 0.0004124160839160839,
+      "loss": 3.3066,
+      "step": 53750
+    },
+    {
+      "epoch": 15.666899644708487,
+      "grad_norm": 0.37823575735092163,
+      "learning_rate": 0.0004122412587412587,
+      "loss": 3.3015,
+      "step": 53800
+    },
+    {
+      "epoch": 15.68146077232221,
+      "grad_norm": 0.4178929030895233,
+      "learning_rate": 0.00041206643356643356,
+      "loss": 3.2922,
+      "step": 53850
+    },
+    {
+      "epoch": 15.69602189993593,
+      "grad_norm": 0.33638760447502136,
+      "learning_rate": 0.0004118916083916083,
+      "loss": 3.297,
+      "step": 53900
+    },
+    {
+      "epoch": 15.710583027549653,
+      "grad_norm": 0.38494572043418884,
+      "learning_rate": 0.00041171678321678316,
+      "loss": 3.2964,
+      "step": 53950
+    },
+    {
+      "epoch": 15.725144155163376,
+      "grad_norm": 0.3939765393733978,
+      "learning_rate": 0.00041154195804195796,
+      "loss": 3.2904,
+      "step": 54000
+    },
+    {
+      "epoch": 15.725144155163376,
+      "eval_accuracy": 0.3723768549322099,
+      "eval_loss": 3.5429139137268066,
+      "eval_runtime": 202.3548,
+      "eval_samples_per_second": 82.261,
+      "eval_steps_per_second": 5.144,
+      "step": 54000
+    },
+    {
+      "epoch": 15.739705282777098,
+      "grad_norm": 0.36110496520996094,
+      "learning_rate": 0.0004113671328671328,
+      "loss": 3.304,
+      "step": 54050
+    },
+    {
+      "epoch": 15.754266410390821,
+      "grad_norm": 0.3761616051197052,
+      "learning_rate": 0.00041119230769230767,
+      "loss": 3.2897,
+      "step": 54100
+    },
+    {
+      "epoch": 15.768827538004544,
+      "grad_norm": 0.38120609521865845,
+      "learning_rate": 0.00041101748251748247,
+      "loss": 3.2999,
+      "step": 54150
+    },
+    {
+      "epoch": 15.783388665618265,
+      "grad_norm": 0.36932608485221863,
+      "learning_rate": 0.0004108426573426573,
+      "loss": 3.2853,
+      "step": 54200
+    },
+    {
+      "epoch": 15.797949793231988,
+      "grad_norm": 0.37278443574905396,
+      "learning_rate": 0.0004106678321678321,
+      "loss": 3.2879,
+      "step": 54250
+    },
+    {
+      "epoch": 15.81251092084571,
+      "grad_norm": 0.38825732469558716,
+      "learning_rate": 0.000410493006993007,
+      "loss": 3.3032,
+      "step": 54300
+    },
+    {
+      "epoch": 15.827072048459433,
+      "grad_norm": 0.3719865679740906,
+      "learning_rate": 0.0004103181818181818,
+      "loss": 3.308,
+      "step": 54350
+    },
+    {
+      "epoch": 15.841633176073156,
+      "grad_norm": 0.39769595861434937,
+      "learning_rate": 0.00041014335664335663,
+      "loss": 3.3053,
+      "step": 54400
+    },
+    {
+      "epoch": 15.856194303686877,
+      "grad_norm": 0.3751024007797241,
+      "learning_rate": 0.00040996853146853143,
+      "loss": 3.2993,
+      "step": 54450
+    },
+    {
+      "epoch": 15.8707554313006,
+      "grad_norm": 0.3529784679412842,
+      "learning_rate": 0.0004097937062937063,
+      "loss": 3.3004,
+      "step": 54500
+    },
+    {
+      "epoch": 15.885316558914322,
+      "grad_norm": 0.37733763456344604,
+      "learning_rate": 0.0004096188811188811,
+      "loss": 3.2994,
+      "step": 54550
+    },
+    {
+      "epoch": 15.899877686528045,
+      "grad_norm": 0.37086179852485657,
+      "learning_rate": 0.00040944405594405594,
+      "loss": 3.2952,
+      "step": 54600
+    },
+    {
+      "epoch": 15.914438814141768,
+      "grad_norm": 0.416424036026001,
+      "learning_rate": 0.0004092692307692307,
+      "loss": 3.3018,
+      "step": 54650
+    },
+    {
+      "epoch": 15.928999941755489,
+      "grad_norm": 0.39904865622520447,
+      "learning_rate": 0.00040909440559440554,
+      "loss": 3.3005,
+      "step": 54700
+    },
+    {
+      "epoch": 15.943561069369212,
+      "grad_norm": 0.38030996918678284,
+      "learning_rate": 0.00040891958041958034,
+      "loss": 3.3018,
+      "step": 54750
+    },
+    {
+      "epoch": 15.958122196982934,
+      "grad_norm": 0.38479670882225037,
+      "learning_rate": 0.0004087447552447552,
+      "loss": 3.3043,
+      "step": 54800
+    },
+    {
+      "epoch": 15.972683324596657,
+      "grad_norm": 0.359994500875473,
+      "learning_rate": 0.00040856993006993005,
+      "loss": 3.3096,
+      "step": 54850
+    },
+    {
+      "epoch": 15.98724445221038,
+      "grad_norm": 0.3941158950328827,
+      "learning_rate": 0.00040839510489510485,
+      "loss": 3.3085,
+      "step": 54900
+    },
+    {
+      "epoch": 16.001747335313645,
+      "grad_norm": 0.4093943238258362,
+      "learning_rate": 0.0004082202797202797,
+      "loss": 3.3071,
+      "step": 54950
+    },
+    {
+      "epoch": 16.01630846292737,
+      "grad_norm": 0.3764057457447052,
+      "learning_rate": 0.0004080454545454545,
+      "loss": 3.1954,
+      "step": 55000
+    },
+    {
+      "epoch": 16.01630846292737,
+      "eval_accuracy": 0.37222978411037905,
+      "eval_loss": 3.551542043685913,
+      "eval_runtime": 224.7042,
+      "eval_samples_per_second": 74.08,
+      "eval_steps_per_second": 4.633,
+      "step": 55000
+    },
+    {
+      "epoch": 16.03086959054109,
+      "grad_norm": 0.38292109966278076,
+      "learning_rate": 0.00040787062937062935,
+      "loss": 3.2015,
+      "step": 55050
+    },
+    {
+      "epoch": 16.045430718154815,
+      "grad_norm": 0.4004324972629547,
+      "learning_rate": 0.00040769580419580415,
+      "loss": 3.1914,
+      "step": 55100
+    },
+    {
+      "epoch": 16.059991845768536,
+      "grad_norm": 0.39811986684799194,
+      "learning_rate": 0.000407520979020979,
+      "loss": 3.2156,
+      "step": 55150
+    },
+    {
+      "epoch": 16.074552973382257,
+      "grad_norm": 0.3610599935054779,
+      "learning_rate": 0.0004073461538461538,
+      "loss": 3.2128,
+      "step": 55200
+    },
+    {
+      "epoch": 16.08911410099598,
+      "grad_norm": 0.3569590449333191,
+      "learning_rate": 0.00040717132867132866,
+      "loss": 3.2195,
+      "step": 55250
+    },
+    {
+      "epoch": 16.103675228609703,
+      "grad_norm": 0.38127443194389343,
+      "learning_rate": 0.00040699650349650346,
+      "loss": 3.2195,
+      "step": 55300
+    },
+    {
+      "epoch": 16.118236356223427,
+      "grad_norm": 0.37626177072525024,
+      "learning_rate": 0.0004068216783216783,
+      "loss": 3.2321,
+      "step": 55350
+    },
+    {
+      "epoch": 16.132797483837148,
+      "grad_norm": 0.39430075883865356,
+      "learning_rate": 0.00040664685314685306,
+      "loss": 3.2147,
+      "step": 55400
+    },
+    {
+      "epoch": 16.14735861145087,
+      "grad_norm": 0.39753103256225586,
+      "learning_rate": 0.0004064720279720279,
+      "loss": 3.2395,
+      "step": 55450
+    },
+    {
+      "epoch": 16.161919739064594,
+      "grad_norm": 0.39400920271873474,
+      "learning_rate": 0.00040629720279720277,
+      "loss": 3.2244,
+      "step": 55500
+    },
+    {
+      "epoch": 16.176480866678315,
+      "grad_norm": 0.3588741719722748,
+      "learning_rate": 0.00040612237762237757,
+      "loss": 3.2348,
+      "step": 55550
+    },
+    {
+      "epoch": 16.19104199429204,
+      "grad_norm": 0.38264986872673035,
+      "learning_rate": 0.0004059475524475524,
+      "loss": 3.2271,
+      "step": 55600
+    },
+    {
+      "epoch": 16.20560312190576,
+      "grad_norm": 0.3960827589035034,
+      "learning_rate": 0.0004057727272727272,
+      "loss": 3.2539,
+      "step": 55650
+    },
+    {
+      "epoch": 16.22016424951948,
+      "grad_norm": 0.37086227536201477,
+      "learning_rate": 0.0004055979020979021,
+      "loss": 3.227,
+      "step": 55700
+    },
+    {
+      "epoch": 16.234725377133206,
+      "grad_norm": 0.36380138993263245,
+      "learning_rate": 0.0004054230769230769,
+      "loss": 3.247,
+      "step": 55750
+    },
+    {
+      "epoch": 16.249286504746927,
+      "grad_norm": 0.3885374069213867,
+      "learning_rate": 0.00040524825174825173,
