diff --git "a/cost_to_hit_frequency_40817/checkpoint-50000/trainer_state.json" "b/cost_to_hit_frequency_40817/checkpoint-50000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/cost_to_hit_frequency_40817/checkpoint-50000/trainer_state.json"
@@ -0,0 +1,7493 @@
+{
+  "best_global_step": 49000,
+  "best_metric": 3.5516107082366943,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/cost_to_hit_frequency_40817/checkpoint-20000",
+  "epoch": 15.114900822447993,
+  "eval_steps": 1000,
+  "global_step": 50000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.015118529269472665,
+      "grad_norm": 0.6569534540176392,
+      "learning_rate": 0.000294,
+      "loss": 8.48,
+      "step": 50
+    },
+    {
+      "epoch": 0.03023705853894533,
+      "grad_norm": 0.9759527444839478,
+      "learning_rate": 0.0005939999999999999,
+      "loss": 6.7387,
+      "step": 100
+    },
+    {
+      "epoch": 0.045355587808417995,
+      "grad_norm": 0.4822550415992737,
+      "learning_rate": 0.0005995549500454132,
+      "loss": 6.3406,
+      "step": 150
+    },
+    {
+      "epoch": 0.06047411707789066,
+      "grad_norm": 0.4631986618041992,
+      "learning_rate": 0.000599100817438692,
+      "loss": 6.137,
+      "step": 200
+    },
+    {
+      "epoch": 0.07559264634736333,
+      "grad_norm": 0.4041077494621277,
+      "learning_rate": 0.0005986466848319708,
+      "loss": 5.9853,
+      "step": 250
+    },
+    {
+      "epoch": 0.09071117561683599,
+      "grad_norm": 0.3731299042701721,
+      "learning_rate": 0.0005981925522252497,
+      "loss": 5.8362,
+      "step": 300
+    },
+    {
+      "epoch": 0.10582970488630866,
+      "grad_norm": 0.44940078258514404,
+      "learning_rate": 0.0005977384196185286,
+      "loss": 5.7301,
+      "step": 350
+    },
+    {
+      "epoch": 0.12094823415578132,
+      "grad_norm": 0.4535612165927887,
+      "learning_rate": 0.0005972842870118073,
+      "loss": 5.6198,
+      "step": 400
+    },
+    {
+      "epoch": 0.13606676342525398,
+      "grad_norm": 0.47181379795074463,
+      "learning_rate": 0.0005968301544050862,
+      "loss": 5.5126,
+      "step": 450
+    },
+    {
+      "epoch": 0.15118529269472666,
+      "grad_norm": 0.47652333974838257,
+      "learning_rate": 0.0005963760217983651,
+      "loss": 5.4013,
+      "step": 500
+    },
+    {
+      "epoch": 0.16630382196419932,
+      "grad_norm": 0.45077475905418396,
+      "learning_rate": 0.0005959218891916439,
+      "loss": 5.3135,
+      "step": 550
+    },
+    {
+      "epoch": 0.18142235123367198,
+      "grad_norm": 0.4728659987449646,
+      "learning_rate": 0.0005954677565849228,
+      "loss": 5.2565,
+      "step": 600
+    },
+    {
+      "epoch": 0.19654088050314467,
+      "grad_norm": 0.447640985250473,
+      "learning_rate": 0.0005950136239782016,
+      "loss": 5.1786,
+      "step": 650
+    },
+    {
+      "epoch": 0.21165940977261732,
+      "grad_norm": 0.4660065174102783,
+      "learning_rate": 0.0005945594913714804,
+      "loss": 5.1181,
+      "step": 700
+    },
+    {
+      "epoch": 0.22677793904208998,
+      "grad_norm": 0.4138477146625519,
+      "learning_rate": 0.0005941053587647593,
+      "loss": 5.0609,
+      "step": 750
+    },
+    {
+      "epoch": 0.24189646831156264,
+      "grad_norm": 0.43173298239707947,
+      "learning_rate": 0.0005936512261580381,
+      "loss": 5.0094,
+      "step": 800
+    },
+    {
+      "epoch": 0.2570149975810353,
+      "grad_norm": 0.41261234879493713,
+      "learning_rate": 0.0005931970935513169,
+      "loss": 4.9896,
+      "step": 850
+    },
+    {
+      "epoch": 0.27213352685050796,
+      "grad_norm": 0.4491656720638275,
+      "learning_rate": 0.0005927429609445958,
+      "loss": 4.9266,
+      "step": 900
+    },
+    {
+      "epoch": 0.28725205611998067,
+      "grad_norm": 0.4212245047092438,
+      "learning_rate": 0.0005922888283378746,
+      "loss": 4.8791,
+      "step": 950
+    },
+    {
+      "epoch": 0.30237058538945333,
+      "grad_norm": 0.4332761764526367,
+      "learning_rate": 0.0005918346957311534,
+      "loss": 4.8382,
+      "step": 1000
+    },
+    {
+      "epoch": 0.30237058538945333,
+      "eval_accuracy": 0.2525349543872818,
+      "eval_loss": 4.76720666885376,
+      "eval_runtime": 182.6488,
+      "eval_samples_per_second": 91.126,
+      "eval_steps_per_second": 5.699,
+      "step": 1000
+    },
+    {
+      "epoch": 0.317489114658926,
+      "grad_norm": 0.5782724022865295,
+      "learning_rate": 0.0005913805631244322,
+      "loss": 4.7861,
+      "step": 1050
+    },
+    {
+      "epoch": 0.33260764392839864,
+      "grad_norm": 0.5215025544166565,
+      "learning_rate": 0.0005909264305177112,
+      "loss": 4.7449,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3477261731978713,
+      "grad_norm": 0.5074264407157898,
+      "learning_rate": 0.0005904722979109899,
+      "loss": 4.7092,
+      "step": 1150
+    },
+    {
+      "epoch": 0.36284470246734396,
+      "grad_norm": 0.4659997820854187,
+      "learning_rate": 0.0005900181653042687,
+      "loss": 4.6644,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3779632317368166,
+      "grad_norm": 0.46156764030456543,
+      "learning_rate": 0.0005895640326975477,
+      "loss": 4.6383,
+      "step": 1250
+    },
+    {
+      "epoch": 0.39308176100628933,
+      "grad_norm": 0.43247443437576294,
+      "learning_rate": 0.0005891099000908265,
+      "loss": 4.5939,
+      "step": 1300
+    },
+    {
+      "epoch": 0.408200290275762,
+      "grad_norm": 0.43695470690727234,
+      "learning_rate": 0.0005886557674841053,
+      "loss": 4.545,
+      "step": 1350
+    },
+    {
+      "epoch": 0.42331881954523465,
+      "grad_norm": 0.4135619103908539,
+      "learning_rate": 0.0005882016348773842,
+      "loss": 4.5432,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4384373488147073,
+      "grad_norm": 0.3893912136554718,
+      "learning_rate": 0.000587747502270663,
+      "loss": 4.5358,
+      "step": 1450
+    },
+    {
+      "epoch": 0.45355587808417996,
+      "grad_norm": 0.4318796694278717,
+      "learning_rate": 0.0005872933696639418,
+      "loss": 4.5063,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4686744073536526,
+      "grad_norm": 0.404938668012619,
+      "learning_rate": 0.0005868392370572207,
+      "loss": 4.4833,
+      "step": 1550
+    },
+    {
+      "epoch": 0.4837929366231253,
+      "grad_norm": 0.39407220482826233,
+      "learning_rate": 0.0005863851044504995,
+      "loss": 4.4573,
+      "step": 1600
+    },
+    {
+      "epoch": 0.498911465892598,
+      "grad_norm": 0.4312690496444702,
+      "learning_rate": 0.0005859309718437783,
+      "loss": 4.4493,
+      "step": 1650
+    },
+    {
+      "epoch": 0.5140299951620706,
+      "grad_norm": 0.42534348368644714,
+      "learning_rate": 0.0005854768392370572,
+      "loss": 4.4358,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5291485244315433,
+      "grad_norm": 0.3989129066467285,
+      "learning_rate": 0.000585022706630336,
+      "loss": 4.4004,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5442670537010159,
+      "grad_norm": 0.4206331968307495,
+      "learning_rate": 0.0005845685740236149,
+      "loss": 4.3903,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5593855829704887,
+      "grad_norm": 0.39128994941711426,
+      "learning_rate": 0.0005841144414168936,
+      "loss": 4.3716,
+      "step": 1850
+    },
+    {
+      "epoch": 0.5745041122399613,
+      "grad_norm": 0.4127030670642853,
+      "learning_rate": 0.0005836603088101725,
+      "loss": 4.3646,
+      "step": 1900
+    },
+    {
+      "epoch": 0.589622641509434,
+      "grad_norm": 0.47868236899375916,
+      "learning_rate": 0.0005832061762034513,
+      "loss": 4.3474,
+      "step": 1950
+    },
+    {
+      "epoch": 0.6047411707789067,
+      "grad_norm": 0.41013041138648987,
+      "learning_rate": 0.0005827520435967303,
+      "loss": 4.3297,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6047411707789067,
+      "eval_accuracy": 0.29954742363508696,
+      "eval_loss": 4.283868312835693,
+      "eval_runtime": 182.004,
+      "eval_samples_per_second": 91.449,
+      "eval_steps_per_second": 5.72,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6198597000483793,
+      "grad_norm": 0.3787741959095001,
+      "learning_rate": 0.0005822979109900091,
+      "loss": 4.3142,
+      "step": 2050
+    },
+    {
+      "epoch": 0.634978229317852,
+      "grad_norm": 0.3691158890724182,
+      "learning_rate": 0.0005818437783832879,
+      "loss": 4.3103,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6500967585873246,
+      "grad_norm": 0.38457924127578735,
+      "learning_rate": 0.0005813896457765666,
+      "loss": 4.2851,
+      "step": 2150
+    },
+    {
+      "epoch": 0.6652152878567973,
+      "grad_norm": 0.40187668800354004,
+      "learning_rate": 0.0005809355131698456,
+      "loss": 4.2813,
+      "step": 2200
+    },
+    {
+      "epoch": 0.68033381712627,
+      "grad_norm": 0.38091689348220825,
+      "learning_rate": 0.0005804813805631244,
+      "loss": 4.2651,
+      "step": 2250
+    },
+    {
+      "epoch": 0.6954523463957426,
+      "grad_norm": 0.4001523554325104,
+      "learning_rate": 0.0005800272479564032,
+      "loss": 4.2552,
+      "step": 2300
+    },
+    {
+      "epoch": 0.7105708756652153,
+      "grad_norm": 0.3821207880973816,
+      "learning_rate": 0.0005795731153496821,
+      "loss": 4.2479,
+      "step": 2350
+    },
+    {
+      "epoch": 0.7256894049346879,
+      "grad_norm": 0.3614497482776642,
+      "learning_rate": 0.0005791189827429609,
+      "loss": 4.2514,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7408079342041606,
+      "grad_norm": 0.40655964612960815,
+      "learning_rate": 0.0005786648501362397,
+      "loss": 4.2292,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7559264634736332,
+      "grad_norm": 0.3501703441143036,
+      "learning_rate": 0.0005782107175295186,
+      "loss": 4.2267,
+      "step": 2500
+    },
+    {
+      "epoch": 0.771044992743106,
+      "grad_norm": 0.41427701711654663,
+      "learning_rate": 0.0005777565849227974,
+      "loss": 4.2247,
+      "step": 2550
+    },
+    {
+      "epoch": 0.7861635220125787,
+      "grad_norm": 0.44100141525268555,
+      "learning_rate": 0.0005773024523160762,
+      "loss": 4.2089,
+      "step": 2600
+    },
+    {
+      "epoch": 0.8012820512820513,
+      "grad_norm": 0.3879771828651428,
+      "learning_rate": 0.000576848319709355,
+      "loss": 4.2073,
+      "step": 2650
+    },
+    {
+      "epoch": 0.816400580551524,
+      "grad_norm": 0.3788333833217621,
+      "learning_rate": 0.0005763941871026339,
+      "loss": 4.2015,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8315191098209966,
+      "grad_norm": 0.366184800863266,
+      "learning_rate": 0.0005759400544959128,
+      "loss": 4.1719,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8466376390904693,
+      "grad_norm": 0.3733968436717987,
+      "learning_rate": 0.0005754859218891917,
+      "loss": 4.1749,
+      "step": 2800
+    },
+    {
+      "epoch": 0.861756168359942,
+      "grad_norm": 0.37004509568214417,
+      "learning_rate": 0.0005750317892824704,
+      "loss": 4.1642,
+      "step": 2850
+    },
+    {
+      "epoch": 0.8768746976294146,
+      "grad_norm": 0.36916959285736084,
+      "learning_rate": 0.0005745776566757492,
+      "loss": 4.1666,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8919932268988873,
+      "grad_norm": 0.39038950204849243,
+      "learning_rate": 0.0005741235240690281,
+      "loss": 4.1523,
+      "step": 2950
+    },
+    {
+      "epoch": 0.9071117561683599,
+      "grad_norm": 0.3550504744052887,
+      "learning_rate": 0.000573669391462307,
+      "loss": 4.1437,
+      "step": 3000
+    },
+    {
+      "epoch": 0.9071117561683599,
+      "eval_accuracy": 0.3157440890648464,
+      "eval_loss": 4.091999530792236,
+      "eval_runtime": 182.3379,
+      "eval_samples_per_second": 91.281,
+      "eval_steps_per_second": 5.709,
+      "step": 3000
+    },
+    {
+      "epoch": 0.9222302854378326,
+      "grad_norm": 0.3827008903026581,
+      "learning_rate": 0.0005732152588555858,
+      "loss": 4.1333,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9373488147073052,
+      "grad_norm": 0.379339337348938,
+      "learning_rate": 0.0005727611262488646,
+      "loss": 4.1309,
+      "step": 3100
+    },
+    {
+      "epoch": 0.9524673439767779,
+      "grad_norm": 0.38624098896980286,
+      "learning_rate": 0.0005723069936421435,
+      "loss": 4.1218,
+      "step": 3150
+    },
+    {
+      "epoch": 0.9675858732462506,
+      "grad_norm": 0.36264657974243164,
+      "learning_rate": 0.0005718528610354223,
+      "loss": 4.1127,
+      "step": 3200
+    },
+    {
+      "epoch": 0.9827044025157232,
+      "grad_norm": 0.3813261389732361,
+      "learning_rate": 0.0005713987284287011,
+      "loss": 4.1095,
+      "step": 3250
+    },
+    {
+      "epoch": 0.997822931785196,
+      "grad_norm": 0.46720796823501587,
+      "learning_rate": 0.00057094459582198,
+      "loss": 4.0923,
+      "step": 3300
+    },
+    {
+      "epoch": 1.012699564586357,
+      "grad_norm": 0.3777443766593933,
+      "learning_rate": 0.0005704904632152588,
+      "loss": 4.0438,
+      "step": 3350
+    },
+    {
+      "epoch": 1.0278180938558297,
+      "grad_norm": 0.35427603125572205,
+      "learning_rate": 0.0005700363306085376,
+      "loss": 4.0293,
+      "step": 3400
+    },
+    {
+      "epoch": 1.0429366231253023,
+      "grad_norm": 0.36464598774909973,
+      "learning_rate": 0.0005695821980018164,
+      "loss": 4.0156,
+      "step": 3450
+    },
+    {
+      "epoch": 1.058055152394775,
+      "grad_norm": 0.39198657870292664,
+      "learning_rate": 0.0005691280653950954,
+      "loss": 4.0147,
+      "step": 3500
+    },
+    {
+      "epoch": 1.0731736816642476,
+      "grad_norm": 0.3432103395462036,
+      "learning_rate": 0.0005686739327883741,
+      "loss": 4.0026,
+      "step": 3550
+    },
+    {
+      "epoch": 1.0882922109337203,
+      "grad_norm": 0.370108962059021,
+      "learning_rate": 0.0005682198001816529,
+      "loss": 4.0078,
+      "step": 3600
+    },
+    {
+      "epoch": 1.103410740203193,
+      "grad_norm": 0.3692167103290558,
+      "learning_rate": 0.0005677656675749318,
+      "loss": 4.0147,
+      "step": 3650
+    },
+    {
+      "epoch": 1.1185292694726656,
+      "grad_norm": 0.354033887386322,
+      "learning_rate": 0.0005673115349682107,
+      "loss": 3.9974,
+      "step": 3700
+    },
+    {
+      "epoch": 1.1336477987421383,
+      "grad_norm": 0.3713250160217285,
+      "learning_rate": 0.0005668574023614895,
+      "loss": 3.9928,