+      "loss": 3.2482,
+      "step": 55800
+    },
+    {
+      "epoch": 16.26384763236065,
+      "grad_norm": 0.3623849153518677,
+      "learning_rate": 0.00040507342657342653,
+      "loss": 3.2442,
+      "step": 55850
+    },
+    {
+      "epoch": 16.278408759974372,
+      "grad_norm": 0.3633931279182434,
+      "learning_rate": 0.0004048986013986014,
+      "loss": 3.2488,
+      "step": 55900
+    },
+    {
+      "epoch": 16.292969887588093,
+      "grad_norm": 0.4284801483154297,
+      "learning_rate": 0.0004047237762237762,
+      "loss": 3.2495,
+      "step": 55950
+    },
+    {
+      "epoch": 16.307531015201818,
+      "grad_norm": 0.36431336402893066,
+      "learning_rate": 0.00040454895104895104,
+      "loss": 3.2552,
+      "step": 56000
+    },
+    {
+      "epoch": 16.307531015201818,
+      "eval_accuracy": 0.3717788139484742,
+      "eval_loss": 3.5528135299682617,
+      "eval_runtime": 181.2816,
+      "eval_samples_per_second": 91.824,
+      "eval_steps_per_second": 5.742,
+      "step": 56000
+    },
+    {
+      "epoch": 16.32209214281554,
+      "grad_norm": 0.4063562750816345,
+      "learning_rate": 0.00040437412587412583,
+      "loss": 3.2598,
+      "step": 56050
+    },
+    {
+      "epoch": 16.336653270429263,
+      "grad_norm": 0.3874892592430115,
+      "learning_rate": 0.0004041993006993007,
+      "loss": 3.2614,
+      "step": 56100
+    },
+    {
+      "epoch": 16.351214398042984,
+      "grad_norm": 0.37387025356292725,
+      "learning_rate": 0.00040402447552447554,
+      "loss": 3.2533,
+      "step": 56150
+    },
+    {
+      "epoch": 16.36577552565671,
+      "grad_norm": 0.38675811886787415,
+      "learning_rate": 0.0004038496503496503,
+      "loss": 3.2581,
+      "step": 56200
+    },
+    {
+      "epoch": 16.38033665327043,
+      "grad_norm": 0.37908634543418884,
+      "learning_rate": 0.00040367482517482514,
+      "loss": 3.2583,
+      "step": 56250
+    },
+    {
+      "epoch": 16.39489778088415,
+      "grad_norm": 0.3693457245826721,
+      "learning_rate": 0.00040349999999999994,
+      "loss": 3.2517,
+      "step": 56300
+    },
+    {
+      "epoch": 16.409458908497875,
+      "grad_norm": 0.38305196166038513,
+      "learning_rate": 0.0004033251748251748,
+      "loss": 3.2707,
+      "step": 56350
+    },
+    {
+      "epoch": 16.424020036111596,
+      "grad_norm": 0.38687750697135925,
+      "learning_rate": 0.0004031503496503496,
+      "loss": 3.2524,
+      "step": 56400
+    },
+    {
+      "epoch": 16.43858116372532,
+      "grad_norm": 0.36994367837905884,
+      "learning_rate": 0.00040297552447552445,
+      "loss": 3.2658,
+      "step": 56450
+    },
+    {
+      "epoch": 16.45314229133904,
+      "grad_norm": 0.3925367295742035,
+      "learning_rate": 0.00040280069930069925,
+      "loss": 3.254,
+      "step": 56500
+    },
+    {
+      "epoch": 16.467703418952762,
+      "grad_norm": 0.37788835167884827,
+      "learning_rate": 0.0004026258741258741,
+      "loss": 3.2669,
+      "step": 56550
+    },
+    {
+      "epoch": 16.482264546566487,
+      "grad_norm": 0.370782732963562,
+      "learning_rate": 0.0004024510489510489,
+      "loss": 3.2512,
+      "step": 56600
+    },
+    {
+      "epoch": 16.496825674180208,
+      "grad_norm": 0.37966829538345337,
+      "learning_rate": 0.00040227622377622376,
+      "loss": 3.2771,
+      "step": 56650
+    },
+    {
+      "epoch": 16.511386801793932,
+      "grad_norm": 0.3862817883491516,
+      "learning_rate": 0.00040210139860139856,
+      "loss": 3.2721,
+      "step": 56700
+    },
+    {
+      "epoch": 16.525947929407653,
+      "grad_norm": 0.38935843110084534,
+      "learning_rate": 0.0004019265734265734,
+      "loss": 3.2803,
+      "step": 56750
+    },
+    {
+      "epoch": 16.540509057021374,
+      "grad_norm": 0.3634398877620697,
+      "learning_rate": 0.0004017517482517482,
+      "loss": 3.2703,
+      "step": 56800
+    },
+    {
+      "epoch": 16.5550701846351,
+      "grad_norm": 0.3821653127670288,
+      "learning_rate": 0.00040157692307692306,
+      "loss": 3.271,
+      "step": 56850
+    },
+    {
+      "epoch": 16.56963131224882,
+      "grad_norm": 0.4057944715023041,
+      "learning_rate": 0.0004014020979020979,
+      "loss": 3.2701,
+      "step": 56900
+    },
+    {
+      "epoch": 16.584192439862544,
+      "grad_norm": 0.38798093795776367,
+      "learning_rate": 0.00040122727272727266,
+      "loss": 3.2679,
+      "step": 56950
+    },
+    {
+      "epoch": 16.598753567476265,
+      "grad_norm": 0.3680713474750519,
+      "learning_rate": 0.0004010524475524475,
+      "loss": 3.2751,
+      "step": 57000
+    },
+    {
+      "epoch": 16.598753567476265,
+      "eval_accuracy": 0.37280078569441766,
+      "eval_loss": 3.543818235397339,
+      "eval_runtime": 181.427,
+      "eval_samples_per_second": 91.75,
+      "eval_steps_per_second": 5.738,
+      "step": 57000
+    },
+    {
+      "epoch": 16.613314695089986,
+      "grad_norm": 0.37805137038230896,
+      "learning_rate": 0.0004008776223776223,
+      "loss": 3.2773,
+      "step": 57050
+    },
+    {
+      "epoch": 16.62787582270371,
+      "grad_norm": 0.3803144097328186,
+      "learning_rate": 0.00040070279720279717,
+      "loss": 3.2769,
+      "step": 57100
+    },
+    {
+      "epoch": 16.642436950317432,
+      "grad_norm": 0.3835834562778473,
+      "learning_rate": 0.00040052797202797197,
+      "loss": 3.2722,
+      "step": 57150
+    },
+    {
+      "epoch": 16.656998077931156,
+      "grad_norm": 0.374383807182312,
+      "learning_rate": 0.0004003531468531468,
+      "loss": 3.2784,
+      "step": 57200
+    },
+    {
+      "epoch": 16.671559205544877,
+      "grad_norm": 0.3727409839630127,
+      "learning_rate": 0.0004001783216783216,
+      "loss": 3.2862,
+      "step": 57250
+    },
+    {
+      "epoch": 16.6861203331586,
+      "grad_norm": 0.37330153584480286,
+      "learning_rate": 0.0004000034965034965,
+      "loss": 3.2933,
+      "step": 57300
+    },
+    {
+      "epoch": 16.700681460772323,
+      "grad_norm": 0.3724009394645691,
+      "learning_rate": 0.0003998286713286713,
+      "loss": 3.2905,
+      "step": 57350
+    },
+    {
+      "epoch": 16.715242588386044,
+      "grad_norm": 0.40461817383766174,
+      "learning_rate": 0.00039965384615384613,
+      "loss": 3.2834,
+      "step": 57400
+    },
+    {
+      "epoch": 16.72980371599977,
+      "grad_norm": 0.3559574782848358,
+      "learning_rate": 0.00039947902097902093,
+      "loss": 3.2804,
+      "step": 57450
+    },
+    {
+      "epoch": 16.74436484361349,
+      "grad_norm": 0.38186269998550415,
+      "learning_rate": 0.0003993041958041958,
+      "loss": 3.2892,
+      "step": 57500
+    },
+    {
+      "epoch": 16.75892597122721,
+      "grad_norm": 0.4213004410266876,
+      "learning_rate": 0.00039912937062937064,
+      "loss": 3.282,
+      "step": 57550
+    },
+    {
+      "epoch": 16.773487098840935,
+      "grad_norm": 0.37131091952323914,
+      "learning_rate": 0.00039895454545454544,
+      "loss": 3.2779,
+      "step": 57600
+    },
+    {
+      "epoch": 16.788048226454656,
+      "grad_norm": 0.37532955408096313,
+      "learning_rate": 0.0003987797202797203,
+      "loss": 3.2911,
+      "step": 57650
+    },
+    {
+      "epoch": 16.80260935406838,
+      "grad_norm": 0.41479650139808655,
+      "learning_rate": 0.00039860489510489504,
+      "loss": 3.2926,
+      "step": 57700
+    },
+    {
+      "epoch": 16.8171704816821,
+      "grad_norm": 0.40581607818603516,
+      "learning_rate": 0.0003984300699300699,
+      "loss": 3.2895,
+      "step": 57750
+    },
+    {
+      "epoch": 16.831731609295822,
+      "grad_norm": 0.37579381465911865,
+      "learning_rate": 0.0003982552447552447,
+      "loss": 3.2862,
+      "step": 57800
+    },
+    {
+      "epoch": 16.846292736909547,
+      "grad_norm": 0.38346877694129944,
+      "learning_rate": 0.00039808041958041955,