+      "step": 3750
+    },
+    {
+      "epoch": 1.148766328011611,
+      "grad_norm": 0.364036500453949,
+      "learning_rate": 0.0005664032697547684,
+      "loss": 4.0023,
+      "step": 3800
+    },
+    {
+      "epoch": 1.1638848572810838,
+      "grad_norm": 0.3609767258167267,
+      "learning_rate": 0.0005659491371480471,
+      "loss": 3.9908,
+      "step": 3850
+    },
+    {
+      "epoch": 1.1790033865505563,
+      "grad_norm": 0.3580855131149292,
+      "learning_rate": 0.000565495004541326,
+      "loss": 3.9918,
+      "step": 3900
+    },
+    {
+      "epoch": 1.1941219158200291,
+      "grad_norm": 0.35853010416030884,
+      "learning_rate": 0.0005650408719346049,
+      "loss": 3.9891,
+      "step": 3950
+    },
+    {
+      "epoch": 1.2092404450895016,
+      "grad_norm": 0.3546011745929718,
+      "learning_rate": 0.0005645867393278837,
+      "loss": 3.9999,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2092404450895016,
+      "eval_accuracy": 0.32564334461599675,
+      "eval_loss": 3.9847805500030518,
+      "eval_runtime": 182.3722,
+      "eval_samples_per_second": 91.264,
+      "eval_steps_per_second": 5.708,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2243589743589745,
+      "grad_norm": 0.3494962453842163,
+      "learning_rate": 0.0005641326067211625,
+      "loss": 3.9874,
+      "step": 4050
+    },
+    {
+      "epoch": 1.239477503628447,
+      "grad_norm": 0.37261122465133667,
+      "learning_rate": 0.0005636784741144414,
+      "loss": 3.9849,
+      "step": 4100
+    },
+    {
+      "epoch": 1.2545960328979198,
+      "grad_norm": 0.3606380224227905,
+      "learning_rate": 0.0005632243415077202,
+      "loss": 3.9697,
+      "step": 4150
+    },
+    {
+      "epoch": 1.2697145621673924,
+      "grad_norm": 0.3712509572505951,
+      "learning_rate": 0.000562770208900999,
+      "loss": 3.9746,
+      "step": 4200
+    },
+    {
+      "epoch": 1.284833091436865,
+      "grad_norm": 0.3441880941390991,
+      "learning_rate": 0.000562316076294278,
+      "loss": 3.9746,
+      "step": 4250
+    },
+    {
+      "epoch": 1.2999516207063377,
+      "grad_norm": 0.37153229117393494,
+      "learning_rate": 0.0005618619436875567,
+      "loss": 3.9607,
+      "step": 4300
+    },
+    {
+      "epoch": 1.3150701499758104,
+      "grad_norm": 0.3481687605381012,
+      "learning_rate": 0.0005614078110808355,
+      "loss": 3.9699,
+      "step": 4350
+    },
+    {
+      "epoch": 1.330188679245283,
+      "grad_norm": 0.3434485197067261,
+      "learning_rate": 0.0005609536784741143,
+      "loss": 3.955,
+      "step": 4400
+    },
+    {
+      "epoch": 1.3453072085147557,
+      "grad_norm": 0.36444127559661865,
+      "learning_rate": 0.0005604995458673933,
+      "loss": 3.959,
+      "step": 4450
+    },
+    {
+      "epoch": 1.3604257377842284,
+      "grad_norm": 0.3475540280342102,
+      "learning_rate": 0.0005600454132606721,
+      "loss": 3.9587,
+      "step": 4500
+    },
+    {
+      "epoch": 1.375544267053701,
+      "grad_norm": 0.35136139392852783,
+      "learning_rate": 0.0005595912806539509,
+      "loss": 3.9493,
+      "step": 4550
+    },
+    {
+      "epoch": 1.3906627963231737,
+      "grad_norm": 0.3854037821292877,
+      "learning_rate": 0.0005591371480472297,
+      "loss": 3.9523,
+      "step": 4600
+    },
+    {
+      "epoch": 1.4057813255926463,
+      "grad_norm": 0.32536762952804565,
+      "learning_rate": 0.0005586830154405086,
+      "loss": 3.9458,
+      "step": 4650
+    },
+    {
+      "epoch": 1.420899854862119,
+      "grad_norm": 0.33678901195526123,
+      "learning_rate": 0.0005582288828337874,
+      "loss": 3.9349,
+      "step": 4700
+    },
+    {
+      "epoch": 1.4360183841315917,
+      "grad_norm": 0.38056421279907227,
+      "learning_rate": 0.0005577747502270663,
+      "loss": 3.9357,
+      "step": 4750
+    },
+    {
+      "epoch": 1.4511369134010643,
+      "grad_norm": 0.3556210994720459,
+      "learning_rate": 0.0005573206176203451,
+      "loss": 3.9439,
+      "step": 4800
+    },
+    {
+      "epoch": 1.466255442670537,
+      "grad_norm": 0.3364984691143036,
+      "learning_rate": 0.0005568664850136239,
+      "loss": 3.9258,
+      "step": 4850
+    },
+    {
+      "epoch": 1.4813739719400096,
+      "grad_norm": 0.34836122393608093,
+      "learning_rate": 0.0005564123524069028,
+      "loss": 3.9111,
+      "step": 4900
+    },
+    {
+      "epoch": 1.4964925012094823,
+      "grad_norm": 0.32042673230171204,
+      "learning_rate": 0.0005559582198001816,
+      "loss": 3.9368,
+      "step": 4950
+    },
+    {
+      "epoch": 1.511611030478955,
+      "grad_norm": 0.34929823875427246,
+      "learning_rate": 0.0005555040871934604,
+      "loss": 3.9251,
+      "step": 5000
+    },
+    {
+      "epoch": 1.511611030478955,
+      "eval_accuracy": 0.33236250223983677,
+      "eval_loss": 3.909000873565674,
+      "eval_runtime": 182.2679,
+      "eval_samples_per_second": 91.316,
+      "eval_steps_per_second": 5.711,
+      "step": 5000
+    },
+    {
+      "epoch": 1.5267295597484276,
+      "grad_norm": 0.3525310158729553,
+      "learning_rate": 0.0005550499545867392,
+      "loss": 3.9314,
+      "step": 5050
+    },
+    {
+      "epoch": 1.5418480890179005,
+      "grad_norm": 0.34068188071250916,
+      "learning_rate": 0.0005545958219800181,
+      "loss": 3.9343,
+      "step": 5100
+    },
+    {
+      "epoch": 1.556966618287373,
+      "grad_norm": 0.3572525084018707,
+      "learning_rate": 0.0005541416893732969,
+      "loss": 3.9272,
+      "step": 5150
+    },
+    {
+      "epoch": 1.5720851475568458,
+      "grad_norm": 0.3401585519313812,
+      "learning_rate": 0.0005536875567665758,
+      "loss": 3.9159,
+      "step": 5200
+    },
+    {
+      "epoch": 1.5872036768263182,
+      "grad_norm": 0.3368934094905853,
+      "learning_rate": 0.0005532334241598547,
+      "loss": 3.9024,
+      "step": 5250
+    },
+    {
+      "epoch": 1.6023222060957911,
+      "grad_norm": 0.33803030848503113,
+      "learning_rate": 0.0005527792915531334,
+      "loss": 3.9133,
+      "step": 5300
+    },
+    {
+      "epoch": 1.6174407353652636,
+      "grad_norm": 0.3491084575653076,
+      "learning_rate": 0.0005523251589464122,
+      "loss": 3.9049,
+      "step": 5350
+    },
+    {
+      "epoch": 1.6325592646347364,
+      "grad_norm": 0.3191988170146942,
+      "learning_rate": 0.0005518710263396912,
+      "loss": 3.897,
+      "step": 5400
+    },
+    {
+      "epoch": 1.6476777939042089,
+      "grad_norm": 0.33832481503486633,
+      "learning_rate": 0.00055141689373297,
+      "loss": 3.8915,
+      "step": 5450
+    },
+    {
+      "epoch": 1.6627963231736818,
+      "grad_norm": 0.3492324650287628,
+      "learning_rate": 0.0005509627611262488,
+      "loss": 3.8975,
+      "step": 5500
+    },
+    {
+      "epoch": 1.6779148524431542,
+      "grad_norm": 0.3200375735759735,
+      "learning_rate": 0.0005505086285195277,
+      "loss": 3.8861,
+      "step": 5550
+    },
+    {
+      "epoch": 1.693033381712627,
+      "grad_norm": 0.3539038896560669,
+      "learning_rate": 0.0005500544959128065,
+      "loss": 3.8972,
+      "step": 5600
+    },
+    {
+      "epoch": 1.7081519109820995,
+      "grad_norm": 0.34983664751052856,
+      "learning_rate": 0.0005496003633060853,
+      "loss": 3.876,
+      "step": 5650
+    },
+    {
+      "epoch": 1.7232704402515724,
+      "grad_norm": 0.33722037076950073,
+      "learning_rate": 0.0005491462306993642,
+      "loss": 3.8811,
+      "step": 5700
+    },
+    {
+      "epoch": 1.738388969521045,
+      "grad_norm": 0.34622326493263245,
+      "learning_rate": 0.000548692098092643,
+      "loss": 3.8913,
+      "step": 5750
+    },
+    {
+      "epoch": 1.7535074987905177,
+      "grad_norm": 0.3671897351741791,
+      "learning_rate": 0.0005482379654859218,
+      "loss": 3.8912,
+      "step": 5800
+    },
+    {
+      "epoch": 1.7686260280599904,
+      "grad_norm": 0.3275216817855835,
+      "learning_rate": 0.0005477838328792006,
+      "loss": 3.8837,
+      "step": 5850
+    },
+    {
+      "epoch": 1.783744557329463,
+      "grad_norm": 0.3239729404449463,
+      "learning_rate": 0.0005473297002724795,
+      "loss": 3.8796,
+      "step": 5900
+    },
+    {
+      "epoch": 1.7988630865989357,
+      "grad_norm": 0.3572208881378174,
+      "learning_rate": 0.0005468755676657584,
+      "loss": 3.8735,
+      "step": 5950
+    },
+    {
+      "epoch": 1.8139816158684083,
+      "grad_norm": 0.3316764831542969,
+      "learning_rate": 0.0005464214350590371,
+      "loss": 3.8666,
+      "step": 6000
+    },
+    {
+      "epoch": 1.8139816158684083,
+      "eval_accuracy": 0.33760983430616326,
+      "eval_loss": 3.853510856628418,
+      "eval_runtime": 182.1185,
+      "eval_samples_per_second": 91.391,
+      "eval_steps_per_second": 5.716,
+      "step": 6000
+    },
+    {
+      "epoch": 1.829100145137881,
+      "grad_norm": 0.3352029025554657,
+      "learning_rate": 0.000545967302452316,
+      "loss": 3.8716,
+      "step": 6050
+    },
+    {
+      "epoch": 1.8442186744073537,
+      "grad_norm": 0.33704429864883423,
+      "learning_rate": 0.0005455131698455948,
+      "loss": 3.8742,
+      "step": 6100
+    },
+    {
+      "epoch": 1.8593372036768263,
+      "grad_norm": 0.32570329308509827,
+      "learning_rate": 0.0005450590372388737,
+      "loss": 3.8545,
+      "step": 6150
+    },
+    {
+      "epoch": 1.874455732946299,
+      "grad_norm": 0.3355255424976349,
+      "learning_rate": 0.0005446049046321526,
+      "loss": 3.8584,
+      "step": 6200
+    },
+    {
+      "epoch": 1.8895742622157716,
+      "grad_norm": 0.3264237642288208,
+      "learning_rate": 0.0005441507720254314,
+      "loss": 3.8647,
+      "step": 6250
+    },
+    {
+      "epoch": 1.9046927914852443,
+      "grad_norm": 0.34331536293029785,
+      "learning_rate": 0.0005436966394187102,
+      "loss": 3.8587,
+      "step": 6300
+    },
+    {
+      "epoch": 1.919811320754717,
+      "grad_norm": 0.3500491976737976,
+      "learning_rate": 0.0005432425068119891,
+      "loss": 3.853,
+      "step": 6350
+    },
+    {
+      "epoch": 1.9349298500241896,
+      "grad_norm": 0.33102309703826904,
+      "learning_rate": 0.0005427883742052679,
+      "loss": 3.8585,
+      "step": 6400
+    },
+    {
+      "epoch": 1.9500483792936623,
+      "grad_norm": 0.3271898329257965,
+      "learning_rate": 0.0005423342415985467,
+      "loss": 3.8577,
+      "step": 6450
+    },
+    {
+      "epoch": 1.965166908563135,
+      "grad_norm": 0.328890323638916,
+      "learning_rate": 0.0005418801089918256,
+      "loss": 3.8379,
+      "step": 6500
+    },
+    {
+      "epoch": 1.9802854378326078,
+      "grad_norm": 0.31617918610572815,
+      "learning_rate": 0.0005414259763851044,
+      "loss": 3.8392,
+      "step": 6550
+    },
+    {
+      "epoch": 1.9954039671020802,
+      "grad_norm": 0.324520081281662,
+      "learning_rate": 0.0005409718437783832,
+      "loss": 3.8461,
+      "step": 6600
+    },
+    {
+      "epoch": 2.0102805999032416,
+      "grad_norm": 0.324241042137146,
+      "learning_rate": 0.000540517711171662,
+      "loss": 3.7744,
+      "step": 6650
+    },
+    {
+      "epoch": 2.025399129172714,
+      "grad_norm": 0.36148601770401,
+      "learning_rate": 0.000540063578564941,
+      "loss": 3.7479,
+      "step": 6700
+    },
+    {
+      "epoch": 2.040517658442187,
+      "grad_norm": 0.32816359400749207,
+      "learning_rate": 0.0005396094459582197,
+      "loss": 3.74,
+      "step": 6750
+    },
+    {
+      "epoch": 2.0556361877116593,
+      "grad_norm": 0.33710262179374695,
+      "learning_rate": 0.0005391553133514985,
+      "loss": 3.7454,
+      "step": 6800
+    },
+    {
+      "epoch": 2.0707547169811322,
+      "grad_norm": 0.3189140558242798,
+      "learning_rate": 0.0005387011807447775,
+      "loss": 3.7555,
+      "step": 6850
+    },
+    {
+      "epoch": 2.0858732462506047,
+      "grad_norm": 0.34328320622444153,
+      "learning_rate": 0.0005382470481380563,
+      "loss": 3.7461,
+      "step": 6900
+    },
+    {
+      "epoch": 2.1009917755200775,
+      "grad_norm": 0.3447844982147217,
+      "learning_rate": 0.0005377929155313351,
+      "loss": 3.7688,
+      "step": 6950
+    },
+    {
+      "epoch": 2.11611030478955,
+      "grad_norm": 0.3423329293727875,
+      "learning_rate": 0.000537338782924614,
+      "loss": 3.748,
+      "step": 7000
+    },
+    {
+      "epoch": 2.11611030478955,
+      "eval_accuracy": 0.3412945715762478,
+      "eval_loss": 3.8153622150421143,
+      "eval_runtime": 182.3197,
+      "eval_samples_per_second": 91.29,
+      "eval_steps_per_second": 5.71,
+      "step": 7000
+    },
+    {
+      "epoch": 2.131228834059023,
+      "grad_norm": 0.3441483974456787,
+      "learning_rate": 0.0005368846503178928,
+      "loss": 3.7481,
+      "step": 7050
+    },
+    {
+      "epoch": 2.1463473633284953,
+      "grad_norm": 0.3269549608230591,
+      "learning_rate": 0.0005364305177111716,
+      "loss": 3.7588,
+      "step": 7100
+    },
+    {
+      "epoch": 2.161465892597968,
+      "grad_norm": 0.330601304769516,
+      "learning_rate": 0.0005359763851044505,
+      "loss": 3.7418,
+      "step": 7150
+    },
+    {
+      "epoch": 2.1765844218674406,
+      "grad_norm": 0.31895920634269714,
+      "learning_rate": 0.0005355222524977293,
+      "loss": 3.7563,
+      "step": 7200
+    },
+    {
+      "epoch": 2.1917029511369135,
+      "grad_norm": 0.33730292320251465,
+      "learning_rate": 0.0005350681198910081,
+      "loss": 3.7622,
+      "step": 7250
+    },
+    {
+      "epoch": 2.206821480406386,
+      "grad_norm": 0.32292884588241577,
+      "learning_rate": 0.000534613987284287,
+      "loss": 3.7546,
+      "step": 7300
+    },
+    {
+      "epoch": 2.221940009675859,
+      "grad_norm": 0.33400362730026245,
+      "learning_rate": 0.0005341598546775658,
+      "loss": 3.7655,
+      "step": 7350
+    },
+    {
+      "epoch": 2.2370585389453312,
+      "grad_norm": 0.3238161504268646,
+      "learning_rate": 0.0005337057220708446,
+      "loss": 3.7483,
+      "step": 7400
+    },
+    {
+      "epoch": 2.252177068214804,
+      "grad_norm": 0.3304459750652313,
+      "learning_rate": 0.0005332515894641234,
+      "loss": 3.7577,
+      "step": 7450
+    },
+    {
+      "epoch": 2.2672955974842766,
+      "grad_norm": 0.34201282262802124,
+      "learning_rate": 0.0005327974568574023,
+      "loss": 3.7453,
+      "step": 7500
+    },
+    {
+      "epoch": 2.2824141267537494,
+      "grad_norm": 0.3679095208644867,
+      "learning_rate": 0.0005323433242506811,
+      "loss": 3.7635,
+      "step": 7550
+    },
+    {
+      "epoch": 2.297532656023222,