+      "loss": 3.2917,
+      "step": 57850
+    },
+    {
+      "epoch": 16.860853864523268,
+      "grad_norm": 0.3862057328224182,
+      "learning_rate": 0.00039790559440559435,
+      "loss": 3.2912,
+      "step": 57900
+    },
+    {
+      "epoch": 16.875414992136992,
+      "grad_norm": 0.36858367919921875,
+      "learning_rate": 0.0003977307692307692,
+      "loss": 3.2872,
+      "step": 57950
+    },
+    {
+      "epoch": 16.889976119750713,
+      "grad_norm": 0.37914177775382996,
+      "learning_rate": 0.000397555944055944,
+      "loss": 3.2981,
+      "step": 58000
+    },
+    {
+      "epoch": 16.889976119750713,
+      "eval_accuracy": 0.37295396977183215,
+      "eval_loss": 3.536424398422241,
+      "eval_runtime": 181.4266,
+      "eval_samples_per_second": 91.751,
+      "eval_steps_per_second": 5.738,
+      "step": 58000
+    },
+    {
+      "epoch": 16.904537247364434,
+      "grad_norm": 0.3646220266819,
+      "learning_rate": 0.00039738111888111885,
+      "loss": 3.2928,
+      "step": 58050
+    },
+    {
+      "epoch": 16.91909837497816,
+      "grad_norm": 0.3765883445739746,
+      "learning_rate": 0.00039720629370629365,
+      "loss": 3.293,
+      "step": 58100
+    },
+    {
+      "epoch": 16.93365950259188,
+      "grad_norm": 0.3603968322277069,
+      "learning_rate": 0.0003970314685314685,
+      "loss": 3.2846,
+      "step": 58150
+    },
+    {
+      "epoch": 16.948220630205604,
+      "grad_norm": 0.38871946930885315,
+      "learning_rate": 0.0003968566433566433,
+      "loss": 3.2851,
+      "step": 58200
+    },
+    {
+      "epoch": 16.962781757819325,
+      "grad_norm": 0.3598012626171112,
+      "learning_rate": 0.00039668181818181816,
+      "loss": 3.3014,
+      "step": 58250
+    },
+    {
+      "epoch": 16.977342885433046,
+      "grad_norm": 0.34035736322402954,
+      "learning_rate": 0.000396506993006993,
+      "loss": 3.301,
+      "step": 58300
+    },
+    {
+      "epoch": 16.99190401304677,
+      "grad_norm": 0.41209015250205994,
+      "learning_rate": 0.0003963321678321678,
+      "loss": 3.2904,
+      "step": 58350
+    },
+    {
+      "epoch": 17.006406896150036,
+      "grad_norm": 0.37024733424186707,
+      "learning_rate": 0.00039615734265734267,
+      "loss": 3.2411,
+      "step": 58400
+    },
+    {
+      "epoch": 17.02096802376376,
+      "grad_norm": 0.37371373176574707,
+      "learning_rate": 0.0003959825174825174,
+      "loss": 3.1822,
+      "step": 58450
+    },
+    {
+      "epoch": 17.03552915137748,
+      "grad_norm": 0.3961547911167145,
+      "learning_rate": 0.00039580769230769227,
+      "loss": 3.1896,
+      "step": 58500
+    },
+    {
+      "epoch": 17.050090278991206,
+      "grad_norm": 0.4139074385166168,
+      "learning_rate": 0.00039563286713286707,
+      "loss": 3.1983,
+      "step": 58550
+    },
+    {
+      "epoch": 17.064651406604927,
+      "grad_norm": 0.36274170875549316,
+      "learning_rate": 0.0003954580419580419,
+      "loss": 3.2071,
+      "step": 58600
+    },
+    {
+      "epoch": 17.07921253421865,
+      "grad_norm": 0.40999776124954224,
+      "learning_rate": 0.0003952832167832167,
+      "loss": 3.2004,
+      "step": 58650
+    },
+    {
+      "epoch": 17.093773661832373,
+      "grad_norm": 0.38540902733802795,
+      "learning_rate": 0.0003951083916083916,
+      "loss": 3.2124,
+      "step": 58700
+    },
+    {
+      "epoch": 17.108334789446094,
+      "grad_norm": 0.41332516074180603,
+      "learning_rate": 0.0003949335664335664,
+      "loss": 3.2163,
+      "step": 58750
+    },
+    {
+      "epoch": 17.122895917059818,
+      "grad_norm": 0.394908607006073,
+      "learning_rate": 0.00039475874125874123,
+      "loss": 3.2062,
+      "step": 58800
+    },
+    {
+      "epoch": 17.13745704467354,
+      "grad_norm": 0.3662570118904114,
+      "learning_rate": 0.00039458391608391603,
+      "loss": 3.2192,
+      "step": 58850
+    },
+    {
+      "epoch": 17.152018172287264,
+      "grad_norm": 0.3641432523727417,
+      "learning_rate": 0.0003944090909090909,
+      "loss": 3.2159,
+      "step": 58900
+    },
+    {
+      "epoch": 17.166579299900985,
+      "grad_norm": 0.3532322943210602,
+      "learning_rate": 0.00039423426573426573,
+      "loss": 3.2213,
+      "step": 58950
+    },
+    {
+      "epoch": 17.181140427514705,
+      "grad_norm": 0.39691367745399475,
+      "learning_rate": 0.00039405944055944053,
+      "loss": 3.2182,
+      "step": 59000
+    },
+    {
+      "epoch": 17.181140427514705,
+      "eval_accuracy": 0.37223095973645287,
+      "eval_loss": 3.5541632175445557,
+      "eval_runtime": 181.4029,
+      "eval_samples_per_second": 91.763,
+      "eval_steps_per_second": 5.739,
+      "step": 59000
+    },
+    {
+      "epoch": 17.19570155512843,
+      "grad_norm": 0.3973633050918579,
+      "learning_rate": 0.0003938846153846154,
+      "loss": 3.2106,
+      "step": 59050
+    },
+    {
+      "epoch": 17.21026268274215,
+      "grad_norm": 0.36320367455482483,
+      "learning_rate": 0.0003937097902097902,
+      "loss": 3.2215,
+      "step": 59100
+    },
+    {
+      "epoch": 17.224823810355876,
+      "grad_norm": 0.4017624855041504,
+      "learning_rate": 0.00039353496503496504,
+      "loss": 3.2241,
+      "step": 59150
+    },
+    {
+      "epoch": 17.239384937969596,
+      "grad_norm": 0.3894210457801819,
+      "learning_rate": 0.0003933601398601398,
+      "loss": 3.2272,
+      "step": 59200
+    },
+    {
+      "epoch": 17.253946065583317,
+      "grad_norm": 0.38473623991012573,
+      "learning_rate": 0.00039318531468531464,
+      "loss": 3.2452,
+      "step": 59250
+    },
+    {
+      "epoch": 17.268507193197042,
+      "grad_norm": 0.3801329731941223,
+      "learning_rate": 0.00039301048951048944,
+      "loss": 3.2393,
+      "step": 59300
+    },
+    {
+      "epoch": 17.283068320810763,
+      "grad_norm": 0.4249562919139862,
+      "learning_rate": 0.0003928356643356643,
+      "loss": 3.2346,
+      "step": 59350
+    },
+    {
+      "epoch": 17.297629448424487,
+      "grad_norm": 0.390713632106781,
+      "learning_rate": 0.0003926608391608391,
+      "loss": 3.2371,
+      "step": 59400
+    },
+    {
+      "epoch": 17.31219057603821,
+      "grad_norm": 0.41129425168037415,
+      "learning_rate": 0.00039248601398601395,
+      "loss": 3.2415,
+      "step": 59450
+    },
+    {
+      "epoch": 17.32675170365193,
+      "grad_norm": 0.3896760940551758,
+      "learning_rate": 0.00039231118881118875,
+      "loss": 3.2395,
+      "step": 59500
+    },
+    {
+      "epoch": 17.341312831265654,
+      "grad_norm": 0.3870900869369507,
+      "learning_rate": 0.0003921363636363636,
+      "loss": 3.253,
+      "step": 59550
+    },
+    {
+      "epoch": 17.355873958879375,
+      "grad_norm": 0.4042428731918335,
+      "learning_rate": 0.00039196153846153846,
+      "loss": 3.2365,
+      "step": 59600
+    },
+    {
+      "epoch": 17.3704350864931,
+      "grad_norm": 0.3771011233329773,
+      "learning_rate": 0.00039178671328671326,
+      "loss": 3.2493,
+      "step": 59650
+    },
+    {
+      "epoch": 17.38499621410682,
+      "grad_norm": 0.3748337924480438,
+      "learning_rate": 0.0003916118881118881,
+      "loss": 3.2468,
+      "step": 59700
+    },
+    {
+      "epoch": 17.39955734172054,
+      "grad_norm": 0.3741134703159332,
+      "learning_rate": 0.0003914370629370629,
+      "loss": 3.2559,
+      "step": 59750
+    },
+    {
+      "epoch": 17.414118469334266,
+      "grad_norm": 0.39321649074554443,
+      "learning_rate": 0.00039126223776223776,
+      "loss": 3.2449,
+      "step": 59800
+    },
+    {
+      "epoch": 17.428679596947987,
+      "grad_norm": 0.3862929344177246,
+      "learning_rate": 0.00039108741258741256,
+      "loss": 3.2443,
+      "step": 59850
+    },
+    {
+      "epoch": 17.44324072456171,
+      "grad_norm": 0.40072572231292725,
+      "learning_rate": 0.0003909125874125874,
+      "loss": 3.2471,