+      "grad_norm": 0.3323363959789276,
+      "learning_rate": 0.00053188919164396,
+      "loss": 3.7442,
+      "step": 7600
+    },
+    {
+      "epoch": 2.3126511852926948,
+      "grad_norm": 0.3315403163433075,
+      "learning_rate": 0.0005314350590372389,
+      "loss": 3.751,
+      "step": 7650
+    },
+    {
+      "epoch": 2.3277697145621676,
+      "grad_norm": 0.35335952043533325,
+      "learning_rate": 0.0005309809264305177,
+      "loss": 3.7569,
+      "step": 7700
+    },
+    {
+      "epoch": 2.34288824383164,
+      "grad_norm": 0.33651435375213623,
+      "learning_rate": 0.0005305267938237964,
+      "loss": 3.7466,
+      "step": 7750
+    },
+    {
+      "epoch": 2.3580067731011125,
+      "grad_norm": 0.31409308314323425,
+      "learning_rate": 0.0005300726612170754,
+      "loss": 3.7599,
+      "step": 7800
+    },
+    {
+      "epoch": 2.3731253023705854,
+      "grad_norm": 0.3148058354854584,
+      "learning_rate": 0.0005296185286103542,
+      "loss": 3.7423,
+      "step": 7850
+    },
+    {
+      "epoch": 2.3882438316400583,
+      "grad_norm": 0.34514766931533813,
+      "learning_rate": 0.000529164396003633,
+      "loss": 3.7253,
+      "step": 7900
+    },
+    {
+      "epoch": 2.4033623609095307,
+      "grad_norm": 0.32518208026885986,
+      "learning_rate": 0.0005287102633969119,
+      "loss": 3.7365,
+      "step": 7950
+    },
+    {
+      "epoch": 2.418480890179003,
+      "grad_norm": 0.3367462158203125,
+      "learning_rate": 0.0005282561307901907,
+      "loss": 3.7526,
+      "step": 8000
+    },
+    {
+      "epoch": 2.418480890179003,
+      "eval_accuracy": 0.3448259887850608,
+      "eval_loss": 3.7824463844299316,
+      "eval_runtime": 182.5201,
+      "eval_samples_per_second": 91.19,
+      "eval_steps_per_second": 5.703,
+      "step": 8000
+    },
+    {
+      "epoch": 2.433599419448476,
+      "grad_norm": 0.3492235541343689,
+      "learning_rate": 0.0005278019981834695,
+      "loss": 3.7511,
+      "step": 8050
+    },
+    {
+      "epoch": 2.448717948717949,
+      "grad_norm": 0.3320801556110382,
+      "learning_rate": 0.0005273478655767484,
+      "loss": 3.7335,
+      "step": 8100
+    },
+    {
+      "epoch": 2.4638364779874213,
+      "grad_norm": 0.32247304916381836,
+      "learning_rate": 0.0005268937329700272,
+      "loss": 3.7359,
+      "step": 8150
+    },
+    {
+      "epoch": 2.478955007256894,
+      "grad_norm": 0.3158959448337555,
+      "learning_rate": 0.000526439600363306,
+      "loss": 3.7384,
+      "step": 8200
+    },
+    {
+      "epoch": 2.4940735365263667,
+      "grad_norm": 0.3267774283885956,
+      "learning_rate": 0.0005259854677565848,
+      "loss": 3.7429,
+      "step": 8250
+    },
+    {
+      "epoch": 2.5091920657958395,
+      "grad_norm": 0.34227943420410156,
+      "learning_rate": 0.0005255313351498637,
+      "loss": 3.7212,
+      "step": 8300
+    },
+    {
+      "epoch": 2.524310595065312,
+      "grad_norm": 0.3359690010547638,
+      "learning_rate": 0.0005250772025431426,
+      "loss": 3.7497,
+      "step": 8350
+    },
+    {
+      "epoch": 2.539429124334785,
+      "grad_norm": 0.3244568407535553,
+      "learning_rate": 0.0005246230699364214,
+      "loss": 3.738,
+      "step": 8400
+    },
+    {
+      "epoch": 2.5545476536042573,
+      "grad_norm": 0.3494398891925812,
+      "learning_rate": 0.0005241689373297002,
+      "loss": 3.7482,
+      "step": 8450
+    },
+    {
+      "epoch": 2.56966618287373,
+      "grad_norm": 0.33279120922088623,
+      "learning_rate": 0.000523714804722979,
+      "loss": 3.7474,
+      "step": 8500
+    },
+    {
+      "epoch": 2.5847847121432026,
+      "grad_norm": 0.3192063868045807,
+      "learning_rate": 0.0005232606721162579,
+      "loss": 3.7344,
+      "step": 8550
+    },
+    {
+      "epoch": 2.5999032414126755,
+      "grad_norm": 0.32627859711647034,
+      "learning_rate": 0.0005228065395095368,
+      "loss": 3.7326,
+      "step": 8600
+    },
+    {
+      "epoch": 2.615021770682148,
+      "grad_norm": 0.32838842272758484,
+      "learning_rate": 0.0005223524069028156,
+      "loss": 3.7382,
+      "step": 8650
+    },
+    {
+      "epoch": 2.630140299951621,
+      "grad_norm": 0.30039238929748535,
+      "learning_rate": 0.0005218982742960944,
+      "loss": 3.7325,
+      "step": 8700
+    },
+    {
+      "epoch": 2.6452588292210932,
+      "grad_norm": 0.35245180130004883,
+      "learning_rate": 0.0005214441416893733,
+      "loss": 3.7375,
+      "step": 8750
+    },
+    {
+      "epoch": 2.660377358490566,
+      "grad_norm": 0.3139996826648712,
+      "learning_rate": 0.0005209900090826521,
+      "loss": 3.7314,
+      "step": 8800
+    },
+    {
+      "epoch": 2.6754958877600385,
+      "grad_norm": 0.32654982805252075,
+      "learning_rate": 0.0005205358764759309,
+      "loss": 3.7347,
+      "step": 8850
+    },
+    {
+      "epoch": 2.6906144170295114,
+      "grad_norm": 0.3103334307670593,
+      "learning_rate": 0.0005200817438692098,
+      "loss": 3.7339,
+      "step": 8900
+    },
+    {
+      "epoch": 2.7057329462989843,
+      "grad_norm": 0.3158484101295471,
+      "learning_rate": 0.0005196276112624886,
+      "loss": 3.7294,
+      "step": 8950
+    },
+    {
+      "epoch": 2.7208514755684567,
+      "grad_norm": 0.3339202404022217,
+      "learning_rate": 0.0005191734786557674,
+      "loss": 3.7241,
+      "step": 9000
+    },
+    {
+      "epoch": 2.7208514755684567,
+      "eval_accuracy": 0.3476167901457528,
+      "eval_loss": 3.7497520446777344,
+      "eval_runtime": 182.7073,
+      "eval_samples_per_second": 91.097,
+      "eval_steps_per_second": 5.698,
+      "step": 9000
+    },
+    {
+      "epoch": 2.735970004837929,
+      "grad_norm": 0.33506032824516296,
+      "learning_rate": 0.0005187193460490462,
+      "loss": 3.7369,
+      "step": 9050
+    },
+    {
+      "epoch": 2.751088534107402,
+      "grad_norm": 0.34870266914367676,
+      "learning_rate": 0.0005182652134423252,
+      "loss": 3.7235,
+      "step": 9100
+    },
+    {
+      "epoch": 2.766207063376875,
+      "grad_norm": 0.33400285243988037,
+      "learning_rate": 0.000517811080835604,
+      "loss": 3.7238,
+      "step": 9150
+    },
+    {
+      "epoch": 2.7813255926463474,
+      "grad_norm": 0.32307007908821106,
+      "learning_rate": 0.0005173569482288827,
+      "loss": 3.7244,
+      "step": 9200
+    },
+    {
+      "epoch": 2.79644412191582,
+      "grad_norm": 0.3319950997829437,
+      "learning_rate": 0.0005169028156221616,
+      "loss": 3.7157,
+      "step": 9250
+    },
+    {
+      "epoch": 2.8115626511852927,
+      "grad_norm": 0.3344840705394745,
+      "learning_rate": 0.0005164486830154405,
+      "loss": 3.7192,
+      "step": 9300
+    },
+    {
+      "epoch": 2.8266811804547656,
+      "grad_norm": 0.30677810311317444,
+      "learning_rate": 0.0005159945504087193,
+      "loss": 3.7245,
+      "step": 9350
+    },
+    {
+      "epoch": 2.841799709724238,
+      "grad_norm": 0.3162977695465088,
+      "learning_rate": 0.0005155404178019982,
+      "loss": 3.7228,
+      "step": 9400
+    },
+    {
+      "epoch": 2.8569182389937104,
+      "grad_norm": 0.3243446946144104,
+      "learning_rate": 0.000515086285195277,
+      "loss": 3.716,
+      "step": 9450
+    },
+    {
+      "epoch": 2.8720367682631833,
+      "grad_norm": 0.3386097252368927,
+      "learning_rate": 0.0005146321525885558,
+      "loss": 3.7195,
+      "step": 9500
+    },
+    {
+      "epoch": 2.887155297532656,
+      "grad_norm": 0.3524998128414154,
+      "learning_rate": 0.0005141780199818347,
+      "loss": 3.7071,
+      "step": 9550
+    },
+    {
+      "epoch": 2.9022738268021286,
+      "grad_norm": 0.3099507689476013,
+      "learning_rate": 0.0005137238873751135,
+      "loss": 3.7196,
+      "step": 9600
+    },
+    {
+      "epoch": 2.917392356071601,
+      "grad_norm": 0.31758880615234375,
+      "learning_rate": 0.0005132697547683923,
+      "loss": 3.7128,
+      "step": 9650
+    },
+    {
+      "epoch": 2.932510885341074,
+      "grad_norm": 0.3277053236961365,
+      "learning_rate": 0.0005128156221616712,
+      "loss": 3.7313,
+      "step": 9700
+    },
+    {
+      "epoch": 2.947629414610547,
+      "grad_norm": 0.30629098415374756,
+      "learning_rate": 0.00051236148955495,
+      "loss": 3.7135,
+      "step": 9750
+    },
+    {
+      "epoch": 2.9627479438800193,
+      "grad_norm": 0.3403231203556061,
+      "learning_rate": 0.0005119073569482288,
+      "loss": 3.7087,
+      "step": 9800
+    },
+    {
+      "epoch": 2.977866473149492,
+      "grad_norm": 0.3153614103794098,
+      "learning_rate": 0.0005114532243415078,
+      "loss": 3.7156,
+      "step": 9850
+    },
+    {
+      "epoch": 2.9929850024189646,
+      "grad_norm": 0.3428856134414673,
+      "learning_rate": 0.0005109990917347865,
+      "loss": 3.7188,
+      "step": 9900
+    },
+    {
+      "epoch": 3.007861635220126,
+      "grad_norm": 0.32290196418762207,
+      "learning_rate": 0.0005105449591280653,
+      "loss": 3.657,
+      "step": 9950
+    },
+    {
+      "epoch": 3.0229801644895984,
+      "grad_norm": 0.3294488787651062,
+      "learning_rate": 0.0005100908265213441,
+      "loss": 3.5917,
+      "step": 10000
+    },
+    {
+      "epoch": 3.0229801644895984,
+      "eval_accuracy": 0.35003781267768785,
+      "eval_loss": 3.730295181274414,
+      "eval_runtime": 182.2568,
+      "eval_samples_per_second": 91.322,
+      "eval_steps_per_second": 5.712,
+      "step": 10000
+    },
+    {
+      "epoch": 3.0380986937590713,
+      "grad_norm": 0.3343505561351776,
+      "learning_rate": 0.0005096366939146231,
+      "loss": 3.5981,
+      "step": 10050
+    },
+    {
+      "epoch": 3.0532172230285437,
+      "grad_norm": 0.32677552103996277,
+      "learning_rate": 0.0005091825613079019,
+      "loss": 3.6168,
+      "step": 10100
+    },
+    {
+      "epoch": 3.0683357522980166,
+      "grad_norm": 0.32819923758506775,
+      "learning_rate": 0.0005087284287011807,
+      "loss": 3.6285,
+      "step": 10150
+    },
+    {
+      "epoch": 3.083454281567489,
+      "grad_norm": 0.34550225734710693,
+      "learning_rate": 0.0005082742960944595,
+      "loss": 3.6147,
+      "step": 10200
+    },
+    {
+      "epoch": 3.098572810836962,
+      "grad_norm": 0.3343297839164734,
+      "learning_rate": 0.0005078201634877384,
+      "loss": 3.6216,
+      "step": 10250
+    },
+    {
+      "epoch": 3.1136913401064343,
+      "grad_norm": 0.3187820315361023,
+      "learning_rate": 0.0005073660308810172,
+      "loss": 3.6035,
+      "step": 10300
+    },
+    {
+      "epoch": 3.128809869375907,
+      "grad_norm": 0.31732386350631714,
+      "learning_rate": 0.0005069118982742961,
+      "loss": 3.6181,
+      "step": 10350
+    },
+    {
+      "epoch": 3.1439283986453797,
+      "grad_norm": 0.33218568563461304,
+      "learning_rate": 0.0005064577656675749,
+      "loss": 3.6207,
+      "step": 10400
+    },
+    {
+      "epoch": 3.1590469279148525,
+      "grad_norm": 0.3314531147480011,
+      "learning_rate": 0.0005060036330608537,
+      "loss": 3.6108,
+      "step": 10450
+    },
+    {
+      "epoch": 3.174165457184325,
+      "grad_norm": 0.3125026226043701,
+      "learning_rate": 0.0005055495004541326,
+      "loss": 3.6393,
+      "step": 10500
+    },
+    {
+      "epoch": 3.189283986453798,
+      "grad_norm": 0.3300221562385559,
+      "learning_rate": 0.0005050953678474114,
+      "loss": 3.6234,
+      "step": 10550
+    },
+    {
+      "epoch": 3.2044025157232703,
+      "grad_norm": 0.3192563056945801,
+      "learning_rate": 0.0005046412352406902,
+      "loss": 3.6309,
+      "step": 10600
+    },
+    {
+      "epoch": 3.219521044992743,
+      "grad_norm": 0.3273380994796753,
+      "learning_rate": 0.000504187102633969,
+      "loss": 3.6277,
+      "step": 10650
+    },
+    {
+      "epoch": 3.2346395742622156,
+      "grad_norm": 0.3166305422782898,
+      "learning_rate": 0.0005037329700272479,
+      "loss": 3.6222,
+      "step": 10700
+    },
+    {
+      "epoch": 3.2497581035316885,
+      "grad_norm": 0.34870728850364685,
+      "learning_rate": 0.0005032788374205267,
+      "loss": 3.6306,
+      "step": 10750
+    },
+    {
+      "epoch": 3.264876632801161,
+      "grad_norm": 0.3375926911830902,
+      "learning_rate": 0.0005028247048138056,
+      "loss": 3.6342,
+      "step": 10800
+    },
+    {
+      "epoch": 3.279995162070634,
+      "grad_norm": 0.3100663721561432,
+      "learning_rate": 0.0005023705722070845,
+      "loss": 3.6279,
+      "step": 10850
+    },
+    {
+      "epoch": 3.2951136913401067,
+      "grad_norm": 0.31706514954566956,
+      "learning_rate": 0.0005019164396003632,
+      "loss": 3.6229,
+      "step": 10900
+    },
+    {
+      "epoch": 3.310232220609579,
+      "grad_norm": 0.32962343096733093,
+      "learning_rate": 0.000501462306993642,
+      "loss": 3.6332,
+      "step": 10950
+    },
+    {
+      "epoch": 3.3253507498790515,
+      "grad_norm": 0.33387279510498047,
+      "learning_rate": 0.000501008174386921,
+      "loss": 3.6298,
+      "step": 11000
+    },
+    {
+      "epoch": 3.3253507498790515,
+      "eval_accuracy": 0.3517179842080337,
+      "eval_loss": 3.7141380310058594,
+      "eval_runtime": 182.4417,
+      "eval_samples_per_second": 91.229,
+      "eval_steps_per_second": 5.706,
+      "step": 11000
+    },
+    {
+      "epoch": 3.3404692791485244,
+      "grad_norm": 0.31837111711502075,
+      "learning_rate": 0.0005005540417801998,
+      "loss": 3.637,
+      "step": 11050
+    },
+    {
+      "epoch": 3.3555878084179973,
+      "grad_norm": 0.33772799372673035,
+      "learning_rate": 0.0005000999091734786,
+      "loss": 3.6324,
+      "step": 11100
+    },
+    {
+      "epoch": 3.3707063376874697,
+      "grad_norm": 0.3386647403240204,
+      "learning_rate": 0.0004996457765667575,
+      "loss": 3.629,
+      "step": 11150
+    },
+    {
+      "epoch": 3.385824866956942,
+      "grad_norm": 0.31823232769966125,
+      "learning_rate": 0.0004991916439600363,
+      "loss": 3.6385,
+      "step": 11200
+    },
+    {
+      "epoch": 3.400943396226415,
+      "grad_norm": 0.32417285442352295,
+      "learning_rate": 0.0004987375113533151,
+      "loss": 3.6336,
+      "step": 11250
+    },
+    {
+      "epoch": 3.416061925495888,
+      "grad_norm": 0.34577304124832153,
+      "learning_rate": 0.000498283378746594,
+      "loss": 3.6371,
+      "step": 11300
+    },
+    {
+      "epoch": 3.4311804547653604,
+      "grad_norm": 0.35057663917541504,
+      "learning_rate": 0.0004978292461398728,
+      "loss": 3.6232,
+      "step": 11350
+    },
+    {
+      "epoch": 3.4462989840348333,
+      "grad_norm": 0.33663854002952576,