+      "step": 59900
+    },
+    {
+      "epoch": 17.457801852175432,
+      "grad_norm": 0.4031667113304138,
+      "learning_rate": 0.00039073776223776216,
+      "loss": 3.2473,
+      "step": 59950
+    },
+    {
+      "epoch": 17.472362979789153,
+      "grad_norm": 0.39461931586265564,
+      "learning_rate": 0.000390562937062937,
+      "loss": 3.261,
+      "step": 60000
+    },
+    {
+      "epoch": 17.472362979789153,
+      "eval_accuracy": 0.3726906295313038,
+      "eval_loss": 3.5467474460601807,
+      "eval_runtime": 181.6124,
+      "eval_samples_per_second": 91.657,
+      "eval_steps_per_second": 5.732,
+      "step": 60000
+    },
+    {
+      "epoch": 17.486924107402878,
+      "grad_norm": 0.4082168638706207,
+      "learning_rate": 0.0003903881118881118,
+      "loss": 3.255,
+      "step": 60050
+    },
+    {
+      "epoch": 17.5014852350166,
+      "grad_norm": 0.4216744899749756,
+      "learning_rate": 0.00039021328671328667,
+      "loss": 3.2661,
+      "step": 60100
+    },
+    {
+      "epoch": 17.516046362630323,
+      "grad_norm": 0.38183650374412537,
+      "learning_rate": 0.00039003846153846147,
+      "loss": 3.2623,
+      "step": 60150
+    },
+    {
+      "epoch": 17.530607490244044,
+      "grad_norm": 0.38617098331451416,
+      "learning_rate": 0.0003898636363636363,
+      "loss": 3.2568,
+      "step": 60200
+    },
+    {
+      "epoch": 17.545168617857765,
+      "grad_norm": 0.40580758452415466,
+      "learning_rate": 0.0003896888111888111,
+      "loss": 3.2626,
+      "step": 60250
+    },
+    {
+      "epoch": 17.55972974547149,
+      "grad_norm": 0.4091379642486572,
+      "learning_rate": 0.000389513986013986,
+      "loss": 3.2525,
+      "step": 60300
+    },
+    {
+      "epoch": 17.57429087308521,
+      "grad_norm": 0.36811408400535583,
+      "learning_rate": 0.00038933916083916083,
+      "loss": 3.2698,
+      "step": 60350
+    },
+    {
+      "epoch": 17.588852000698935,
+      "grad_norm": 0.36896032094955444,
+      "learning_rate": 0.00038916433566433563,
+      "loss": 3.2636,
+      "step": 60400
+    },
+    {
+      "epoch": 17.603413128312656,
+      "grad_norm": 0.3820686340332031,
+      "learning_rate": 0.0003889895104895105,
+      "loss": 3.2738,
+      "step": 60450
+    },
+    {
+      "epoch": 17.617974255926377,
+      "grad_norm": 0.4073737561702728,
+      "learning_rate": 0.0003888146853146853,
+      "loss": 3.2677,
+      "step": 60500
+    },
+    {
+      "epoch": 17.6325353835401,
+      "grad_norm": 0.3798710107803345,
+      "learning_rate": 0.00038863986013986014,
+      "loss": 3.2739,
+      "step": 60550
+    },
+    {
+      "epoch": 17.647096511153823,
+      "grad_norm": 0.413150429725647,
+      "learning_rate": 0.00038846503496503494,
+      "loss": 3.2687,
+      "step": 60600
+    },
+    {
+      "epoch": 17.661657638767547,
+      "grad_norm": 0.37528276443481445,
+      "learning_rate": 0.0003882902097902098,
+      "loss": 3.2653,
+      "step": 60650
+    },
+    {
+      "epoch": 17.676218766381268,
+      "grad_norm": 0.40495121479034424,
+      "learning_rate": 0.00038811538461538454,
+      "loss": 3.2781,
+      "step": 60700
+    },
+    {
+      "epoch": 17.690779893994993,
+      "grad_norm": 0.3710327446460724,
+      "learning_rate": 0.0003879405594405594,
+      "loss": 3.258,
+      "step": 60750
+    },
+    {
+      "epoch": 17.705341021608714,
+      "grad_norm": 0.3918648064136505,
+      "learning_rate": 0.0003877657342657342,
+      "loss": 3.266,
+      "step": 60800
+    },
+    {
+      "epoch": 17.719902149222435,
+      "grad_norm": 0.3934940993785858,
+      "learning_rate": 0.00038759090909090905,
+      "loss": 3.2684,
+      "step": 60850
+    },
+    {
+      "epoch": 17.73446327683616,
+      "grad_norm": 0.41869670152664185,
+      "learning_rate": 0.00038741608391608384,
+      "loss": 3.2637,
+      "step": 60900
+    },
+    {
+      "epoch": 17.74902440444988,
+      "grad_norm": 0.3695789575576782,
+      "learning_rate": 0.0003872412587412587,
+      "loss": 3.2686,
+      "step": 60950
+    },
+    {
+      "epoch": 17.763585532063605,
+      "grad_norm": 0.4143275320529938,
+      "learning_rate": 0.00038706643356643355,
+      "loss": 3.2801,
+      "step": 61000
+    },
+    {
+      "epoch": 17.763585532063605,
+      "eval_accuracy": 0.37312643411685675,
+      "eval_loss": 3.5362603664398193,
+      "eval_runtime": 181.741,
+      "eval_samples_per_second": 91.592,
+      "eval_steps_per_second": 5.728,
+      "step": 61000
+    },
+    {
+      "epoch": 17.778146659677326,
+      "grad_norm": 0.36733829975128174,
+      "learning_rate": 0.00038689160839160835,
+      "loss": 3.2712,
+      "step": 61050
+    },
+    {
+      "epoch": 17.792707787291047,
+      "grad_norm": 0.39132675528526306,
+      "learning_rate": 0.0003867167832167832,
+      "loss": 3.2772,
+      "step": 61100
+    },
+    {
+      "epoch": 17.80726891490477,
+      "grad_norm": 0.3637419044971466,
+      "learning_rate": 0.000386541958041958,
+      "loss": 3.2711,
+      "step": 61150
+    },
+    {
+      "epoch": 17.821830042518492,
+      "grad_norm": 0.39945223927497864,
+      "learning_rate": 0.00038636713286713286,
+      "loss": 3.2662,
+      "step": 61200
+    },
+    {
+      "epoch": 17.836391170132217,
+      "grad_norm": 0.3926716148853302,
+      "learning_rate": 0.00038619230769230766,
+      "loss": 3.2793,
+      "step": 61250
+    },
+    {
+      "epoch": 17.850952297745938,
+      "grad_norm": 0.38200077414512634,
+      "learning_rate": 0.0003860174825174825,
+      "loss": 3.2801,
+      "step": 61300
+    },
+    {
+      "epoch": 17.86551342535966,
+      "grad_norm": 0.383569598197937,
+      "learning_rate": 0.0003858426573426573,
+      "loss": 3.2805,
+      "step": 61350
+    },
+    {
+      "epoch": 17.880074552973383,
+      "grad_norm": 0.39642050862312317,
+      "learning_rate": 0.00038566783216783217,
+      "loss": 3.2796,
+      "step": 61400
+    },
+    {
+      "epoch": 17.894635680587104,
+      "grad_norm": 0.36815083026885986,
+      "learning_rate": 0.0003854930069930069,
+      "loss": 3.2764,
+      "step": 61450
+    },
+    {
+      "epoch": 17.90919680820083,
+      "grad_norm": 0.3897499442100525,
+      "learning_rate": 0.00038531818181818177,
+      "loss": 3.2772,
+      "step": 61500
+    },
+    {
+      "epoch": 17.92375793581455,
+      "grad_norm": 0.37837308645248413,
+      "learning_rate": 0.00038514335664335657,
+      "loss": 3.2797,
+      "step": 61550
+    },
+    {
+      "epoch": 17.93831906342827,
+      "grad_norm": 0.41439804434776306,
+      "learning_rate": 0.0003849685314685314,
+      "loss": 3.2879,
+      "step": 61600
+    },
+    {
+      "epoch": 17.952880191041995,
+      "grad_norm": 0.37204301357269287,
+      "learning_rate": 0.0003847937062937062,
+      "loss": 3.2851,
+      "step": 61650
+    },
+    {
+      "epoch": 17.967441318655716,
+      "grad_norm": 0.3963753283023834,
+      "learning_rate": 0.0003846188811188811,
+      "loss": 3.2828,
+      "step": 61700
+    },
+    {
+      "epoch": 17.98200244626944,
+      "grad_norm": 0.3774401545524597,
+      "learning_rate": 0.00038444405594405593,
+      "loss": 3.281,
+      "step": 61750
+    },
+    {
+      "epoch": 17.99656357388316,
+      "grad_norm": 0.38163232803344727,
+      "learning_rate": 0.00038426923076923073,
+      "loss": 3.2773,
+      "step": 61800
+    },
+    {
+      "epoch": 18.01106645698643,
+      "grad_norm": 0.39165937900543213,
+      "learning_rate": 0.0003840944055944056,
+      "loss": 3.2078,
+      "step": 61850
+    },
+    {
+      "epoch": 18.02562758460015,
+      "grad_norm": 0.3841058313846588,
+      "learning_rate": 0.0003839195804195804,
+      "loss": 3.1829,
+      "step": 61900
+    },
+    {
+      "epoch": 18.040188712213872,
+      "grad_norm": 0.40802210569381714,
+      "learning_rate": 0.00038374475524475523,
+      "loss": 3.1849,