+      "learning_rate": 0.0004973751135331516,
+      "loss": 3.6322,
+      "step": 11400
+    },
+    {
+      "epoch": 3.4614175133043057,
+      "grad_norm": 0.32272565364837646,
+      "learning_rate": 0.0004969209809264304,
+      "loss": 3.6351,
+      "step": 11450
+    },
+    {
+      "epoch": 3.4765360425737786,
+      "grad_norm": 0.34051501750946045,
+      "learning_rate": 0.0004964668483197093,
+      "loss": 3.6287,
+      "step": 11500
+    },
+    {
+      "epoch": 3.491654571843251,
+      "grad_norm": 0.3307649493217468,
+      "learning_rate": 0.0004960127157129882,
+      "loss": 3.6389,
+      "step": 11550
+    },
+    {
+      "epoch": 3.506773101112724,
+      "grad_norm": 0.3179891109466553,
+      "learning_rate": 0.0004955585831062669,
+      "loss": 3.6281,
+      "step": 11600
+    },
+    {
+      "epoch": 3.5218916303821963,
+      "grad_norm": 0.32227689027786255,
+      "learning_rate": 0.0004951044504995458,
+      "loss": 3.6423,
+      "step": 11650
+    },
+    {
+      "epoch": 3.537010159651669,
+      "grad_norm": 0.35890889167785645,
+      "learning_rate": 0.0004946503178928246,
+      "loss": 3.6468,
+      "step": 11700
+    },
+    {
+      "epoch": 3.5521286889211416,
+      "grad_norm": 0.36508727073669434,
+      "learning_rate": 0.0004941961852861035,
+      "loss": 3.6366,
+      "step": 11750
+    },
+    {
+      "epoch": 3.5672472181906145,
+      "grad_norm": 0.32733410596847534,
+      "learning_rate": 0.0004937420526793824,
+      "loss": 3.6355,
+      "step": 11800
+    },
+    {
+      "epoch": 3.582365747460087,
+      "grad_norm": 0.3346957564353943,
+      "learning_rate": 0.0004932879200726612,
+      "loss": 3.6388,
+      "step": 11850
+    },
+    {
+      "epoch": 3.59748427672956,
+      "grad_norm": 0.32891422510147095,
+      "learning_rate": 0.00049283378746594,
+      "loss": 3.629,
+      "step": 11900
+    },
+    {
+      "epoch": 3.6126028059990323,
+      "grad_norm": 0.3372475206851959,
+      "learning_rate": 0.0004923796548592189,
+      "loss": 3.6244,
+      "step": 11950
+    },
+    {
+      "epoch": 3.627721335268505,
+      "grad_norm": 0.3311806917190552,
+      "learning_rate": 0.0004919255222524977,
+      "loss": 3.6253,
+      "step": 12000
+    },
+    {
+      "epoch": 3.627721335268505,
+      "eval_accuracy": 0.35409750215283,
+      "eval_loss": 3.6940906047821045,
+      "eval_runtime": 182.4197,
+      "eval_samples_per_second": 91.24,
+      "eval_steps_per_second": 5.707,
+      "step": 12000
+    },
+    {
+      "epoch": 3.6428398645379776,
+      "grad_norm": 0.31895172595977783,
+      "learning_rate": 0.0004914713896457765,
+      "loss": 3.6391,
+      "step": 12050
+    },
+    {
+      "epoch": 3.6579583938074505,
+      "grad_norm": 0.3537571430206299,
+      "learning_rate": 0.0004910172570390554,
+      "loss": 3.6328,
+      "step": 12100
+    },
+    {
+      "epoch": 3.6730769230769234,
+      "grad_norm": 0.336915522813797,
+      "learning_rate": 0.0004905631244323342,
+      "loss": 3.6389,
+      "step": 12150
+    },
+    {
+      "epoch": 3.688195452346396,
+      "grad_norm": 0.33801692724227905,
+      "learning_rate": 0.000490108991825613,
+      "loss": 3.6309,
+      "step": 12200
+    },
+    {
+      "epoch": 3.7033139816158682,
+      "grad_norm": 0.3218366801738739,
+      "learning_rate": 0.0004896548592188918,
+      "loss": 3.625,
+      "step": 12250
+    },
+    {
+      "epoch": 3.718432510885341,
+      "grad_norm": 0.33120447397232056,
+      "learning_rate": 0.0004892007266121708,
+      "loss": 3.6343,
+      "step": 12300
+    },
+    {
+      "epoch": 3.733551040154814,
+      "grad_norm": 0.32918858528137207,
+      "learning_rate": 0.0004887465940054495,
+      "loss": 3.6272,
+      "step": 12350
+    },
+    {
+      "epoch": 3.7486695694242864,
+      "grad_norm": 0.3185270130634308,
+      "learning_rate": 0.0004882924613987283,
+      "loss": 3.6325,
+      "step": 12400
+    },
+    {
+      "epoch": 3.763788098693759,
+      "grad_norm": 0.31657835841178894,
+      "learning_rate": 0.00048783832879200717,
+      "loss": 3.6286,
+      "step": 12450
+    },
+    {
+      "epoch": 3.7789066279632317,
+      "grad_norm": 0.31327733397483826,
+      "learning_rate": 0.00048738419618528606,
+      "loss": 3.6213,
+      "step": 12500
+    },
+    {
+      "epoch": 3.7940251572327046,
+      "grad_norm": 0.32215988636016846,
+      "learning_rate": 0.0004869300635785649,
+      "loss": 3.6203,
+      "step": 12550
+    },
+    {
+      "epoch": 3.809143686502177,
+      "grad_norm": 0.3287181556224823,
+      "learning_rate": 0.0004864759309718437,
+      "loss": 3.6269,
+      "step": 12600
+    },
+    {
+      "epoch": 3.8242622157716495,
+      "grad_norm": 0.3348924219608307,
+      "learning_rate": 0.0004860217983651226,
+      "loss": 3.6146,
+      "step": 12650
+    },
+    {
+      "epoch": 3.8393807450411224,
+      "grad_norm": 0.3488834500312805,
+      "learning_rate": 0.00048556766575840143,
+      "loss": 3.6344,
+      "step": 12700
+    },
+    {
+      "epoch": 3.8544992743105952,
+      "grad_norm": 0.32925039529800415,
+      "learning_rate": 0.00048511353315168026,
+      "loss": 3.6386,
+      "step": 12750
+    },
+    {
+      "epoch": 3.8696178035800677,
+      "grad_norm": 0.33648958802223206,
+      "learning_rate": 0.00048465940054495904,
+      "loss": 3.6341,
+      "step": 12800
+    },
+    {
+      "epoch": 3.8847363328495406,
+      "grad_norm": 0.3223407566547394,
+      "learning_rate": 0.0004842052679382379,
+      "loss": 3.6252,
+      "step": 12850
+    },
+    {
+      "epoch": 3.899854862119013,
+      "grad_norm": 0.3572065830230713,
+      "learning_rate": 0.00048375113533151676,
+      "loss": 3.6246,
+      "step": 12900
+    },
+    {
+      "epoch": 3.914973391388486,
+      "grad_norm": 0.34701117873191833,
+      "learning_rate": 0.0004832970027247956,
+      "loss": 3.6167,
+      "step": 12950
+    },
+    {
+      "epoch": 3.9300919206579583,
+      "grad_norm": 0.33134695887565613,
+      "learning_rate": 0.0004828428701180744,
+      "loss": 3.6323,
+      "step": 13000
+    },
+    {
+      "epoch": 3.9300919206579583,
+      "eval_accuracy": 0.3554904337217598,
+      "eval_loss": 3.676191568374634,
+      "eval_runtime": 182.5829,
+      "eval_samples_per_second": 91.159,
+      "eval_steps_per_second": 5.702,
+      "step": 13000
+    },
+    {
+      "epoch": 3.945210449927431,
+      "grad_norm": 0.3298724293708801,
+      "learning_rate": 0.0004823887375113533,
+      "loss": 3.6291,
+      "step": 13050
+    },
+    {
+      "epoch": 3.9603289791969036,
+      "grad_norm": 0.3415639102458954,
+      "learning_rate": 0.00048193460490463213,
+      "loss": 3.6256,
+      "step": 13100
+    },
+    {
+      "epoch": 3.9754475084663765,
+      "grad_norm": 0.3238424062728882,
+      "learning_rate": 0.0004814804722979109,
+      "loss": 3.6304,
+      "step": 13150
+    },
+    {
+      "epoch": 3.990566037735849,
+      "grad_norm": 0.31136244535446167,
+      "learning_rate": 0.00048102633969118974,
+      "loss": 3.6211,
+      "step": 13200
+    },
+    {
+      "epoch": 4.00544267053701,
+      "grad_norm": 0.3414509892463684,
+      "learning_rate": 0.0004805722070844686,
+      "loss": 3.5841,
+      "step": 13250
+    },
+    {
+      "epoch": 4.020561199806483,
+      "grad_norm": 0.3479083776473999,
+      "learning_rate": 0.00048011807447774746,
+      "loss": 3.5103,
+      "step": 13300
+    },
+    {
+      "epoch": 4.035679729075955,
+      "grad_norm": 0.32956740260124207,
+      "learning_rate": 0.0004796639418710263,
+      "loss": 3.5115,
+      "step": 13350
+    },
+    {
+      "epoch": 4.050798258345428,
+      "grad_norm": 0.32761216163635254,
+      "learning_rate": 0.00047920980926430517,
+      "loss": 3.5252,
+      "step": 13400
+    },
+    {
+      "epoch": 4.065916787614901,
+      "grad_norm": 0.32798823714256287,
+      "learning_rate": 0.000478755676657584,
+      "loss": 3.529,
+      "step": 13450
+    },
+    {
+      "epoch": 4.081035316884374,
+      "grad_norm": 0.3446704149246216,
+      "learning_rate": 0.00047830154405086283,
+      "loss": 3.5224,
+      "step": 13500
+    },
+    {
+      "epoch": 4.096153846153846,
+      "grad_norm": 0.3277835547924042,
+      "learning_rate": 0.0004778474114441416,
+      "loss": 3.5262,
+      "step": 13550
+    },
+    {
+      "epoch": 4.111272375423319,
+      "grad_norm": 0.3266860544681549,
+      "learning_rate": 0.0004773932788374205,
+      "loss": 3.5234,
+      "step": 13600
+    },
+    {
+      "epoch": 4.126390904692792,
+      "grad_norm": 0.33199459314346313,
+      "learning_rate": 0.0004769391462306993,
+      "loss": 3.5329,
+      "step": 13650
+    },
+    {
+      "epoch": 4.1415094339622645,
+      "grad_norm": 0.3167458176612854,
+      "learning_rate": 0.00047648501362397816,
+      "loss": 3.5315,
+      "step": 13700
+    },
+    {
+      "epoch": 4.1566279632317364,
+      "grad_norm": 0.33163997530937195,
+      "learning_rate": 0.000476030881017257,
+      "loss": 3.539,
+      "step": 13750
+    },
+    {
+      "epoch": 4.171746492501209,
+      "grad_norm": 0.3242467939853668,
+      "learning_rate": 0.00047557674841053587,
+      "loss": 3.5322,
+      "step": 13800
+    },
+    {
+      "epoch": 4.186865021770682,
+      "grad_norm": 0.3438900113105774,
+      "learning_rate": 0.0004751226158038147,
+      "loss": 3.5399,
+      "step": 13850
+    },
+    {
+      "epoch": 4.201983551040155,
+      "grad_norm": 0.34759825468063354,
+      "learning_rate": 0.0004746684831970935,
+      "loss": 3.5452,
+      "step": 13900
+    },
+    {
+      "epoch": 4.217102080309627,
+      "grad_norm": 0.3428172171115875,
+      "learning_rate": 0.0004742143505903723,
+      "loss": 3.5468,
+      "step": 13950
+    },
+    {
+      "epoch": 4.2322206095791,
+      "grad_norm": 0.3243090808391571,
+      "learning_rate": 0.0004737602179836512,
+      "loss": 3.5399,
+      "step": 14000
+    },
+    {
+      "epoch": 4.2322206095791,
+      "eval_accuracy": 0.35642869605990957,
+      "eval_loss": 3.6728808879852295,
+      "eval_runtime": 182.5974,
+      "eval_samples_per_second": 91.151,
+      "eval_steps_per_second": 5.701,
+      "step": 14000
+    },
+    {
+      "epoch": 4.247339138848573,
+      "grad_norm": 0.3331063985824585,
+      "learning_rate": 0.00047330608537693,
+      "loss": 3.5417,
+      "step": 14050
+    },
+    {
+      "epoch": 4.262457668118046,
+      "grad_norm": 0.32661089301109314,
+      "learning_rate": 0.00047285195277020886,
+      "loss": 3.5406,
+      "step": 14100
+    },
+    {
+      "epoch": 4.277576197387518,
+      "grad_norm": 0.33398354053497314,
+      "learning_rate": 0.00047239782016348774,
+      "loss": 3.5467,
+      "step": 14150
+    },
+    {
+      "epoch": 4.292694726656991,
+      "grad_norm": 0.3145262897014618,
+      "learning_rate": 0.00047194368755676657,
+      "loss": 3.5434,
+      "step": 14200
+    },
+    {
+      "epoch": 4.3078132559264635,
+      "grad_norm": 0.3353433609008789,
+      "learning_rate": 0.00047148955495004535,
+      "loss": 3.5485,
+      "step": 14250
+    },
+    {
+      "epoch": 4.322931785195936,
+      "grad_norm": 0.33363988995552063,
+      "learning_rate": 0.0004710354223433242,
+      "loss": 3.5426,
+      "step": 14300
+    },
+    {
+      "epoch": 4.338050314465409,
+      "grad_norm": 0.33351442217826843,
+      "learning_rate": 0.00047058128973660306,
+      "loss": 3.5549,
+      "step": 14350
+    },
+    {
+      "epoch": 4.353168843734881,
+      "grad_norm": 0.3099449872970581,
+      "learning_rate": 0.0004701271571298819,
+      "loss": 3.5518,
+      "step": 14400
+    },
+    {
+      "epoch": 4.368287373004354,
+      "grad_norm": 0.3316916823387146,
+      "learning_rate": 0.0004696730245231607,
+      "loss": 3.557,
+      "step": 14450
+    },
+    {
+      "epoch": 4.383405902273827,
+      "grad_norm": 0.32294154167175293,
+      "learning_rate": 0.00046921889191643956,
+      "loss": 3.5462,
+      "step": 14500
+    },
+    {
+      "epoch": 4.3985244315433,
+      "grad_norm": 0.32864052057266235,
+      "learning_rate": 0.00046876475930971844,
+      "loss": 3.5644,
+      "step": 14550
+    },
+    {
+      "epoch": 4.413642960812772,
+      "grad_norm": 0.3232649564743042,
+      "learning_rate": 0.0004683106267029972,
+      "loss": 3.5485,
+      "step": 14600
+    },
+    {
+      "epoch": 4.428761490082245,
+      "grad_norm": 0.33318766951560974,
+      "learning_rate": 0.00046785649409627605,
+      "loss": 3.5564,
+      "step": 14650
+    },
+    {
+      "epoch": 4.443880019351718,
+      "grad_norm": 0.3175857663154602,
+      "learning_rate": 0.0004674023614895549,
+      "loss": 3.5574,
+      "step": 14700
+    },
+    {
+      "epoch": 4.4589985486211905,
+      "grad_norm": 0.3410588502883911,
+      "learning_rate": 0.00046694822888283376,
+      "loss": 3.5549,
+      "step": 14750
+    },
+    {
+      "epoch": 4.4741170778906625,
+      "grad_norm": 0.33438578248023987,
+      "learning_rate": 0.0004664940962761126,
+      "loss": 3.5644,
+      "step": 14800
+    },
+    {
+      "epoch": 4.489235607160135,
+      "grad_norm": 0.3205425441265106,
+      "learning_rate": 0.0004660399636693914,
+      "loss": 3.5554,
+      "step": 14850
+    },
+    {
+      "epoch": 4.504354136429608,
+      "grad_norm": 0.3268115222454071,
+      "learning_rate": 0.0004655858310626703,
+      "loss": 3.568,
+      "step": 14900
+    },
+    {
+      "epoch": 4.519472665699081,
+      "grad_norm": 0.32924684882164,
+      "learning_rate": 0.0004651316984559491,
+      "loss": 3.5596,
+      "step": 14950
+    },
+    {
+      "epoch": 4.534591194968553,
+      "grad_norm": 0.3317072093486786,
+      "learning_rate": 0.0004646775658492279,
+      "loss": 3.5538,
+      "step": 15000
+    },
+    {
+      "epoch": 4.534591194968553,
+      "eval_accuracy": 0.35780751842074693,
+      "eval_loss": 3.6592884063720703,
+      "eval_runtime": 182.5425,
+      "eval_samples_per_second": 91.179,
+      "eval_steps_per_second": 5.703,
+      "step": 15000
+    },
+    {
+      "epoch": 4.549709724238026,
+      "grad_norm": 0.3319750428199768,
+      "learning_rate": 0.00046422343324250675,
+      "loss": 3.5586,
+      "step": 15050
+    },
+    {
+      "epoch": 4.564828253507499,
+      "grad_norm": 0.33184248208999634,
+      "learning_rate": 0.00046376930063578563,
+      "loss": 3.5611,
+      "step": 15100
+    },
+    {
+      "epoch": 4.579946782776972,
+      "grad_norm": 0.3535155653953552,
+      "learning_rate": 0.00046331516802906446,
+      "loss": 3.56,
+      "step": 15150
+    },
+    {
+      "epoch": 4.595065312046444,
+      "grad_norm": 0.322519987821579,