+      "step": 61950
+    },
+    {
+      "epoch": 18.054749839827597,
+      "grad_norm": 0.3776838779449463,
+      "learning_rate": 0.00038356993006993003,
+      "loss": 3.1875,
+      "step": 62000
+    },
+    {
+      "epoch": 18.054749839827597,
+      "eval_accuracy": 0.372360748854999,
+      "eval_loss": 3.554863929748535,
+      "eval_runtime": 181.5806,
+      "eval_samples_per_second": 91.673,
+      "eval_steps_per_second": 5.733,
+      "step": 62000
+    },
+    {
+      "epoch": 18.069310967441318,
+      "grad_norm": 0.3968208432197571,
+      "learning_rate": 0.0003833951048951049,
+      "loss": 3.1835,
+      "step": 62050
+    },
+    {
+      "epoch": 18.083872095055042,
+      "grad_norm": 0.40352922677993774,
+      "learning_rate": 0.0003832202797202797,
+      "loss": 3.182,
+      "step": 62100
+    },
+    {
+      "epoch": 18.098433222668763,
+      "grad_norm": 0.3768850564956665,
+      "learning_rate": 0.00038304545454545454,
+      "loss": 3.2048,
+      "step": 62150
+    },
+    {
+      "epoch": 18.112994350282484,
+      "grad_norm": 0.4197027385234833,
+      "learning_rate": 0.0003828706293706293,
+      "loss": 3.2066,
+      "step": 62200
+    },
+    {
+      "epoch": 18.12755547789621,
+      "grad_norm": 0.4093753397464752,
+      "learning_rate": 0.00038269580419580414,
+      "loss": 3.2117,
+      "step": 62250
+    },
+    {
+      "epoch": 18.14211660550993,
+      "grad_norm": 0.37517985701560974,
+      "learning_rate": 0.00038252097902097894,
+      "loss": 3.2091,
+      "step": 62300
+    },
+    {
+      "epoch": 18.156677733123654,
+      "grad_norm": 0.3777500092983246,
+      "learning_rate": 0.0003823461538461538,
+      "loss": 3.2062,
+      "step": 62350
+    },
+    {
+      "epoch": 18.171238860737375,
+      "grad_norm": 0.40095406770706177,
+      "learning_rate": 0.00038217132867132865,
+      "loss": 3.2225,
+      "step": 62400
+    },
+    {
+      "epoch": 18.185799988351096,
+      "grad_norm": 0.3697664141654968,
+      "learning_rate": 0.00038199650349650345,
+      "loss": 3.2109,
+      "step": 62450
+    },
+    {
+      "epoch": 18.20036111596482,
+      "grad_norm": 0.36581069231033325,
+      "learning_rate": 0.0003818216783216783,
+      "loss": 3.2116,
+      "step": 62500
+    },
+    {
+      "epoch": 18.214922243578542,
+      "grad_norm": 0.3837885856628418,
+      "learning_rate": 0.0003816468531468531,
+      "loss": 3.2055,
+      "step": 62550
+    },
+    {
+      "epoch": 18.229483371192266,
+      "grad_norm": 0.42037078738212585,
+      "learning_rate": 0.00038147202797202796,
+      "loss": 3.2034,
+      "step": 62600
+    },
+    {
+      "epoch": 18.244044498805987,
+      "grad_norm": 0.4067479968070984,
+      "learning_rate": 0.00038129720279720276,
+      "loss": 3.2087,
+      "step": 62650
+    },
+    {
+      "epoch": 18.25860562641971,
+      "grad_norm": 0.3840962052345276,
+      "learning_rate": 0.0003811223776223776,
+      "loss": 3.2361,
+      "step": 62700
+    },
+    {
+      "epoch": 18.273166754033433,
+      "grad_norm": 0.375093936920166,
+      "learning_rate": 0.0003809475524475524,
+      "loss": 3.2333,
+      "step": 62750
+    },
+    {
+      "epoch": 18.287727881647154,
+      "grad_norm": 0.38850006461143494,
+      "learning_rate": 0.00038077272727272726,
+      "loss": 3.2303,
+      "step": 62800
+    },
+    {
+      "epoch": 18.30228900926088,
+      "grad_norm": 0.4177054166793823,
+      "learning_rate": 0.00038059790209790206,
+      "loss": 3.2337,
+      "step": 62850
+    },
+    {
+      "epoch": 18.3168501368746,
+      "grad_norm": 0.40772518515586853,
+      "learning_rate": 0.0003804230769230769,
+      "loss": 3.2256,
+      "step": 62900
+    },
+    {
+      "epoch": 18.33141126448832,
+      "grad_norm": 0.3933784067630768,
+      "learning_rate": 0.00038024825174825166,
+      "loss": 3.2287,
+      "step": 62950
+    },
+    {
+      "epoch": 18.345972392102045,
+      "grad_norm": 0.3855026066303253,
+      "learning_rate": 0.0003800734265734265,
+      "loss": 3.2276,
+      "step": 63000
+    },
+    {
+      "epoch": 18.345972392102045,
+      "eval_accuracy": 0.37281912546116874,
+      "eval_loss": 3.5480048656463623,
+      "eval_runtime": 181.5485,
+      "eval_samples_per_second": 91.689,
+      "eval_steps_per_second": 5.734,
+      "step": 63000
+    },
+    {
+      "epoch": 18.360533519715766,
+      "grad_norm": 0.39833077788352966,
+      "learning_rate": 0.0003798986013986013,
+      "loss": 3.2216,
+      "step": 63050
+    },
+    {
+      "epoch": 18.37509464732949,
+      "grad_norm": 0.39158907532691956,
+      "learning_rate": 0.00037972377622377617,
+      "loss": 3.2331,
+      "step": 63100
+    },
+    {
+      "epoch": 18.38965577494321,
+      "grad_norm": 0.3841773569583893,
+      "learning_rate": 0.000379548951048951,
+      "loss": 3.2408,
+      "step": 63150
+    },
+    {
+      "epoch": 18.404216902556932,
+      "grad_norm": 0.392995685338974,
+      "learning_rate": 0.0003793741258741258,
+      "loss": 3.2356,
+      "step": 63200
+    },
+    {
+      "epoch": 18.418778030170657,
+      "grad_norm": 0.40546339750289917,
+      "learning_rate": 0.0003791993006993007,
+      "loss": 3.2303,
+      "step": 63250
+    },
+    {
+      "epoch": 18.433339157784378,
+      "grad_norm": 0.41558903455734253,
+      "learning_rate": 0.0003790244755244755,
+      "loss": 3.2352,
+      "step": 63300
+    },
+    {
+      "epoch": 18.447900285398102,
+      "grad_norm": 0.3937760889530182,
+      "learning_rate": 0.00037884965034965033,
+      "loss": 3.232,
+      "step": 63350
+    },
+    {
+      "epoch": 18.462461413011823,
+      "grad_norm": 0.3903751075267792,
+      "learning_rate": 0.00037867482517482513,
+      "loss": 3.2268,
+      "step": 63400
+    },
+    {
+      "epoch": 18.477022540625548,
+      "grad_norm": 0.41812342405319214,
+      "learning_rate": 0.0003785,
+      "loss": 3.2368,
+      "step": 63450
+    },
+    {
+      "epoch": 18.49158366823927,
+      "grad_norm": 0.37955912947654724,
+      "learning_rate": 0.0003783251748251748,
+      "loss": 3.2381,
+      "step": 63500
+    },
+    {
+      "epoch": 18.50614479585299,
+      "grad_norm": 0.39163386821746826,
+      "learning_rate": 0.00037815034965034964,
+      "loss": 3.2593,
+      "step": 63550
+    },
+    {
+      "epoch": 18.520705923466714,
+      "grad_norm": 0.3994816541671753,
+      "learning_rate": 0.00037797552447552444,
+      "loss": 3.2357,
+      "step": 63600
+    },
+    {
+      "epoch": 18.535267051080435,
+      "grad_norm": 0.4218083322048187,
+      "learning_rate": 0.0003778006993006993,
+      "loss": 3.2518,
+      "step": 63650
+    },
+    {
+      "epoch": 18.54982817869416,
+      "grad_norm": 0.4084559679031372,
+      "learning_rate": 0.00037762587412587404,
+      "loss": 3.2486,
+      "step": 63700
+    },
+    {
+      "epoch": 18.56438930630788,
+      "grad_norm": 0.36928969621658325,
+      "learning_rate": 0.0003774510489510489,
+      "loss": 3.24,
+      "step": 63750
+    },
+    {
+      "epoch": 18.5789504339216,
+      "grad_norm": 0.3866930603981018,
+      "learning_rate": 0.0003772762237762238,
+      "loss": 3.25,
+      "step": 63800
+    },
+    {
+      "epoch": 18.593511561535326,
+      "grad_norm": 0.40036076307296753,
+      "learning_rate": 0.00037710139860139854,
+      "loss": 3.2509,
+      "step": 63850
+    },
+    {
+      "epoch": 18.608072689149047,
+      "grad_norm": 0.38957899808883667,
+      "learning_rate": 0.0003769265734265734,
+      "loss": 3.248,
+      "step": 63900
+    },
+    {
+      "epoch": 18.62263381676277,
+      "grad_norm": 0.3876812756061554,
+      "learning_rate": 0.0003767517482517482,
+      "loss": 3.254,
+      "step": 63950
+    },
+    {
+      "epoch": 18.637194944376493,
+      "grad_norm": 0.40378889441490173,
+      "learning_rate": 0.00037657692307692305,
+      "loss": 3.2535,
+      "step": 64000
+    },
+    {