+      "learning_rate": 0.0004628610354223433,
+      "loss": 3.5594,
+      "step": 15200
+    },
+    {
+      "epoch": 4.610183841315917,
+      "grad_norm": 0.31665462255477905,
+      "learning_rate": 0.00046240690281562207,
+      "loss": 3.5673,
+      "step": 15250
+    },
+    {
+      "epoch": 4.6253023705853895,
+      "grad_norm": 0.341170072555542,
+      "learning_rate": 0.00046195277020890096,
+      "loss": 3.5536,
+      "step": 15300
+    },
+    {
+      "epoch": 4.640420899854862,
+      "grad_norm": 0.3396996557712555,
+      "learning_rate": 0.0004614986376021798,
+      "loss": 3.5653,
+      "step": 15350
+    },
+    {
+      "epoch": 4.655539429124335,
+      "grad_norm": 0.33313286304473877,
+      "learning_rate": 0.0004610445049954586,
+      "loss": 3.5667,
+      "step": 15400
+    },
+    {
+      "epoch": 4.670657958393807,
+      "grad_norm": 0.3346698582172394,
+      "learning_rate": 0.00046059037238873745,
+      "loss": 3.5552,
+      "step": 15450
+    },
+    {
+      "epoch": 4.68577648766328,
+      "grad_norm": 0.32503342628479004,
+      "learning_rate": 0.00046013623978201633,
+      "loss": 3.5628,
+      "step": 15500
+    },
+    {
+      "epoch": 4.700895016932753,
+      "grad_norm": 0.3317323327064514,
+      "learning_rate": 0.00045968210717529516,
+      "loss": 3.5598,
+      "step": 15550
+    },
+    {
+      "epoch": 4.716013546202225,
+      "grad_norm": 0.31406593322753906,
+      "learning_rate": 0.00045922797456857394,
+      "loss": 3.5652,
+      "step": 15600
+    },
+    {
+      "epoch": 4.731132075471698,
+      "grad_norm": 0.35221582651138306,
+      "learning_rate": 0.0004587738419618528,
+      "loss": 3.5625,
+      "step": 15650
+    },
+    {
+      "epoch": 4.746250604741171,
+      "grad_norm": 0.3168243169784546,
+      "learning_rate": 0.00045831970935513166,
+      "loss": 3.5446,
+      "step": 15700
+    },
+    {
+      "epoch": 4.761369134010644,
+      "grad_norm": 0.3266247510910034,
+      "learning_rate": 0.0004578655767484105,
+      "loss": 3.5824,
+      "step": 15750
+    },
+    {
+      "epoch": 4.7764876632801165,
+      "grad_norm": 0.3345303535461426,
+      "learning_rate": 0.0004574114441416893,
+      "loss": 3.5577,
+      "step": 15800
+    },
+    {
+      "epoch": 4.7916061925495885,
+      "grad_norm": 0.36775925755500793,
+      "learning_rate": 0.0004569573115349682,
+      "loss": 3.5736,
+      "step": 15850
+    },
+    {
+      "epoch": 4.806724721819061,
+      "grad_norm": 0.31279629468917847,
+      "learning_rate": 0.00045650317892824703,
+      "loss": 3.5719,
+      "step": 15900
+    },
+    {
+      "epoch": 4.821843251088534,
+      "grad_norm": 0.32946261763572693,
+      "learning_rate": 0.0004560490463215258,
+      "loss": 3.5472,
+      "step": 15950
+    },
+    {
+      "epoch": 4.836961780358006,
+      "grad_norm": 0.3353503942489624,
+      "learning_rate": 0.00045559491371480464,
+      "loss": 3.5512,
+      "step": 16000
+    },
+    {
+      "epoch": 4.836961780358006,
+      "eval_accuracy": 0.3589908106727694,
+      "eval_loss": 3.6440577507019043,
+      "eval_runtime": 182.5579,
+      "eval_samples_per_second": 91.171,
+      "eval_steps_per_second": 5.702,
+      "step": 16000
+    },
+    {
+      "epoch": 4.852080309627479,
+      "grad_norm": 0.3155944347381592,
+      "learning_rate": 0.0004551407811080835,
+      "loss": 3.5697,
+      "step": 16050
+    },
+    {
+      "epoch": 4.867198838896952,
+      "grad_norm": 0.34022998809814453,
+      "learning_rate": 0.00045468664850136235,
+      "loss": 3.5635,
+      "step": 16100
+    },
+    {
+      "epoch": 4.882317368166425,
+      "grad_norm": 0.349436491727829,
+      "learning_rate": 0.0004542325158946412,
+      "loss": 3.5606,
+      "step": 16150
+    },
+    {
+      "epoch": 4.897435897435898,
+      "grad_norm": 0.32045167684555054,
+      "learning_rate": 0.00045377838328792,
+      "loss": 3.5483,
+      "step": 16200
+    },
+    {
+      "epoch": 4.91255442670537,
+      "grad_norm": 0.3273170292377472,
+      "learning_rate": 0.0004533242506811989,
+      "loss": 3.561,
+      "step": 16250
+    },
+    {
+      "epoch": 4.927672955974843,
+      "grad_norm": 0.3492163121700287,
+      "learning_rate": 0.00045287011807447773,
+      "loss": 3.5538,
+      "step": 16300
+    },
+    {
+      "epoch": 4.9427914852443156,
+      "grad_norm": 0.3319571018218994,
+      "learning_rate": 0.0004524159854677565,
+      "loss": 3.5678,
+      "step": 16350
+    },
+    {
+      "epoch": 4.957910014513788,
+      "grad_norm": 0.3193274438381195,
+      "learning_rate": 0.0004519618528610354,
+      "loss": 3.5604,
+      "step": 16400
+    },
+    {
+      "epoch": 4.97302854378326,
+      "grad_norm": 0.3205724060535431,
+      "learning_rate": 0.0004515077202543142,
+      "loss": 3.5501,
+      "step": 16450
+    },
+    {
+      "epoch": 4.988147073052733,
+      "grad_norm": 0.32597634196281433,
+      "learning_rate": 0.00045105358764759305,
+      "loss": 3.5648,
+      "step": 16500
+    },
+    {
+      "epoch": 5.003023705853894,
+      "grad_norm": 0.3266983926296234,
+      "learning_rate": 0.0004505994550408719,
+      "loss": 3.5279,
+      "step": 16550
+    },
+    {
+      "epoch": 5.018142235123367,
+      "grad_norm": 0.32244426012039185,
+      "learning_rate": 0.00045014532243415077,
+      "loss": 3.4542,
+      "step": 16600
+    },
+    {
+      "epoch": 5.03326076439284,
+      "grad_norm": 0.32874855399131775,
+      "learning_rate": 0.0004496911898274296,
+      "loss": 3.4671,
+      "step": 16650
+    },
+    {
+      "epoch": 5.048379293662313,
+      "grad_norm": 0.34664833545684814,
+      "learning_rate": 0.0004492370572207084,
+      "loss": 3.4582,
+      "step": 16700
+    },
+    {
+      "epoch": 5.063497822931785,
+      "grad_norm": 0.3226701021194458,
+      "learning_rate": 0.0004487829246139872,
+      "loss": 3.4471,
+      "step": 16750
+    },
+    {
+      "epoch": 5.078616352201258,
+      "grad_norm": 0.3484433889389038,
+      "learning_rate": 0.0004483287920072661,
+      "loss": 3.4685,
+      "step": 16800
+    },
+    {
+      "epoch": 5.093734881470731,
+      "grad_norm": 0.338821142911911,
+      "learning_rate": 0.0004478746594005449,
+      "loss": 3.4738,
+      "step": 16850
+    },
+    {
+      "epoch": 5.1088534107402035,
+      "grad_norm": 0.3581444323062897,
+      "learning_rate": 0.00044742052679382375,
+      "loss": 3.4712,
+      "step": 16900
+    },
+    {
+      "epoch": 5.1239719400096755,
+      "grad_norm": 0.34923458099365234,
+      "learning_rate": 0.00044696639418710264,
+      "loss": 3.4607,
+      "step": 16950
+    },
+    {
+      "epoch": 5.139090469279148,
+      "grad_norm": 0.33646735548973083,
+      "learning_rate": 0.00044651226158038147,
+      "loss": 3.4722,
+      "step": 17000
+    },
+    {
+      "epoch": 5.139090469279148,
+      "eval_accuracy": 0.35967545999545686,
+      "eval_loss": 3.646660327911377,
+      "eval_runtime": 182.5899,
+      "eval_samples_per_second": 91.155,
+      "eval_steps_per_second": 5.701,
+      "step": 17000
+    },
+    {
+      "epoch": 5.154208998548621,
+      "grad_norm": 0.33612239360809326,
+      "learning_rate": 0.00044605812897366025,
+      "loss": 3.4686,
+      "step": 17050
+    },
+    {
+      "epoch": 5.169327527818094,
+      "grad_norm": 0.3267473876476288,
+      "learning_rate": 0.0004456039963669391,
+      "loss": 3.474,
+      "step": 17100
+    },
+    {
+      "epoch": 5.184446057087566,
+      "grad_norm": 0.33728793263435364,
+      "learning_rate": 0.00044514986376021796,
+      "loss": 3.4742,
+      "step": 17150
+    },
+    {
+      "epoch": 5.199564586357039,
+      "grad_norm": 0.33321622014045715,
+      "learning_rate": 0.0004446957311534968,
+      "loss": 3.4825,
+      "step": 17200
+    },
+    {
+      "epoch": 5.214683115626512,
+      "grad_norm": 0.3262038230895996,
+      "learning_rate": 0.0004442415985467756,
+      "loss": 3.4769,
+      "step": 17250
+    },
+    {
+      "epoch": 5.229801644895985,
+      "grad_norm": 0.3380536139011383,
+      "learning_rate": 0.00044378746594005445,
+      "loss": 3.4779,
+      "step": 17300
+    },
+    {
+      "epoch": 5.244920174165458,
+      "grad_norm": 0.32373255491256714,
+      "learning_rate": 0.00044333333333333334,
+      "loss": 3.4721,
+      "step": 17350
+    },
+    {
+      "epoch": 5.26003870343493,
+      "grad_norm": 0.34807828068733215,
+      "learning_rate": 0.0004428792007266121,
+      "loss": 3.4817,
+      "step": 17400
+    },
+    {
+      "epoch": 5.2751572327044025,
+      "grad_norm": 0.34097328782081604,
+      "learning_rate": 0.00044242506811989095,
+      "loss": 3.4845,
+      "step": 17450
+    },
+    {
+      "epoch": 5.290275761973875,
+      "grad_norm": 0.33616551756858826,
+      "learning_rate": 0.0004419709355131698,
+      "loss": 3.5016,
+      "step": 17500
+    },
+    {
+      "epoch": 5.305394291243347,
+      "grad_norm": 0.3330867886543274,
+      "learning_rate": 0.00044151680290644866,
+      "loss": 3.4844,
+      "step": 17550
+    },
+    {
+      "epoch": 5.32051282051282,
+      "grad_norm": 0.3476291298866272,
+      "learning_rate": 0.0004410626702997275,
+      "loss": 3.4877,
+      "step": 17600
+    },
+    {
+      "epoch": 5.335631349782293,
+      "grad_norm": 0.3363373577594757,
+      "learning_rate": 0.0004406085376930063,
+      "loss": 3.4931,
+      "step": 17650
+    },
+    {
+      "epoch": 5.350749879051766,
+      "grad_norm": 0.35095226764678955,
+      "learning_rate": 0.0004401544050862852,
+      "loss": 3.4904,
+      "step": 17700
+    },
+    {
+      "epoch": 5.365868408321239,
+      "grad_norm": 0.35196179151535034,
+      "learning_rate": 0.000439700272479564,
+      "loss": 3.5003,
+      "step": 17750
+    },
+    {
+      "epoch": 5.380986937590711,
+      "grad_norm": 0.3490413427352905,
+      "learning_rate": 0.0004392461398728428,
+      "loss": 3.4914,
+      "step": 17800
+    },
+    {
+      "epoch": 5.396105466860184,
+      "grad_norm": 0.3287891447544098,
+      "learning_rate": 0.00043879200726612165,
+      "loss": 3.482,
+      "step": 17850
+    },
+    {
+      "epoch": 5.411223996129657,
+      "grad_norm": 0.3373957574367523,
+      "learning_rate": 0.00043833787465940053,
+      "loss": 3.4928,
+      "step": 17900
+    },
+    {
+      "epoch": 5.4263425253991295,
+      "grad_norm": 0.3305628299713135,
+      "learning_rate": 0.00043788374205267936,
+      "loss": 3.4918,
+      "step": 17950
+    },
+    {
+      "epoch": 5.4414610546686015,
+      "grad_norm": 0.33179208636283875,
+      "learning_rate": 0.0004374296094459582,
+      "loss": 3.5072,
+      "step": 18000
+    },
+    {
+      "epoch": 5.4414610546686015,
+      "eval_accuracy": 0.36049544013909796,
+      "eval_loss": 3.6349031925201416,
+      "eval_runtime": 182.6046,
+      "eval_samples_per_second": 91.148,
+      "eval_steps_per_second": 5.701,
+      "step": 18000
+    },
+    {
+      "epoch": 5.456579583938074,
+      "grad_norm": 0.3481210172176361,
+      "learning_rate": 0.00043697547683923697,
+      "loss": 3.5004,
+      "step": 18050
+    },
+    {
+      "epoch": 5.471698113207547,
+      "grad_norm": 0.34053078293800354,
+      "learning_rate": 0.00043652134423251585,
+      "loss": 3.4979,
+      "step": 18100
+    },
+    {
+      "epoch": 5.48681664247702,
+      "grad_norm": 0.3321893811225891,
+      "learning_rate": 0.0004360672116257947,
+      "loss": 3.5042,
+      "step": 18150
+    },
+    {
+      "epoch": 5.501935171746492,
+      "grad_norm": 0.3595956861972809,
+      "learning_rate": 0.0004356130790190735,
+      "loss": 3.4927,
+      "step": 18200
+    },
+    {
+      "epoch": 5.517053701015965,
+      "grad_norm": 0.33346107602119446,
+      "learning_rate": 0.00043515894641235235,
+      "loss": 3.4992,
+      "step": 18250
+    },
+    {
+      "epoch": 5.532172230285438,
+      "grad_norm": 0.3272981345653534,
+      "learning_rate": 0.00043470481380563123,
+      "loss": 3.4941,
+      "step": 18300
+    },
+    {
+      "epoch": 5.547290759554911,
+      "grad_norm": 0.3254975378513336,
+      "learning_rate": 0.00043425068119891006,
+      "loss": 3.5031,
+      "step": 18350
+    },
+    {
+      "epoch": 5.562409288824383,
+      "grad_norm": 0.3381115198135376,
+      "learning_rate": 0.00043379654859218884,
+      "loss": 3.5026,
+      "step": 18400
+    },
+    {
+      "epoch": 5.577527818093856,
+      "grad_norm": 0.35261309146881104,
+      "learning_rate": 0.0004333424159854677,
+      "loss": 3.5006,
+      "step": 18450
+    },
+    {
+      "epoch": 5.5926463473633286,
+      "grad_norm": 0.3541949987411499,
+      "learning_rate": 0.00043288828337874655,
+      "loss": 3.5055,
+      "step": 18500
+    },
+    {
+      "epoch": 5.607764876632801,
+      "grad_norm": 0.34113311767578125,
+      "learning_rate": 0.0004324341507720254,
+      "loss": 3.5077,
+      "step": 18550
+    },
+    {
+      "epoch": 5.622883405902273,
+      "grad_norm": 0.32559531927108765,
+      "learning_rate": 0.0004319800181653042,
+      "loss": 3.509,
+      "step": 18600
+    },
+    {
+      "epoch": 5.638001935171746,
+      "grad_norm": 0.3446948528289795,
+      "learning_rate": 0.0004315258855585831,
+      "loss": 3.504,
+      "step": 18650
+    },
+    {
+      "epoch": 5.653120464441219,
+      "grad_norm": 0.33571872115135193,
+      "learning_rate": 0.00043107175295186193,
+      "loss": 3.4972,
+      "step": 18700
+    },
+    {
+      "epoch": 5.668238993710692,
+      "grad_norm": 0.331429660320282,
+      "learning_rate": 0.0004306176203451407,
+      "loss": 3.4938,
+      "step": 18750
+    },
+    {
+      "epoch": 5.683357522980165,
+      "grad_norm": 0.3403501510620117,
+      "learning_rate": 0.00043016348773841954,
+      "loss": 3.5087,
+      "step": 18800
+    },
+    {
+      "epoch": 5.698476052249637,
+      "grad_norm": 0.3406428098678589,
+      "learning_rate": 0.0004297093551316984,
+      "loss": 3.5111,
+      "step": 18850
+    },
+    {
+      "epoch": 5.71359458151911,
+      "grad_norm": 0.33290889859199524,
+      "learning_rate": 0.00042925522252497725,
+      "loss": 3.4946,
+      "step": 18900
+    },
+    {
+      "epoch": 5.728713110788583,
+      "grad_norm": 0.3506568670272827,
+      "learning_rate": 0.0004288010899182561,
+      "loss": 3.4977,
+      "step": 18950
+    },
+    {
+      "epoch": 5.743831640058055,
+      "grad_norm": 0.32340604066848755,
+      "learning_rate": 0.0004283469573115349,
+      "loss": 3.501,
+      "step": 19000
+    },
+    {
+      "epoch": 5.743831640058055,
+      "eval_accuracy": 0.3617825526473342,
+      "eval_loss": 3.6229019165039062,
+      "eval_runtime": 182.4602,
+      "eval_samples_per_second": 91.22,
+      "eval_steps_per_second": 5.705,
+      "step": 19000
+    },
+    {
+      "epoch": 5.758950169327528,
+      "grad_norm": 0.3353550434112549,
+      "learning_rate": 0.0004278928247048138,
+      "loss": 3.4941,
+      "step": 19050
+    },
+    {
+      "epoch": 5.7740686985970004,
+      "grad_norm": 0.33119282126426697,
+      "learning_rate": 0.00042743869209809263,
+      "loss": 3.5137,
+      "step": 19100
+    },
+    {
+      "epoch": 5.789187227866473,
+      "grad_norm": 0.33732518553733826,
+      "learning_rate": 0.0004269845594913714,
+      "loss": 3.5052,
+      "step": 19150
+    },
+    {
+      "epoch": 5.804305757135946,
+      "grad_norm": 0.32491031289100647,
+      "learning_rate": 0.0004265304268846503,
+      "loss": 3.505,
+      "step": 19200
+    },
+    {
+      "epoch": 5.819424286405418,
+      "grad_norm": 0.328688383102417,
+      "learning_rate": 0.0004260762942779291,
+      "loss": 3.5022,
+      "step": 19250
+    },
+    {
+      "epoch": 5.834542815674891,
+      "grad_norm": 0.33090704679489136,
+      "learning_rate": 0.00042562216167120795,
+      "loss": 3.5023,
+      "step": 19300
+    },
+    {
+      "epoch": 5.849661344944364,
+      "grad_norm": 0.3403943181037903,
+      "learning_rate": 0.0004251680290644868,
+      "loss": 3.4987,
+      "step": 19350
+    },
+    {
+      "epoch": 5.864779874213837,
+      "grad_norm": 0.33841657638549805,
+      "learning_rate": 0.00042471389645776567,
+      "loss": 3.5105,
+      "step": 19400
+    },
+    {
+      "epoch": 5.879898403483309,
+      "grad_norm": 0.31852486729621887,
+      "learning_rate": 0.0004242597638510445,
+      "loss": 3.5051,
+      "step": 19450
+    },
+    {
+      "epoch": 5.895016932752782,
+      "grad_norm": 0.3277070224285126,
+      "learning_rate": 0.0004238056312443233,
+      "loss": 3.4959,
+      "step": 19500
+    },
+    {
+      "epoch": 5.910135462022255,
+      "grad_norm": 0.32635408639907837,
+      "learning_rate": 0.0004233514986376021,
+      "loss": 3.5042,
+      "step": 19550
+    },
+    {
+      "epoch": 5.9252539912917275,
+      "grad_norm": 0.32446935772895813,
+      "learning_rate": 0.000422897366030881,
+      "loss": 3.5092,
+      "step": 19600
+    },
+    {
+      "epoch": 5.9403725205611995,
+      "grad_norm": 0.33674439787864685,
+      "learning_rate": 0.0004224432334241598,
+      "loss": 3.4991,
+      "step": 19650
+    },
+    {
+      "epoch": 5.955491049830672,
+      "grad_norm": 0.32446402311325073,
+      "learning_rate": 0.00042198910081743865,
+      "loss": 3.5077,
+      "step": 19700
+    },
+    {
+      "epoch": 5.970609579100145,
+      "grad_norm": 0.3281049132347107,
+      "learning_rate": 0.0004215349682107175,
+      "loss": 3.5125,
+      "step": 19750
+    },
+    {
+      "epoch": 5.985728108369618,
+      "grad_norm": 0.33700883388519287,
+      "learning_rate": 0.00042108083560399637,
+      "loss": 3.5084,
+      "step": 19800
+    },
+    {
+      "epoch": 6.000604741170779,
+      "grad_norm": 0.35103023052215576,
+      "learning_rate": 0.00042062670299727515,
+      "loss": 3.5109,
+      "step": 19850
+    },
+    {
+      "epoch": 6.015723270440252,
+      "grad_norm": 0.33696576952934265,
+      "learning_rate": 0.000420172570390554,
+      "loss": 3.4014,
+      "step": 19900
+    },
+    {
+      "epoch": 6.030841799709724,
+      "grad_norm": 0.3408699631690979,
+      "learning_rate": 0.00041971843778383286,
+      "loss": 3.3988,
+      "step": 19950
+    },
+    {
+      "epoch": 6.045960328979197,
+      "grad_norm": 0.340057909488678,
+      "learning_rate": 0.0004192643051771117,
+      "loss": 3.411,
+      "step": 20000
+    },
+    {
+      "epoch": 6.045960328979197,
+      "eval_accuracy": 0.3621478635601953,
+      "eval_loss": 3.6226511001586914,
+      "eval_runtime": 181.7969,
+      "eval_samples_per_second": 91.553,
+      "eval_steps_per_second": 5.726,
+      "step": 20000
+    },
+    {
+      "epoch": 6.06107885824867,
+      "grad_norm": 0.36212554574012756,
+      "learning_rate": 0.0004188101725703905,
+      "loss": 3.4062,
+      "step": 20050
+    },
+    {
+      "epoch": 6.0761973875181425,
+      "grad_norm": 0.35196641087532043,
+      "learning_rate": 0.00041835603996366935,
+      "loss": 3.4145,
+      "step": 20100
+    },
+    {
+      "epoch": 6.0913159167876145,
+      "grad_norm": 0.35778287053108215,
+      "learning_rate": 0.00041790190735694824,
+      "loss": 3.4204,
+      "step": 20150
+    },
+    {
+      "epoch": 6.106434446057087,
+      "grad_norm": 0.34076035022735596,
+      "learning_rate": 0.000417447774750227,
+      "loss": 3.4187,
+      "step": 20200
+    },
+    {
+      "epoch": 6.12155297532656,
+      "grad_norm": 0.3436349034309387,
+      "learning_rate": 0.00041699364214350585,
+      "loss": 3.4213,
+      "step": 20250
+    },
+    {
+      "epoch": 6.136671504596033,
+      "grad_norm": 0.32700756192207336,
+      "learning_rate": 0.0004165395095367847,
+      "loss": 3.4231,
+      "step": 20300
+    },
+    {
+      "epoch": 6.151790033865505,
+      "grad_norm": 0.3409166932106018,
+      "learning_rate": 0.00041608537693006356,
+      "loss": 3.408,
+      "step": 20350
+    },
+    {
+      "epoch": 6.166908563134978,
+      "grad_norm": 0.3603594899177551,
+      "learning_rate": 0.0004156312443233424,
+      "loss": 3.4209,
+      "step": 20400
+    },
+    {
+      "epoch": 6.182027092404451,
+      "grad_norm": 0.3426113426685333,
+      "learning_rate": 0.0004151771117166212,
+      "loss": 3.4202,
+      "step": 20450
+    },
+    {
+      "epoch": 6.197145621673924,
+      "grad_norm": 0.3426719009876251,
+      "learning_rate": 0.0004147229791099,
+      "loss": 3.4261,
+      "step": 20500
+    },
+    {
+      "epoch": 6.212264150943396,
+      "grad_norm": 0.34170880913734436,
+      "learning_rate": 0.0004142688465031789,
+      "loss": 3.423,
+      "step": 20550
+    },
+    {
+      "epoch": 6.227382680212869,
+      "grad_norm": 0.33598586916923523,
+      "learning_rate": 0.0004138147138964577,
+      "loss": 3.4305,
+      "step": 20600
+    },
+    {
+      "epoch": 6.2425012094823416,
+      "grad_norm": 0.331582635641098,
+      "learning_rate": 0.00041336058128973655,
+      "loss": 3.4281,
+      "step": 20650
+    },
+    {
+      "epoch": 6.257619738751814,
+      "grad_norm": 0.33808058500289917,
+      "learning_rate": 0.00041290644868301543,
+      "loss": 3.4321,
+      "step": 20700
+    },
+    {
+      "epoch": 6.272738268021287,
+      "grad_norm": 0.34110236167907715,
+      "learning_rate": 0.00041245231607629426,
+      "loss": 3.4264,
+      "step": 20750
+    },
+    {
+      "epoch": 6.287856797290759,
+      "grad_norm": 0.34383150935173035,
+      "learning_rate": 0.0004119981834695731,
+      "loss": 3.438,
+      "step": 20800
+    },
+    {
+      "epoch": 6.302975326560232,
+      "grad_norm": 0.34354880452156067,
+      "learning_rate": 0.00041154405086285187,
+      "loss": 3.4231,
+      "step": 20850
+    },
+    {
+      "epoch": 6.318093855829705,
+      "grad_norm": 0.3443782329559326,
+      "learning_rate": 0.00041108991825613075,
+      "loss": 3.4324,
+      "step": 20900
+    },
+    {
+      "epoch": 6.333212385099178,
+      "grad_norm": 0.3713175654411316,
+      "learning_rate": 0.0004106357856494096,
+      "loss": 3.4445,
+      "step": 20950
+    },
+    {
+      "epoch": 6.34833091436865,
+      "grad_norm": 0.32927024364471436,
+      "learning_rate": 0.0004101816530426884,
+      "loss": 3.4363,
+      "step": 21000
+    },
+    {
+      "epoch": 6.34833091436865,
+      "eval_accuracy": 0.36282722192984806,
+      "eval_loss": 3.619049310684204,
+      "eval_runtime": 181.9714,
+      "eval_samples_per_second": 91.465,
+      "eval_steps_per_second": 5.721,
+      "step": 21000
+    },
+    {
+      "epoch": 6.363449443638123,
+      "grad_norm": 0.33876073360443115,
+      "learning_rate": 0.00040972752043596725,
+      "loss": 3.4475,
+      "step": 21050
+    },
+    {
+      "epoch": 6.378567972907596,
+      "grad_norm": 0.3428070843219757,
+      "learning_rate": 0.00040927338782924613,
+      "loss": 3.4501,
+      "step": 21100
+    },
+    {
+      "epoch": 6.393686502177069,
+      "grad_norm": 0.33244138956069946,
+      "learning_rate": 0.00040881925522252496,
+      "loss": 3.4431,
+      "step": 21150
+    },
+    {
+      "epoch": 6.408805031446541,
+      "grad_norm": 0.3557736575603485,
+      "learning_rate": 0.00040836512261580374,
+      "loss": 3.4562,
+      "step": 21200
+    },
+    {
+      "epoch": 6.4239235607160134,
+      "grad_norm": 0.3500126600265503,
+      "learning_rate": 0.00040791099000908257,
+      "loss": 3.451,
+      "step": 21250
+    },
+    {
+      "epoch": 6.439042089985486,
+      "grad_norm": 0.3625536561012268,
+      "learning_rate": 0.00040745685740236145,
+      "loss": 3.4482,
+      "step": 21300
+    },
+    {
+      "epoch": 6.454160619254959,
+      "grad_norm": 0.34508174657821655,
+      "learning_rate": 0.0004070027247956403,
+      "loss": 3.4541,
+      "step": 21350
+    },
+    {
+      "epoch": 6.469279148524431,
+      "grad_norm": 0.3331073224544525,
+      "learning_rate": 0.0004065485921889191,
+      "loss": 3.4521,
+      "step": 21400
+    },
+    {
+      "epoch": 6.484397677793904,
+      "grad_norm": 0.332691490650177,
+      "learning_rate": 0.000406094459582198,
+      "loss": 3.4507,
+      "step": 21450
+    },
+    {
+      "epoch": 6.499516207063377,
+      "grad_norm": 0.32340583205223083,
+      "learning_rate": 0.00040564032697547683,
+      "loss": 3.4496,
+      "step": 21500
+    },
+    {
+      "epoch": 6.51463473633285,
+      "grad_norm": 0.33922895789146423,
+      "learning_rate": 0.0004051861943687556,
+      "loss": 3.4528,
+      "step": 21550
+    },
+    {
+      "epoch": 6.529753265602322,
+      "grad_norm": 0.342859148979187,
+      "learning_rate": 0.00040473206176203444,
+      "loss": 3.4543,
+      "step": 21600
+    },
+    {
+      "epoch": 6.544871794871795,
+      "grad_norm": 0.3467361032962799,
+      "learning_rate": 0.0004042779291553133,
+      "loss": 3.4578,
+      "step": 21650
+    },
+    {
+      "epoch": 6.559990324141268,
+      "grad_norm": 0.3421052396297455,
+      "learning_rate": 0.00040382379654859215,
+      "loss": 3.4531,
+      "step": 21700
+    },
+    {
+      "epoch": 6.5751088534107405,
+      "grad_norm": 0.3430230915546417,
+      "learning_rate": 0.000403369663941871,
+      "loss": 3.4529,
+      "step": 21750
+    },
+    {
+      "epoch": 6.590227382680213,
+      "grad_norm": 0.3446337878704071,
+      "learning_rate": 0.0004029155313351498,
+      "loss": 3.4538,
+      "step": 21800
+    },
+    {
+      "epoch": 6.605345911949685,
+      "grad_norm": 0.3413420617580414,
+      "learning_rate": 0.0004024613987284287,
+      "loss": 3.4459,
+      "step": 21850
+    },
+    {
+      "epoch": 6.620464441219158,
+      "grad_norm": 0.36404335498809814,
+      "learning_rate": 0.00040200726612170753,
+      "loss": 3.4481,
+      "step": 21900
+    },
+    {
+      "epoch": 6.635582970488631,
+      "grad_norm": 0.3348482549190521,
+      "learning_rate": 0.0004015531335149863,
+      "loss": 3.4477,
+      "step": 21950
+    },
+    {
+      "epoch": 6.650701499758103,
+      "grad_norm": 0.33620691299438477,
+      "learning_rate": 0.0004010990009082652,
+      "loss": 3.4492,
+      "step": 22000
+    },
+    {
+      "epoch": 6.650701499758103,
+      "eval_accuracy": 0.3638837664625064,
+      "eval_loss": 3.6092684268951416,
+      "eval_runtime": 181.7457,
+      "eval_samples_per_second": 91.579,
+      "eval_steps_per_second": 5.728,
+      "step": 22000
+    },
+    {
+      "epoch": 6.665820029027576,
+      "grad_norm": 0.3584924042224884,
+      "learning_rate": 0.000400644868301544,
+      "loss": 3.4621,
+      "step": 22050
+    },
+    {
+      "epoch": 6.680938558297049,
+      "grad_norm": 0.342899352312088,
+      "learning_rate": 0.00040019073569482285,
+      "loss": 3.4605,
+      "step": 22100
+    },
+    {
+      "epoch": 6.696057087566522,
+      "grad_norm": 0.3478437066078186,
+      "learning_rate": 0.0003997366030881017,
+      "loss": 3.4763,
+      "step": 22150
+    },
+    {
+      "epoch": 6.711175616835995,
+      "grad_norm": 0.3423757553100586,
+      "learning_rate": 0.00039928247048138057,
+      "loss": 3.4587,
+      "step": 22200
+    },
+    {
+      "epoch": 6.726294146105467,
+      "grad_norm": 0.3426273465156555,
+      "learning_rate": 0.0003988283378746594,
+      "loss": 3.453,
+      "step": 22250
+    },
+    {
+      "epoch": 6.7414126753749395,
+      "grad_norm": 0.3257652223110199,
+      "learning_rate": 0.0003983742052679382,
+      "loss": 3.4499,
+      "step": 22300
+    },
+    {
+      "epoch": 6.756531204644412,
+      "grad_norm": 0.32835474610328674,
+      "learning_rate": 0.000397920072661217,
+      "loss": 3.4683,
+      "step": 22350
+    },
+    {
+      "epoch": 6.771649733913884,
+      "grad_norm": 0.3534278869628906,
+      "learning_rate": 0.0003974659400544959,
+      "loss": 3.461,
+      "step": 22400
+    },
+    {
+      "epoch": 6.786768263183357,
+      "grad_norm": 0.3387344479560852,
+      "learning_rate": 0.0003970118074477747,
+      "loss": 3.4594,
+      "step": 22450
+    },
+    {
+      "epoch": 6.80188679245283,
+      "grad_norm": 0.3469718396663666,
+      "learning_rate": 0.00039655767484105355,
+      "loss": 3.4633,
+      "step": 22500
+    },
+    {
+      "epoch": 6.817005321722303,
+      "grad_norm": 0.34182408452033997,
+      "learning_rate": 0.0003961035422343324,
+      "loss": 3.4561,
+      "step": 22550
+    },
+    {
+      "epoch": 6.832123850991776,
+      "grad_norm": 0.35019853711128235,
+      "learning_rate": 0.00039564940962761127,
+      "loss": 3.4721,
+      "step": 22600
+    },
+    {
+      "epoch": 6.847242380261248,
+      "grad_norm": 0.32239803671836853,
+      "learning_rate": 0.00039519527702089005,
+      "loss": 3.4599,
+      "step": 22650
+    },
+    {
+      "epoch": 6.862360909530721,
+      "grad_norm": 0.35213080048561096,
+      "learning_rate": 0.0003947411444141689,
+      "loss": 3.4652,
+      "step": 22700
+    },
+    {
+      "epoch": 6.877479438800194,
+      "grad_norm": 0.3404783010482788,
+      "learning_rate": 0.00039428701180744776,
+      "loss": 3.4592,
+      "step": 22750
+    },
+    {
+      "epoch": 6.8925979680696665,
+      "grad_norm": 0.33358967304229736,
+      "learning_rate": 0.0003938328792007266,
+      "loss": 3.4612,
+      "step": 22800
+    },
+    {
+      "epoch": 6.9077164973391385,
+      "grad_norm": 0.3438171446323395,
+      "learning_rate": 0.0003933787465940054,
+      "loss": 3.4415,