+      "epoch": 18.637194944376493,
+      "eval_accuracy": 0.37327244687522115,
+      "eval_loss": 3.53926157951355,
+      "eval_runtime": 181.376,
+      "eval_samples_per_second": 91.776,
+      "eval_steps_per_second": 5.739,
+      "step": 64000
+    },
+    {
+      "epoch": 18.651756071990214,
+      "grad_norm": 0.40529847145080566,
+      "learning_rate": 0.00037640209790209785,
+      "loss": 3.2524,
+      "step": 64050
+    },
+    {
+      "epoch": 18.666317199603938,
+      "grad_norm": 0.3858989179134369,
+      "learning_rate": 0.0003762272727272727,
+      "loss": 3.266,
+      "step": 64100
+    },
+    {
+      "epoch": 18.68087832721766,
+      "grad_norm": 0.402764230966568,
+      "learning_rate": 0.0003760524475524475,
+      "loss": 3.2595,
+      "step": 64150
+    },
+    {
+      "epoch": 18.695439454831384,
+      "grad_norm": 0.42223840951919556,
+      "learning_rate": 0.00037587762237762236,
+      "loss": 3.2551,
+      "step": 64200
+    },
+    {
+      "epoch": 18.710000582445105,
+      "grad_norm": 0.3946334719657898,
+      "learning_rate": 0.00037570279720279716,
+      "loss": 3.2581,
+      "step": 64250
+    },
+    {
+      "epoch": 18.724561710058826,
+      "grad_norm": 0.38452982902526855,
+      "learning_rate": 0.000375527972027972,
+      "loss": 3.2568,
+      "step": 64300
+    },
+    {
+      "epoch": 18.73912283767255,
+      "grad_norm": 0.3887118697166443,
+      "learning_rate": 0.0003753531468531468,
+      "loss": 3.2629,
+      "step": 64350
+    },
+    {
+      "epoch": 18.75368396528627,
+      "grad_norm": 0.40216881036758423,
+      "learning_rate": 0.00037517832167832167,
+      "loss": 3.2633,
+      "step": 64400
+    },
+    {
+      "epoch": 18.768245092899996,
+      "grad_norm": 0.4059422016143799,
+      "learning_rate": 0.0003750034965034965,
+      "loss": 3.2735,
+      "step": 64450
+    },
+    {
+      "epoch": 18.782806220513717,
+      "grad_norm": 0.39602193236351013,
+      "learning_rate": 0.00037482867132867127,
+      "loss": 3.2626,
+      "step": 64500
+    },
+    {
+      "epoch": 18.797367348127437,
+      "grad_norm": 0.39334166049957275,
+      "learning_rate": 0.0003746538461538462,
+      "loss": 3.2659,
+      "step": 64550
+    },
+    {
+      "epoch": 18.811928475741162,
+      "grad_norm": 0.3972336947917938,
+      "learning_rate": 0.0003744790209790209,
+      "loss": 3.2565,
+      "step": 64600
+    },
+    {
+      "epoch": 18.826489603354883,
+      "grad_norm": 0.3721438944339752,
+      "learning_rate": 0.0003743041958041958,
+      "loss": 3.2685,
+      "step": 64650
+    },
+    {
+      "epoch": 18.841050730968607,
+      "grad_norm": 0.4226655066013336,
+      "learning_rate": 0.0003741293706293706,
+      "loss": 3.2777,
+      "step": 64700
+    },
+    {
+      "epoch": 18.85561185858233,
+      "grad_norm": 0.38680168986320496,
+      "learning_rate": 0.0003739545454545454,
+      "loss": 3.2635,
+      "step": 64750
+    },
+    {
+      "epoch": 18.87017298619605,
+      "grad_norm": 0.3778873682022095,
+      "learning_rate": 0.0003737797202797202,
+      "loss": 3.2742,
+      "step": 64800
+    },
+    {
+      "epoch": 18.884734113809774,
+      "grad_norm": 0.38358810544013977,
+      "learning_rate": 0.0003736048951048951,
+      "loss": 3.2617,
+      "step": 64850
+    },
+    {
+      "epoch": 18.899295241423495,
+      "grad_norm": 0.4202938377857208,
+      "learning_rate": 0.0003734300699300699,
+      "loss": 3.268,
+      "step": 64900
+    },
+    {
+      "epoch": 18.91385636903722,
+      "grad_norm": 0.37639421224594116,
+      "learning_rate": 0.00037325524475524473,
+      "loss": 3.2762,
+      "step": 64950
+    },
+    {
+      "epoch": 18.92841749665094,
+      "grad_norm": 0.3673178553581238,
+      "learning_rate": 0.00037308041958041953,
+      "loss": 3.2763,
+      "step": 65000
+    },
+    {
+      "epoch": 18.92841749665094,
+      "eval_accuracy": 0.3738111187422306,
+      "eval_loss": 3.5290017127990723,
+      "eval_runtime": 181.6306,
+      "eval_samples_per_second": 91.648,
+      "eval_steps_per_second": 5.731,
+      "step": 65000
+    },
+    {
+      "epoch": 18.94297862426466,
+      "grad_norm": 0.3887439966201782,
+      "learning_rate": 0.0003729055944055944,
+      "loss": 3.269,
+      "step": 65050
+    },
+    {
+      "epoch": 18.957539751878386,
+      "grad_norm": 0.3968895971775055,
+      "learning_rate": 0.0003727307692307692,
+      "loss": 3.2634,
+      "step": 65100
+    },
+    {
+      "epoch": 18.972100879492107,
+      "grad_norm": 0.35834673047065735,
+      "learning_rate": 0.00037255594405594404,
+      "loss": 3.2796,
+      "step": 65150
+    },
+    {
+      "epoch": 18.98666200710583,
+      "grad_norm": 0.39905858039855957,
+      "learning_rate": 0.0003723811188811189,
+      "loss": 3.2806,
+      "step": 65200
+    },
+    {
+      "epoch": 19.001164890209097,
+      "grad_norm": 0.4308378994464874,
+      "learning_rate": 0.00037220629370629364,
+      "loss": 3.2609,
+      "step": 65250
+    },
+    {
+      "epoch": 19.01572601782282,
+      "grad_norm": 0.382627010345459,
+      "learning_rate": 0.00037203146853146855,
+      "loss": 3.1625,
+      "step": 65300
+    },
+    {
+      "epoch": 19.030287145436542,
+      "grad_norm": 0.3864773213863373,
+      "learning_rate": 0.0003718566433566433,
+      "loss": 3.161,
+      "step": 65350
+    },
+    {
+      "epoch": 19.044848273050263,
+      "grad_norm": 0.44940927624702454,
+      "learning_rate": 0.00037168181818181815,
+      "loss": 3.1772,
+      "step": 65400
+    },
+    {
+      "epoch": 19.059409400663988,
+      "grad_norm": 0.38539499044418335,
+      "learning_rate": 0.00037150699300699295,
+      "loss": 3.1838,
+      "step": 65450
+    },
+    {
+      "epoch": 19.07397052827771,
+      "grad_norm": 0.3811834454536438,
+      "learning_rate": 0.0003713321678321678,
+      "loss": 3.1818,
+      "step": 65500
+    },
+    {
+      "epoch": 19.088531655891433,
+      "grad_norm": 0.3792768120765686,
+      "learning_rate": 0.0003711573426573426,
+      "loss": 3.178,
+      "step": 65550
+    },
+    {
+      "epoch": 19.103092783505154,
+      "grad_norm": 0.40748918056488037,
+      "learning_rate": 0.00037098251748251746,
+      "loss": 3.1902,
+      "step": 65600
+    },
+    {
+      "epoch": 19.11765391111888,
+      "grad_norm": 0.3937475383281708,
+      "learning_rate": 0.00037080769230769226,
+      "loss": 3.1973,
+      "step": 65650
+    },
+    {
+      "epoch": 19.1322150387326,
+      "grad_norm": 0.3816971778869629,
+      "learning_rate": 0.0003706328671328671,
+      "loss": 3.1828,
+      "step": 65700
+    },
+    {
+      "epoch": 19.14677616634632,
+      "grad_norm": 0.4283738136291504,
+      "learning_rate": 0.0003704580419580419,
+      "loss": 3.1961,
+      "step": 65750
+    },
+    {
+      "epoch": 19.161337293960045,
+      "grad_norm": 0.40591931343078613,
+      "learning_rate": 0.00037028321678321676,
+      "loss": 3.1943,
+      "step": 65800
+    },
+    {
+      "epoch": 19.175898421573766,
+      "grad_norm": 0.43688419461250305,
+      "learning_rate": 0.0003701083916083916,
+      "loss": 3.2046,
+      "step": 65850
+    },
+    {
+      "epoch": 19.19045954918749,
+      "grad_norm": 0.42176738381385803,
+      "learning_rate": 0.0003699335664335664,
+      "loss": 3.1941,
+      "step": 65900
+    },
+    {
+      "epoch": 19.20502067680121,
+      "grad_norm": 0.4293349087238312,
+      "learning_rate": 0.00036975874125874127,
+      "loss": 3.1965,
+      "step": 65950
+    },
+    {
+      "epoch": 19.219581804414933,
+      "grad_norm": 0.401883989572525,
+      "learning_rate": 0.00036958391608391607,
+      "loss": 3.2052,
+      "step": 66000
+    },
+    {
+      "epoch": 19.219581804414933,
+      "eval_accuracy": 0.3729365705059401,
+      "eval_loss": 3.549788236618042,
+      "eval_runtime": 181.5209,
+      "eval_samples_per_second": 91.703,