+      "step": 22850
+    },
+    {
+      "epoch": 6.922835026608611,
+      "grad_norm": 0.3446924090385437,
+      "learning_rate": 0.00039292461398728425,
+      "loss": 3.4612,
+      "step": 22900
+    },
+    {
+      "epoch": 6.937953555878084,
+      "grad_norm": 0.34483805298805237,
+      "learning_rate": 0.00039247048138056314,
+      "loss": 3.475,
+      "step": 22950
+    },
+    {
+      "epoch": 6.953072085147557,
+      "grad_norm": 0.3411277234554291,
+      "learning_rate": 0.0003920163487738419,
+      "loss": 3.4632,
+      "step": 23000
+    },
+    {
+      "epoch": 6.953072085147557,
+      "eval_accuracy": 0.36477981875311283,
+      "eval_loss": 3.597496271133423,
+      "eval_runtime": 181.6718,
+      "eval_samples_per_second": 91.616,
+      "eval_steps_per_second": 5.73,
+      "step": 23000
+    },
+    {
+      "epoch": 6.968190614417029,
+      "grad_norm": 0.3497656583786011,
+      "learning_rate": 0.00039156221616712075,
+      "loss": 3.4725,
+      "step": 23050
+    },
+    {
+      "epoch": 6.983309143686502,
+      "grad_norm": 0.3395549952983856,
+      "learning_rate": 0.0003911080835603996,
+      "loss": 3.4574,
+      "step": 23100
+    },
+    {
+      "epoch": 6.998427672955975,
+      "grad_norm": 0.35077255964279175,
+      "learning_rate": 0.00039065395095367846,
+      "loss": 3.4717,
+      "step": 23150
+    },
+    {
+      "epoch": 7.013304305757136,
+      "grad_norm": 0.3488810956478119,
+      "learning_rate": 0.0003901998183469573,
+      "loss": 3.3635,
+      "step": 23200
+    },
+    {
+      "epoch": 7.028422835026609,
+      "grad_norm": 0.3391869068145752,
+      "learning_rate": 0.0003897456857402361,
+      "loss": 3.3509,
+      "step": 23250
+    },
+    {
+      "epoch": 7.043541364296082,
+      "grad_norm": 0.3469946086406708,
+      "learning_rate": 0.0003892915531335149,
+      "loss": 3.3675,
+      "step": 23300
+    },
+    {
+      "epoch": 7.058659893565554,
+      "grad_norm": 0.36358246207237244,
+      "learning_rate": 0.0003888374205267938,
+      "loss": 3.3506,
+      "step": 23350
+    },
+    {
+      "epoch": 7.0737784228350264,
+      "grad_norm": 0.33452320098876953,
+      "learning_rate": 0.0003883832879200726,
+      "loss": 3.3665,
+      "step": 23400
+    },
+    {
+      "epoch": 7.088896952104499,
+      "grad_norm": 0.34786656498908997,
+      "learning_rate": 0.00038792915531335145,
+      "loss": 3.3834,
+      "step": 23450
+    },
+    {
+      "epoch": 7.104015481373972,
+      "grad_norm": 0.36504653096199036,
+      "learning_rate": 0.00038747502270663033,
+      "loss": 3.3682,
+      "step": 23500
+    },
+    {
+      "epoch": 7.119134010643444,
+      "grad_norm": 0.345797061920166,
+      "learning_rate": 0.00038702089009990916,
+      "loss": 3.3713,
+      "step": 23550
+    },
+    {
+      "epoch": 7.134252539912917,
+      "grad_norm": 0.3616042137145996,
+      "learning_rate": 0.000386566757493188,
+      "loss": 3.3793,
+      "step": 23600
+    },
+    {
+      "epoch": 7.14937106918239,
+      "grad_norm": 0.33136123418807983,
+      "learning_rate": 0.00038611262488646677,
+      "loss": 3.3777,
+      "step": 23650
+    },
+    {
+      "epoch": 7.164489598451863,
+      "grad_norm": 0.36124685406684875,
+      "learning_rate": 0.00038565849227974565,
+      "loss": 3.3797,
+      "step": 23700
+    },
+    {
+      "epoch": 7.179608127721336,
+      "grad_norm": 0.3644281327724457,
+      "learning_rate": 0.0003852043596730245,
+      "loss": 3.3808,
+      "step": 23750
+    },
+    {
+      "epoch": 7.194726656990808,
+      "grad_norm": 0.3486228287220001,
+      "learning_rate": 0.0003847502270663033,
+      "loss": 3.3862,
+      "step": 23800
+    },
+    {
+      "epoch": 7.209845186260281,
+      "grad_norm": 0.350700706243515,
+      "learning_rate": 0.00038429609445958215,
+      "loss": 3.3776,
+      "step": 23850
+    },
+    {
+      "epoch": 7.2249637155297535,
+      "grad_norm": 0.35353735089302063,
+      "learning_rate": 0.00038384196185286103,
+      "loss": 3.3948,
+      "step": 23900
+    },
+    {
+      "epoch": 7.240082244799226,
+      "grad_norm": 0.34329304099082947,
+      "learning_rate": 0.00038338782924613986,
+      "loss": 3.3934,
+      "step": 23950
+    },
+    {
+      "epoch": 7.255200774068698,
+      "grad_norm": 0.37160131335258484,
+      "learning_rate": 0.00038293369663941864,
+      "loss": 3.3846,
+      "step": 24000
+    },
+    {
+      "epoch": 7.255200774068698,
+      "eval_accuracy": 0.36480168802565616,
+      "eval_loss": 3.606091260910034,
+      "eval_runtime": 181.8857,
+      "eval_samples_per_second": 91.508,
+      "eval_steps_per_second": 5.723,
+      "step": 24000
+    },
+    {
+      "epoch": 7.270319303338171,
+      "grad_norm": 0.36367154121398926,
+      "learning_rate": 0.00038247956403269747,
+      "loss": 3.3927,
+      "step": 24050
+    },
+    {
+      "epoch": 7.285437832607644,
+      "grad_norm": 0.3284398913383484,
+      "learning_rate": 0.00038202543142597635,
+      "loss": 3.4085,
+      "step": 24100
+    },
+    {
+      "epoch": 7.300556361877117,
+      "grad_norm": 0.36100390553474426,
+      "learning_rate": 0.0003815712988192552,
+      "loss": 3.4062,
+      "step": 24150
+    },
+    {
+      "epoch": 7.315674891146589,
+      "grad_norm": 0.3739434778690338,
+      "learning_rate": 0.000381117166212534,
+      "loss": 3.3941,
+      "step": 24200
+    },
+    {
+      "epoch": 7.330793420416062,
+      "grad_norm": 0.3636663854122162,
+      "learning_rate": 0.0003806630336058129,
+      "loss": 3.3991,
+      "step": 24250
+    },
+    {
+      "epoch": 7.345911949685535,
+      "grad_norm": 0.3598420023918152,
+      "learning_rate": 0.00038020890099909173,
+      "loss": 3.4083,
+      "step": 24300
+    },
+    {
+      "epoch": 7.361030478955008,
+      "grad_norm": 0.37003642320632935,
+      "learning_rate": 0.0003797547683923705,
+      "loss": 3.4077,
+      "step": 24350
+    },
+    {
+      "epoch": 7.37614900822448,
+      "grad_norm": 0.34832462668418884,
+      "learning_rate": 0.00037930063578564934,
+      "loss": 3.3993,
+      "step": 24400
+    },
+    {
+      "epoch": 7.3912675374939525,
+      "grad_norm": 0.3355371356010437,
+      "learning_rate": 0.0003788465031789282,
+      "loss": 3.3961,
+      "step": 24450
+    },
+    {
+      "epoch": 7.406386066763425,
+      "grad_norm": 0.3435608148574829,
+      "learning_rate": 0.00037839237057220705,
+      "loss": 3.4002,
+      "step": 24500
+    },
+    {
+      "epoch": 7.421504596032898,
+      "grad_norm": 0.3562789559364319,
+      "learning_rate": 0.0003779382379654859,
+      "loss": 3.4164,
+      "step": 24550
+    },
+    {
+      "epoch": 7.43662312530237,
+      "grad_norm": 0.33656221628189087,
+      "learning_rate": 0.0003774841053587647,
+      "loss": 3.4101,
+      "step": 24600
+    },
+    {
+      "epoch": 7.451741654571843,
+      "grad_norm": 0.3509146273136139,
+      "learning_rate": 0.0003770299727520436,
+      "loss": 3.4069,
+      "step": 24650
+    },
+    {
+      "epoch": 7.466860183841316,
+      "grad_norm": 0.3496881425380707,
+      "learning_rate": 0.00037657584014532243,
+      "loss": 3.4107,
+      "step": 24700
+    },
+    {
+      "epoch": 7.481978713110789,
+      "grad_norm": 0.3623145520687103,
+      "learning_rate": 0.0003761217075386012,
+      "loss": 3.4143,
+      "step": 24750
+    },
+    {
+      "epoch": 7.497097242380261,
+      "grad_norm": 0.3448486924171448,
+      "learning_rate": 0.00037566757493188004,
+      "loss": 3.4059,
+      "step": 24800
+    },
+    {
+      "epoch": 7.512215771649734,
+      "grad_norm": 0.35689592361450195,
+      "learning_rate": 0.0003752134423251589,
+      "loss": 3.4122,
+      "step": 24850
+    },
+    {
+      "epoch": 7.527334300919207,
+      "grad_norm": 0.36233118176460266,
+      "learning_rate": 0.00037475930971843775,
+      "loss": 3.4118,
+      "step": 24900
+    },
+    {
+      "epoch": 7.5424528301886795,
+      "grad_norm": 0.3479229211807251,
+      "learning_rate": 0.0003743051771117166,
+      "loss": 3.4052,
+      "step": 24950
+    },
+    {
+      "epoch": 7.5575713594581515,
+      "grad_norm": 0.338264524936676,
+      "learning_rate": 0.00037385104450499547,
+      "loss": 3.4205,
+      "step": 25000
+    },
+    {
+      "epoch": 7.5575713594581515,
+      "eval_accuracy": 0.36550926481149393,
+      "eval_loss": 3.5999324321746826,
+      "eval_runtime": 182.6684,
+      "eval_samples_per_second": 91.116,
+      "eval_steps_per_second": 5.699,
+      "step": 25000
+    },
+    {
+      "epoch": 7.572689888727624,
+      "grad_norm": 0.3575372099876404,
+      "learning_rate": 0.0003733969118982743,
+      "loss": 3.4177,
+      "step": 25050
+    },
+    {
+      "epoch": 7.587808417997097,
+      "grad_norm": 0.356697678565979,
+      "learning_rate": 0.0003729427792915531,
+      "loss": 3.4052,
+      "step": 25100
+    },
+    {
+      "epoch": 7.60292694726657,
+      "grad_norm": 0.340265154838562,
+      "learning_rate": 0.0003724886466848319,
+      "loss": 3.4063,
+      "step": 25150
+    },
+    {
+      "epoch": 7.618045476536043,
+      "grad_norm": 0.36485445499420166,
+      "learning_rate": 0.0003720345140781108,
+      "loss": 3.4231,
+      "step": 25200
+    },
+    {
+      "epoch": 7.633164005805515,
+      "grad_norm": 0.3679625988006592,
+      "learning_rate": 0.0003715803814713896,
+      "loss": 3.4219,
+      "step": 25250
+    },
+    {
+      "epoch": 7.648282535074988,
+      "grad_norm": 0.36495083570480347,
+      "learning_rate": 0.00037112624886466845,
+      "loss": 3.4118,
+      "step": 25300
+    },
+    {
+      "epoch": 7.663401064344461,
+      "grad_norm": 0.3572041094303131,
+      "learning_rate": 0.0003706721162579473,
+      "loss": 3.4184,
+      "step": 25350
+    },
+    {
+      "epoch": 7.678519593613933,
+      "grad_norm": 0.3356753885746002,
+      "learning_rate": 0.00037021798365122617,
+      "loss": 3.4145,
+      "step": 25400
+    },
+    {
+      "epoch": 7.693638122883406,
+      "grad_norm": 0.34812238812446594,
+      "learning_rate": 0.00036976385104450495,
+      "loss": 3.4273,
+      "step": 25450
+    },
+    {
+      "epoch": 7.7087566521528785,
+      "grad_norm": 0.35833728313446045,
+      "learning_rate": 0.0003693097184377838,
+      "loss": 3.4167,
+      "step": 25500
+    },
+    {
+      "epoch": 7.723875181422351,
+      "grad_norm": 0.35254955291748047,
+      "learning_rate": 0.0003688555858310626,
+      "loss": 3.4091,
+      "step": 25550
+    },
+    {
+      "epoch": 7.738993710691824,
+      "grad_norm": 0.36359110474586487,
+      "learning_rate": 0.0003684014532243415,
+      "loss": 3.4176,
+      "step": 25600
+    },
+    {
+      "epoch": 7.754112239961296,
+      "grad_norm": 0.3410344421863556,
+      "learning_rate": 0.0003679473206176203,
+      "loss": 3.4118,
+      "step": 25650
+    },
+    {
+      "epoch": 7.769230769230769,
+      "grad_norm": 0.3466271162033081,
+      "learning_rate": 0.00036749318801089915,
+      "loss": 3.4288,
+      "step": 25700
+    },
+    {
+      "epoch": 7.784349298500242,
+      "grad_norm": 0.34525296092033386,
+      "learning_rate": 0.00036703905540417804,
+      "loss": 3.4064,
+      "step": 25750
+    },
+    {
+      "epoch": 7.799467827769715,
+      "grad_norm": 0.34507691860198975,
+      "learning_rate": 0.0003665849227974568,
+      "loss": 3.4162,
+      "step": 25800
+    },
+    {
+      "epoch": 7.814586357039187,
+      "grad_norm": 0.3572472333908081,
+      "learning_rate": 0.00036613079019073565,
+      "loss": 3.4211,
+      "step": 25850
+    },
+    {
+      "epoch": 7.82970488630866,
+      "grad_norm": 0.3335019052028656,
+      "learning_rate": 0.0003656766575840145,
+      "loss": 3.4183,
+      "step": 25900
+    },
+    {
+      "epoch": 7.844823415578133,
+      "grad_norm": 0.34102049469947815,
+      "learning_rate": 0.00036522252497729336,
+      "loss": 3.4322,
+      "step": 25950
+    },
+    {
+      "epoch": 7.8599419448476056,
+      "grad_norm": 0.3497759699821472,
+      "learning_rate": 0.0003647683923705722,
+      "loss": 3.4231,
+      "step": 26000
+    },
+    {
+      "epoch": 7.8599419448476056,
+      "eval_accuracy": 0.3660651676103375,
+      "eval_loss": 3.5885255336761475,
+      "eval_runtime": 182.172,
+      "eval_samples_per_second": 91.364,
+      "eval_steps_per_second": 5.714,
+      "step": 26000
+    },
+    {
+      "epoch": 7.8750604741170775,
+      "grad_norm": 0.35175567865371704,
+      "learning_rate": 0.000364314259763851,
+      "loss": 3.4266,
+      "step": 26050
+    },
+    {
+      "epoch": 7.89017900338655,
+      "grad_norm": 0.3355766534805298,
+      "learning_rate": 0.0003638601271571298,
+      "loss": 3.4316,
+      "step": 26100
+    },
+    {
+      "epoch": 7.905297532656023,
+      "grad_norm": 0.36761733889579773,
+      "learning_rate": 0.0003634059945504087,
+      "loss": 3.4175,
+      "step": 26150
+    },
+    {
+      "epoch": 7.920416061925496,
+      "grad_norm": 0.35893869400024414,
+      "learning_rate": 0.0003629518619436875,
+      "loss": 3.4332,
+      "step": 26200
+    },
+    {
+      "epoch": 7.935534591194968,
+      "grad_norm": 0.3662582337856293,
+      "learning_rate": 0.00036249772933696635,
+      "loss": 3.4186,
+      "step": 26250
+    },
+    {
+      "epoch": 7.950653120464441,
+      "grad_norm": 0.36315032839775085,
+      "learning_rate": 0.0003620435967302452,
+      "loss": 3.4248,
+      "step": 26300
+    },
+    {
+      "epoch": 7.965771649733914,
+      "grad_norm": 0.3308250308036804,
+      "learning_rate": 0.00036158946412352406,
+      "loss": 3.4402,
+      "step": 26350
+    },
+    {
+      "epoch": 7.980890179003387,
+      "grad_norm": 0.3648611307144165,
+      "learning_rate": 0.0003611353315168029,
+      "loss": 3.4168,
+      "step": 26400
+    },
+    {
+      "epoch": 7.996008708272859,
+      "grad_norm": 0.342036634683609,
+      "learning_rate": 0.00036068119891008167,
+      "loss": 3.4285,
+      "step": 26450
+    },
+    {
+      "epoch": 8.01088534107402,
+      "grad_norm": 0.37517133355140686,
+      "learning_rate": 0.00036022706630336055,
+      "loss": 3.3465,
+      "step": 26500
+    },
+    {
+      "epoch": 8.026003870343493,
+      "grad_norm": 0.346508651971817,
+      "learning_rate": 0.0003597729336966394,
+      "loss": 3.3161,
+      "step": 26550
+    },
+    {
+      "epoch": 8.041122399612966,
+      "grad_norm": 0.37528109550476074,
+      "learning_rate": 0.0003593188010899182,
+      "loss": 3.3215,
+      "step": 26600
+    },
+    {
+      "epoch": 8.056240928882438,
+      "grad_norm": 0.3728856146335602,
+      "learning_rate": 0.00035886466848319705,