+      "eval_steps_per_second": 5.735,
+      "step": 66000
+    },
+    {
+      "epoch": 19.234142932028657,
+      "grad_norm": 0.38579362630844116,
+      "learning_rate": 0.0003694090909090909,
+      "loss": 3.2108,
+      "step": 66050
+    },
+    {
+      "epoch": 19.248704059642378,
+      "grad_norm": 0.3706747889518738,
+      "learning_rate": 0.00036923426573426567,
+      "loss": 3.1956,
+      "step": 66100
+    },
+    {
+      "epoch": 19.263265187256103,
+      "grad_norm": 0.40372592210769653,
+      "learning_rate": 0.0003690594405594405,
+      "loss": 3.206,
+      "step": 66150
+    },
+    {
+      "epoch": 19.277826314869824,
+      "grad_norm": 0.37341922521591187,
+      "learning_rate": 0.0003688846153846153,
+      "loss": 3.2149,
+      "step": 66200
+    },
+    {
+      "epoch": 19.292387442483545,
+      "grad_norm": 0.40485063195228577,
+      "learning_rate": 0.0003687097902097902,
+      "loss": 3.2128,
+      "step": 66250
+    },
+    {
+      "epoch": 19.30694857009727,
+      "grad_norm": 0.37851768732070923,
+      "learning_rate": 0.000368534965034965,
+      "loss": 3.2091,
+      "step": 66300
+    },
+    {
+      "epoch": 19.32150969771099,
+      "grad_norm": 0.3929597735404968,
+      "learning_rate": 0.00036836013986013983,
+      "loss": 3.2174,
+      "step": 66350
+    },
+    {
+      "epoch": 19.336070825324715,
+      "grad_norm": 0.4164958894252777,
+      "learning_rate": 0.00036818531468531463,
+      "loss": 3.2248,
+      "step": 66400
+    },
+    {
+      "epoch": 19.350631952938436,
+      "grad_norm": 0.3765912353992462,
+      "learning_rate": 0.0003680104895104895,
+      "loss": 3.226,
+      "step": 66450
+    },
+    {
+      "epoch": 19.365193080552157,
+      "grad_norm": 0.4018316864967346,
+      "learning_rate": 0.0003678356643356643,
+      "loss": 3.2108,
+      "step": 66500
+    },
+    {
+      "epoch": 19.37975420816588,
+      "grad_norm": 0.3959610164165497,
+      "learning_rate": 0.00036766083916083914,
+      "loss": 3.2259,
+      "step": 66550
+    },
+    {
+      "epoch": 19.394315335779602,
+      "grad_norm": 0.39516445994377136,
+      "learning_rate": 0.000367486013986014,
+      "loss": 3.2286,
+      "step": 66600
+    },
+    {
+      "epoch": 19.408876463393327,
+      "grad_norm": 0.38694578409194946,
+      "learning_rate": 0.0003673111888111888,
+      "loss": 3.2329,
+      "step": 66650
+    },
+    {
+      "epoch": 19.423437591007048,
+      "grad_norm": 0.38315701484680176,
+      "learning_rate": 0.00036713636363636365,
+      "loss": 3.2293,
+      "step": 66700
+    },
+    {
+      "epoch": 19.43799871862077,
+      "grad_norm": 0.3903602063655853,
+      "learning_rate": 0.00036696153846153844,
+      "loss": 3.2431,
+      "step": 66750
+    },
+    {
+      "epoch": 19.452559846234493,
+      "grad_norm": 0.3970508277416229,
+      "learning_rate": 0.0003667867132867133,
+      "loss": 3.2321,
+      "step": 66800
+    },
+    {
+      "epoch": 19.467120973848214,
+      "grad_norm": 0.39490050077438354,
+      "learning_rate": 0.00036661188811188804,
+      "loss": 3.2436,
+      "step": 66850
+    },
+    {
+      "epoch": 19.48168210146194,
+      "grad_norm": 0.3896945118904114,
+      "learning_rate": 0.0003664370629370629,
+      "loss": 3.2351,
+      "step": 66900
+    },
+    {
+      "epoch": 19.49624322907566,
+      "grad_norm": 0.4222089946269989,
+      "learning_rate": 0.0003662622377622377,
+      "loss": 3.2298,
+      "step": 66950
+    },
+    {
+      "epoch": 19.51080435668938,
+      "grad_norm": 0.4079953134059906,
+      "learning_rate": 0.00036608741258741255,
+      "loss": 3.2232,
+      "step": 67000
+    },
+    {
+      "epoch": 19.51080435668938,
+      "eval_accuracy": 0.37318897742398227,
+      "eval_loss": 3.5430922508239746,
+      "eval_runtime": 181.509,
+      "eval_samples_per_second": 91.709,
+      "eval_steps_per_second": 5.735,
+      "step": 67000
+    },
+    {
+      "epoch": 19.525365484303105,
+      "grad_norm": 0.40674111247062683,
+      "learning_rate": 0.00036591258741258735,
+      "loss": 3.2444,
+      "step": 67050
+    },
+    {
+      "epoch": 19.539926611916826,
+      "grad_norm": 0.3936384618282318,
+      "learning_rate": 0.0003657377622377622,
+      "loss": 3.2507,
+      "step": 67100
+    },
+    {
+      "epoch": 19.55448773953055,
+      "grad_norm": 0.39949601888656616,
+      "learning_rate": 0.000365562937062937,
+      "loss": 3.2333,
+      "step": 67150
+    },
+    {
+      "epoch": 19.56904886714427,
+      "grad_norm": 0.4255499839782715,
+      "learning_rate": 0.00036538811188811186,
+      "loss": 3.2361,
+      "step": 67200
+    },
+    {
+      "epoch": 19.583609994757992,
+      "grad_norm": 0.40485048294067383,
+      "learning_rate": 0.0003652132867132867,
+      "loss": 3.2366,
+      "step": 67250
+    },
+    {
+      "epoch": 19.598171122371717,
+      "grad_norm": 0.4169028699398041,
+      "learning_rate": 0.0003650384615384615,
+      "loss": 3.2392,
+      "step": 67300
+    },
+    {
+      "epoch": 19.612732249985438,
+      "grad_norm": 0.37912827730178833,
+      "learning_rate": 0.00036486363636363637,
+      "loss": 3.2355,
+      "step": 67350
+    },
+    {
+      "epoch": 19.627293377599162,
+      "grad_norm": 0.3838323652744293,
+      "learning_rate": 0.00036468881118881117,
+      "loss": 3.2375,
+      "step": 67400
+    },
+    {
+      "epoch": 19.641854505212883,
+      "grad_norm": 0.3855572044849396,
+      "learning_rate": 0.000364513986013986,
+      "loss": 3.2405,
+      "step": 67450
+    },
+    {
+      "epoch": 19.656415632826604,
+      "grad_norm": 0.3959762752056122,
+      "learning_rate": 0.0003643391608391608,
+      "loss": 3.2475,
+      "step": 67500
+    },
+    {
+      "epoch": 19.67097676044033,
+      "grad_norm": 0.3651852607727051,
+      "learning_rate": 0.0003641643356643357,
+      "loss": 3.2349,
+      "step": 67550
+    },
+    {
+      "epoch": 19.68553788805405,
+      "grad_norm": 0.39558374881744385,
+      "learning_rate": 0.0003639895104895104,
+      "loss": 3.2591,
+      "step": 67600
+    },
+    {
+      "epoch": 19.700099015667774,
+      "grad_norm": 0.39178308844566345,
+      "learning_rate": 0.0003638146853146853,
+      "loss": 3.251,
+      "step": 67650
+    },
+    {
+      "epoch": 19.714660143281495,
+      "grad_norm": 0.3785051107406616,
+      "learning_rate": 0.00036363986013986007,
+      "loss": 3.2491,
+      "step": 67700
+    },
+    {
+      "epoch": 19.729221270895216,
+      "grad_norm": 0.42383629083633423,
+      "learning_rate": 0.0003634650349650349,
+      "loss": 3.246,
+      "step": 67750
+    },
+    {
+      "epoch": 19.74378239850894,
+      "grad_norm": 0.3946218192577362,
+      "learning_rate": 0.0003632902097902097,
+      "loss": 3.2454,
+      "step": 67800
+    },
+    {
+      "epoch": 19.758343526122662,
+      "grad_norm": 0.41260185837745667,
+      "learning_rate": 0.0003631153846153846,
+      "loss": 3.2526,
+      "step": 67850
+    },
+    {
+      "epoch": 19.772904653736386,
+      "grad_norm": 0.36741843819618225,
+      "learning_rate": 0.00036294055944055943,
+      "loss": 3.2621,
+      "step": 67900
+    },
+    {
+      "epoch": 19.787465781350107,
+      "grad_norm": 0.37260907888412476,
+      "learning_rate": 0.00036276573426573423,
+      "loss": 3.2524,
+      "step": 67950
+    },
+    {
+      "epoch": 19.802026908963832,
+      "grad_norm": 0.393512487411499,
+      "learning_rate": 0.0003625909090909091,
+      "loss": 3.2532,
+      "step": 68000
+    },
+    {
+      "epoch": 19.802026908963832,
+      "eval_accuracy": 0.37361055693404244,
+      "eval_loss": 3.5359911918640137,
+      "eval_runtime": 181.8052,
+      "eval_samples_per_second": 91.56,
+      "eval_steps_per_second": 5.726,
+      "step": 68000
+    },
+    {
+      "epoch": 19.816588036577553,
+      "grad_norm": 0.41025984287261963,
+      "learning_rate": 0.0003624160839160839,
+      "loss": 3.2568,
+      "step": 68050
+    },
+    {
+      "epoch": 19.831149164191274,
+      "grad_norm": 0.37306198477745056,
+      "learning_rate": 0.00036224125874125874,
+      "loss": 3.252,
+      "step": 68100
+    },
+    {
+      "epoch": 19.845710291805,
+      "grad_norm": 0.41727498173713684,
+      "learning_rate": 0.00036206643356643354,
+      "loss": 3.2563,
+      "step": 68150
+    },
+    {
+      "epoch": 19.86027141941872,
+      "grad_norm": 0.41464343667030334,
+      "learning_rate": 0.0003618916083916084,
+      "loss": 3.2577,
+      "step": 68200
+    },
+    {
+      "epoch": 19.874832547032444,
+      "grad_norm": 0.40288522839546204,
+      "learning_rate": 0.0003617167832167832,
+      "loss": 3.2604,
+      "step": 68250
+    },
+    {
+      "epoch": 19.889393674646165,
+      "grad_norm": 0.3966323733329773,
+      "learning_rate": 0.00036154195804195805,
+      "loss": 3.2573,
+      "step": 68300
+    },
+    {
+      "epoch": 19.903954802259886,
+      "grad_norm": 0.4013204574584961,
+      "learning_rate": 0.0003613671328671328,
+      "loss": 3.2623,
+      "step": 68350
+    },
+    {
+      "epoch": 19.91851592987361,
+      "grad_norm": 0.4066911041736603,
+      "learning_rate": 0.00036119230769230765,
+      "loss": 3.2603,
+      "step": 68400
+    },
+    {
+      "epoch": 19.93307705748733,
+      "grad_norm": 0.4191489517688751,
+      "learning_rate": 0.00036101748251748245,
+      "loss": 3.26,
+      "step": 68450
+    },
+    {
+      "epoch": 19.947638185101056,
+      "grad_norm": 0.39228692650794983,
+      "learning_rate": 0.0003608426573426573,
+      "loss": 3.2565,
+      "step": 68500
+    },
+    {
+      "epoch": 19.962199312714777,
+      "grad_norm": 0.40354210138320923,
+      "learning_rate": 0.0003606678321678321,
+      "loss": 3.2659,
+      "step": 68550
+    },
+    {
+      "epoch": 19.976760440328498,
+      "grad_norm": 0.424032986164093,
+      "learning_rate": 0.00036049300699300696,
+      "loss": 3.2405,
+      "step": 68600
+    },
+    {
+      "epoch": 19.991321567942222,
+      "grad_norm": 0.40421053767204285,
+      "learning_rate": 0.0003603181818181818,
+      "loss": 3.2657,
+      "step": 68650
+    },
+    {
+      "epoch": 20.005824451045488,
+      "grad_norm": 0.37901127338409424,
+      "learning_rate": 0.0003601433566433566,
+      "loss": 3.2116,
+      "step": 68700
+    },
+    {
+      "epoch": 20.020385578659212,
+      "grad_norm": 0.39838650822639465,
+      "learning_rate": 0.00035996853146853146,
+      "loss": 3.1563,
+      "step": 68750
+    },
+    {
+      "epoch": 20.034946706272933,
+      "grad_norm": 0.3853570818901062,
+      "learning_rate": 0.00035979370629370626,
+      "loss": 3.1651,
+      "step": 68800
+    },
+    {
+      "epoch": 20.049507833886658,
+      "grad_norm": 0.37622562050819397,
+      "learning_rate": 0.0003596188811188811,
+      "loss": 3.1671,
+      "step": 68850
+    },
+    {
+      "epoch": 20.06406896150038,
+      "grad_norm": 0.40107908844947815,
+      "learning_rate": 0.0003594440559440559,
+      "loss": 3.171,
+      "step": 68900
+    },
+    {
+      "epoch": 20.0786300891141,
+      "grad_norm": 0.3853244185447693,
+      "learning_rate": 0.00035926923076923077,
+      "loss": 3.1719,
+      "step": 68950
+    },
+    {
+      "epoch": 20.093191216727824,
+      "grad_norm": 0.437876433134079,
+      "learning_rate": 0.00035909440559440557,
+      "loss": 3.1732,
+      "step": 69000
+    },
+    {
+      "epoch": 20.093191216727824,
+      "eval_accuracy": 0.37331006690958235,
+      "eval_loss": 3.5458712577819824,
+      "eval_runtime": 181.7798,
+      "eval_samples_per_second": 91.572,
+      "eval_steps_per_second": 5.727,
+      "step": 69000
+    },
+    {
+      "epoch": 20.107752344341545,
+      "grad_norm": 0.4051014482975006,
+      "learning_rate": 0.0003589195804195804,
+      "loss": 3.1833,
+      "step": 69050
+    },
+    {
+      "epoch": 20.12231347195527,
+      "grad_norm": 0.42431265115737915,
+      "learning_rate": 0.00035874475524475517,
+      "loss": 3.1786,
+      "step": 69100
+    },
+    {
+      "epoch": 20.13687459956899,
+      "grad_norm": 0.3915984332561493,
+      "learning_rate": 0.00035856993006993,
+      "loss": 3.1951,
+      "step": 69150
+    },
+    {
+      "epoch": 20.15143572718271,
+      "grad_norm": 0.39397016167640686,
+      "learning_rate": 0.0003583951048951048,
+      "loss": 3.1773,
+      "step": 69200
+    },
+    {
+      "epoch": 20.165996854796436,
+      "grad_norm": 0.41162973642349243,
+      "learning_rate": 0.0003582202797202797,
+      "loss": 3.1971,
+      "step": 69250
+    },
+    {
+      "epoch": 20.180557982410157,
+      "grad_norm": 0.40703409910202026,
+      "learning_rate": 0.00035804545454545453,
+      "loss": 3.1815,
+      "step": 69300
+    },
+    {
+      "epoch": 20.19511911002388,
+      "grad_norm": 0.4034010171890259,
+      "learning_rate": 0.00035787062937062933,
+      "loss": 3.1882,
+      "step": 69350
+    },
+    {
+      "epoch": 20.209680237637603,
+      "grad_norm": 0.40180739760398865,
+      "learning_rate": 0.0003576958041958042,
+      "loss": 3.1876,
+      "step": 69400
+    },
+    {
+      "epoch": 20.224241365251324,
+      "grad_norm": 0.414236456155777,
+      "learning_rate": 0.000357520979020979,
+      "loss": 3.1873,
+      "step": 69450
+    },
+    {
+      "epoch": 20.238802492865048,
+      "grad_norm": 0.3842925727367401,
+      "learning_rate": 0.00035734615384615384,
+      "loss": 3.2091,
+      "step": 69500
+    },
+    {
+      "epoch": 20.25336362047877,
+      "grad_norm": 0.3949846625328064,
+      "learning_rate": 0.00035717132867132864,
+      "loss": 3.2013,
+      "step": 69550
+    },
+    {
+      "epoch": 20.267924748092494,
+      "grad_norm": 0.4168473482131958,
+      "learning_rate": 0.0003569965034965035,
+      "loss": 3.2028,
+      "step": 69600
+    },
+    {
+      "epoch": 20.282485875706215,
+      "grad_norm": 0.3983443081378937,
+      "learning_rate": 0.0003568216783216783,
+      "loss": 3.1992,
+      "step": 69650
+    },
+    {
+      "epoch": 20.297047003319935,
+      "grad_norm": 0.4090621769428253,
+      "learning_rate": 0.00035664685314685314,
+      "loss": 3.1973,
+      "step": 69700
+    },
+    {
+      "epoch": 20.31160813093366,
+      "grad_norm": 0.4229073226451874,
+      "learning_rate": 0.00035647202797202794,
+      "loss": 3.2145,
+      "step": 69750
+    },
+    {
+      "epoch": 20.32616925854738,
+      "grad_norm": 0.4256856143474579,
+      "learning_rate": 0.0003562972027972028,
+      "loss": 3.2116,
+      "step": 69800
+    },
+    {
+      "epoch": 20.340730386161106,
+      "grad_norm": 0.40367722511291504,
+      "learning_rate": 0.00035612237762237754,
+      "loss": 3.2001,
+      "step": 69850
+    },
+    {
+      "epoch": 20.355291513774826,
+      "grad_norm": 0.43100783228874207,
+      "learning_rate": 0.0003559475524475524,
+      "loss": 3.2091,
+      "step": 69900
+    },
+    {
+      "epoch": 20.369852641388547,
+      "grad_norm": 0.4087181091308594,
+      "learning_rate": 0.0003557727272727272,
+      "loss": 3.2042,
+      "step": 69950
+    },
+    {
+      "epoch": 20.384413769002272,
+      "grad_norm": 0.3920936584472656,
+      "learning_rate": 0.00035559790209790205,
+      "loss": 3.2108,
+      "step": 70000
+    },
+    {
+      "epoch": 20.384413769002272,
+      "eval_accuracy": 0.3731025689075589,
+      "eval_loss": 3.5489280223846436,
+      "eval_runtime": 182.0706,
+      "eval_samples_per_second": 91.426,
+      "eval_steps_per_second": 5.718,
+      "step": 70000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 171700,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 50,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 20,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 5
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4631308623872e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}