+      "loss": 3.3283,
+      "step": 26650
+    },
+    {
+      "epoch": 8.07135945815191,
+      "grad_norm": 0.3430670201778412,
+      "learning_rate": 0.00035841053587647593,
+      "loss": 3.3318,
+      "step": 26700
+    },
+    {
+      "epoch": 8.086477987421384,
+      "grad_norm": 0.3535607159137726,
+      "learning_rate": 0.00035795640326975476,
+      "loss": 3.3282,
+      "step": 26750
+    },
+    {
+      "epoch": 8.101596516690856,
+      "grad_norm": 0.3596191108226776,
+      "learning_rate": 0.00035750227066303354,
+      "loss": 3.3194,
+      "step": 26800
+    },
+    {
+      "epoch": 8.116715045960328,
+      "grad_norm": 0.36048582196235657,
+      "learning_rate": 0.00035704813805631237,
+      "loss": 3.3234,
+      "step": 26850
+    },
+    {
+      "epoch": 8.131833575229802,
+      "grad_norm": 0.3535524904727936,
+      "learning_rate": 0.00035659400544959125,
+      "loss": 3.3467,
+      "step": 26900
+    },
+    {
+      "epoch": 8.146952104499274,
+      "grad_norm": 0.3580203950405121,
+      "learning_rate": 0.0003561398728428701,
+      "loss": 3.3281,
+      "step": 26950
+    },
+    {
+      "epoch": 8.162070633768748,
+      "grad_norm": 0.36240702867507935,
+      "learning_rate": 0.0003556857402361489,
+      "loss": 3.349,
+      "step": 27000
+    },
+    {
+      "epoch": 8.162070633768748,
+      "eval_accuracy": 0.36599462156987517,
+      "eval_loss": 3.598177671432495,
+      "eval_runtime": 182.0613,
+      "eval_samples_per_second": 91.42,
+      "eval_steps_per_second": 5.718,
+      "step": 27000
+    },
+    {
+      "epoch": 8.17718916303822,
+      "grad_norm": 0.36847227811813354,
+      "learning_rate": 0.0003552316076294278,
+      "loss": 3.3383,
+      "step": 27050
+    },
+    {
+      "epoch": 8.192307692307692,
+      "grad_norm": 0.3737618029117584,
+      "learning_rate": 0.00035477747502270663,
+      "loss": 3.3452,
+      "step": 27100
+    },
+    {
+      "epoch": 8.207426221577165,
+      "grad_norm": 0.36412081122398376,
+      "learning_rate": 0.00035432334241598546,
+      "loss": 3.3605,
+      "step": 27150
+    },
+    {
+      "epoch": 8.222544750846637,
+      "grad_norm": 0.3439239263534546,
+      "learning_rate": 0.00035386920980926424,
+      "loss": 3.3609,
+      "step": 27200
+    },
+    {
+      "epoch": 8.237663280116111,
+      "grad_norm": 0.35170575976371765,
+      "learning_rate": 0.0003534150772025431,
+      "loss": 3.3691,
+      "step": 27250
+    },
+    {
+      "epoch": 8.252781809385583,
+      "grad_norm": 0.36688730120658875,
+      "learning_rate": 0.00035296094459582195,
+      "loss": 3.3521,
+      "step": 27300
+    },
+    {
+      "epoch": 8.267900338655055,
+      "grad_norm": 0.3887461721897125,
+      "learning_rate": 0.0003525068119891008,
+      "loss": 3.3552,
+      "step": 27350
+    },
+    {
+      "epoch": 8.283018867924529,
+      "grad_norm": 0.37762829661369324,
+      "learning_rate": 0.0003520526793823796,
+      "loss": 3.3567,
+      "step": 27400
+    },
+    {
+      "epoch": 8.298137397194001,
+      "grad_norm": 0.3905569911003113,
+      "learning_rate": 0.0003515985467756585,
+      "loss": 3.3724,
+      "step": 27450
+    },
+    {
+      "epoch": 8.313255926463473,
+      "grad_norm": 0.3466584384441376,
+      "learning_rate": 0.00035114441416893733,
+      "loss": 3.3576,
+      "step": 27500
+    },
+    {
+      "epoch": 8.328374455732947,
+      "grad_norm": 0.3696681559085846,
+      "learning_rate": 0.0003506902815622161,
+      "loss": 3.3526,
+      "step": 27550
+    },
+    {
+      "epoch": 8.343492985002419,
+      "grad_norm": 0.3654014468193054,
+      "learning_rate": 0.00035023614895549494,
+      "loss": 3.3696,
+      "step": 27600
+    },
+    {
+      "epoch": 8.358611514271892,
+      "grad_norm": 0.35293635725975037,
+      "learning_rate": 0.0003497820163487738,
+      "loss": 3.3705,
+      "step": 27650
+    },
+    {
+      "epoch": 8.373730043541364,
+      "grad_norm": 0.35331737995147705,
+      "learning_rate": 0.00034932788374205265,
+      "loss": 3.3732,
+      "step": 27700
+    },
+    {
+      "epoch": 8.388848572810836,
+      "grad_norm": 0.35930585861206055,
+      "learning_rate": 0.0003488737511353315,
+      "loss": 3.3677,
+      "step": 27750
+    },
+    {
+      "epoch": 8.40396710208031,
+      "grad_norm": 0.3438582122325897,
+      "learning_rate": 0.00034841961852861037,
+      "loss": 3.3588,
+      "step": 27800
+    },
+    {
+      "epoch": 8.419085631349782,
+      "grad_norm": 0.34405338764190674,
+      "learning_rate": 0.0003479654859218892,
+      "loss": 3.3703,
+      "step": 27850
+    },
+    {
+      "epoch": 8.434204160619254,
+      "grad_norm": 0.34330499172210693,
+      "learning_rate": 0.000347511353315168,
+      "loss": 3.3554,
+      "step": 27900
+    },
+    {
+      "epoch": 8.449322689888728,
+      "grad_norm": 0.38266411423683167,
+      "learning_rate": 0.0003470572207084468,
+      "loss": 3.3699,
+      "step": 27950
+    },
+    {
+      "epoch": 8.4644412191582,
+      "grad_norm": 0.3567814528942108,
+      "learning_rate": 0.0003466030881017257,
+      "loss": 3.3654,
+      "step": 28000
+    },
+    {
+      "epoch": 8.4644412191582,
+      "eval_accuracy": 0.3666786830088921,
+      "eval_loss": 3.588674306869507,
+      "eval_runtime": 182.8049,
+      "eval_samples_per_second": 91.048,
+      "eval_steps_per_second": 5.695,
+      "step": 28000
+    },
+    {
+      "epoch": 8.479559748427674,
+      "grad_norm": 0.3407067358493805,
+      "learning_rate": 0.0003461489554950045,
+      "loss": 3.3759,
+      "step": 28050
+    },
+    {
+      "epoch": 8.494678277697146,
+      "grad_norm": 0.36050164699554443,
+      "learning_rate": 0.00034569482288828335,
+      "loss": 3.3732,
+      "step": 28100
+    },
+    {
+      "epoch": 8.509796806966618,
+      "grad_norm": 0.34035587310791016,
+      "learning_rate": 0.0003452406902815622,
+      "loss": 3.3653,
+      "step": 28150
+    },
+    {
+      "epoch": 8.524915336236091,
+      "grad_norm": 0.3827550709247589,
+      "learning_rate": 0.00034478655767484107,
+      "loss": 3.3745,
+      "step": 28200
+    },
+    {
+      "epoch": 8.540033865505563,
+      "grad_norm": 0.367535799741745,
+      "learning_rate": 0.00034433242506811984,
+      "loss": 3.3773,
+      "step": 28250
+    },
+    {
+      "epoch": 8.555152394775035,
+      "grad_norm": 0.35586029291152954,
+      "learning_rate": 0.0003438782924613987,
+      "loss": 3.381,
+      "step": 28300
+    },
+    {
+      "epoch": 8.57027092404451,
+      "grad_norm": 0.3650664687156677,
+      "learning_rate": 0.0003434241598546775,
+      "loss": 3.379,
+      "step": 28350
+    },
+    {
+      "epoch": 8.585389453313981,
+      "grad_norm": 0.35350197553634644,
+      "learning_rate": 0.0003429700272479564,
+      "loss": 3.3727,
+      "step": 28400
+    },
+    {
+      "epoch": 8.600507982583455,
+      "grad_norm": 0.3827017545700073,
+      "learning_rate": 0.0003425158946412352,
+      "loss": 3.3785,
+      "step": 28450
+    },
+    {
+      "epoch": 8.615626511852927,
+      "grad_norm": 0.3796612024307251,
+      "learning_rate": 0.00034206176203451405,
+      "loss": 3.3805,
+      "step": 28500
+    },
+    {
+      "epoch": 8.630745041122399,
+      "grad_norm": 0.36361196637153625,
+      "learning_rate": 0.00034160762942779294,
+      "loss": 3.3769,
+      "step": 28550
+    },
+    {
+      "epoch": 8.645863570391873,
+      "grad_norm": 0.365169882774353,
+      "learning_rate": 0.0003411534968210717,
+      "loss": 3.3815,
+      "step": 28600
+    },
+    {
+      "epoch": 8.660982099661345,
+      "grad_norm": 0.3632723391056061,
+      "learning_rate": 0.00034069936421435054,
+      "loss": 3.3771,
+      "step": 28650
+    },
+    {
+      "epoch": 8.676100628930818,
+      "grad_norm": 0.3668626546859741,
+      "learning_rate": 0.0003402452316076294,
+      "loss": 3.3811,
+      "step": 28700
+    },
+    {
+      "epoch": 8.69121915820029,
+      "grad_norm": 0.35274365544319153,
+      "learning_rate": 0.00033979109900090826,
+      "loss": 3.3836,
+      "step": 28750
+    },
+    {
+      "epoch": 8.706337687469762,
+      "grad_norm": 0.35408350825309753,
+      "learning_rate": 0.0003393369663941871,
+      "loss": 3.3814,
+      "step": 28800
+    },
+    {
+      "epoch": 8.721456216739236,
+      "grad_norm": 0.34329164028167725,
+      "learning_rate": 0.0003388828337874659,
+      "loss": 3.3864,
+      "step": 28850
+    },
+    {
+      "epoch": 8.736574746008708,
+      "grad_norm": 0.3636920750141144,
+      "learning_rate": 0.0003384287011807447,
+      "loss": 3.3855,
+      "step": 28900
+    },
+    {
+      "epoch": 8.75169327527818,
+      "grad_norm": 0.36424097418785095,
+      "learning_rate": 0.0003379745685740236,
+      "loss": 3.3914,
+      "step": 28950
+    },
+    {
+      "epoch": 8.766811804547654,
+      "grad_norm": 0.3490569293498993,
+      "learning_rate": 0.0003375204359673024,
+      "loss": 3.3779,
+      "step": 29000
+    },
+    {
+      "epoch": 8.766811804547654,
+      "eval_accuracy": 0.3673816743020998,
+      "eval_loss": 3.579127073287964,
+      "eval_runtime": 181.9617,
+      "eval_samples_per_second": 91.47,
+      "eval_steps_per_second": 5.721,
+      "step": 29000
+    },
+    {
+      "epoch": 8.781930333817126,
+      "grad_norm": 0.36123400926589966,
+      "learning_rate": 0.00033706630336058124,
+      "loss": 3.389,
+      "step": 29050
+    },
+    {
+      "epoch": 8.7970488630866,
+      "grad_norm": 0.34295931458473206,
+      "learning_rate": 0.0003366121707538601,
+      "loss": 3.3975,
+      "step": 29100
+    },
+    {
+      "epoch": 8.812167392356072,
+      "grad_norm": 0.38402318954467773,
+      "learning_rate": 0.00033615803814713896,
+      "loss": 3.3888,
+      "step": 29150
+    },
+    {
+      "epoch": 8.827285921625544,
+      "grad_norm": 0.3944171667098999,
+      "learning_rate": 0.0003357039055404178,
+      "loss": 3.3868,
+      "step": 29200
+    },
+    {
+      "epoch": 8.842404450895017,
+      "grad_norm": 0.3558756113052368,
+      "learning_rate": 0.00033524977293369657,
+      "loss": 3.3993,
+      "step": 29250
+    },
+    {
+      "epoch": 8.85752298016449,
+      "grad_norm": 0.3514937162399292,
+      "learning_rate": 0.00033479564032697545,
+      "loss": 3.3866,
+      "step": 29300
+    },
+    {
+      "epoch": 8.872641509433961,
+      "grad_norm": 0.363704651594162,
+      "learning_rate": 0.0003343415077202543,
+      "loss": 3.3843,
+      "step": 29350
+    },
+    {
+      "epoch": 8.887760038703435,
+      "grad_norm": 0.35803845524787903,
+      "learning_rate": 0.0003338873751135331,
+      "loss": 3.398,
+      "step": 29400
+    },
+    {
+      "epoch": 8.902878567972907,
+      "grad_norm": 0.39813512563705444,
+      "learning_rate": 0.00033343324250681194,
+      "loss": 3.4079,
+      "step": 29450
+    },
+    {
+      "epoch": 8.917997097242381,
+      "grad_norm": 0.34561270475387573,
+      "learning_rate": 0.00033297910990009083,
+      "loss": 3.3967,
+      "step": 29500
+    },
+    {
+      "epoch": 8.933115626511853,
+      "grad_norm": 0.377175509929657,
+      "learning_rate": 0.00033252497729336966,
+      "loss": 3.3931,
+      "step": 29550
+    },
+    {
+      "epoch": 8.948234155781325,
+      "grad_norm": 0.3505743145942688,
+      "learning_rate": 0.00033207084468664844,
+      "loss": 3.3993,
+      "step": 29600
+    },
+    {
+      "epoch": 8.963352685050799,
+      "grad_norm": 0.3531405031681061,
+      "learning_rate": 0.00033161671207992727,
+      "loss": 3.3932,
+      "step": 29650
+    },
+    {
+      "epoch": 8.97847121432027,
+      "grad_norm": 0.3554518222808838,
+      "learning_rate": 0.00033116257947320615,
+      "loss": 3.3915,
+      "step": 29700
+    },
+    {
+      "epoch": 8.993589743589745,
+      "grad_norm": 0.34890449047088623,
+      "learning_rate": 0.000330708446866485,
+      "loss": 3.3907,
+      "step": 29750
+    },
+    {
+      "epoch": 9.008466376390905,
+      "grad_norm": 0.38579925894737244,
+      "learning_rate": 0.0003302543142597638,
+      "loss": 3.3333,
+      "step": 29800
+    },
+    {
+      "epoch": 9.023584905660377,
+      "grad_norm": 0.3463684618473053,
+      "learning_rate": 0.00032980018165304264,
+      "loss": 3.2836,
+      "step": 29850
+    },
+    {
+      "epoch": 9.03870343492985,
+      "grad_norm": 0.3551153838634491,
+      "learning_rate": 0.00032934604904632153,
+      "loss": 3.2839,
+      "step": 29900
+    },
+    {
+      "epoch": 9.053821964199322,
+      "grad_norm": 0.36974823474884033,
+      "learning_rate": 0.00032889191643960036,
+      "loss": 3.2932,
+      "step": 29950
+    },
+    {
+      "epoch": 9.068940493468796,
+      "grad_norm": 0.3919062316417694,
+      "learning_rate": 0.00032843778383287914,
+      "loss": 3.2906,
+      "step": 30000
+    },
+    {
+      "epoch": 9.068940493468796,
+      "eval_accuracy": 0.3672455204440074,
+      "eval_loss": 3.5892333984375,
+      "eval_runtime": 182.034,
+      "eval_samples_per_second": 91.433,
+      "eval_steps_per_second": 5.719,
+      "step": 30000
+    },
+    {
+      "epoch": 9.084059022738268,
+      "grad_norm": 0.3697023093700409,
+      "learning_rate": 0.000327983651226158,
+      "loss": 3.2994,
+      "step": 30050
+    },
+    {
+      "epoch": 9.09917755200774,
+      "grad_norm": 0.35424408316612244,
+      "learning_rate": 0.00032752951861943685,
+      "loss": 3.3047,
+      "step": 30100
+    },
+    {
+      "epoch": 9.114296081277214,
+      "grad_norm": 0.3747372329235077,
+      "learning_rate": 0.0003270753860127157,
+      "loss": 3.3119,
+      "step": 30150
+    },
+    {
+      "epoch": 9.129414610546686,
+      "grad_norm": 0.36294251680374146,
+      "learning_rate": 0.0003266212534059945,
+      "loss": 3.3163,
+      "step": 30200
+    },
+    {
+      "epoch": 9.144533139816158,
+      "grad_norm": 0.3802247941493988,
+      "learning_rate": 0.0003261671207992734,
+      "loss": 3.3031,
+      "step": 30250
+    },
+    {
+      "epoch": 9.159651669085632,
+      "grad_norm": 0.3664504885673523,
+      "learning_rate": 0.00032571298819255223,
+      "loss": 3.3157,
+      "step": 30300
+    },
+    {
+      "epoch": 9.174770198355104,
+      "grad_norm": 0.3593531548976898,
+      "learning_rate": 0.000325258855585831,
+      "loss": 3.3118,
+      "step": 30350
+    },
+    {
+      "epoch": 9.189888727624577,
+      "grad_norm": 0.38480937480926514,
+      "learning_rate": 0.00032480472297910984,
+      "loss": 3.3224,
+      "step": 30400
+    },
+    {
+      "epoch": 9.20500725689405,
+      "grad_norm": 0.3553849458694458,
+      "learning_rate": 0.0003243505903723887,
+      "loss": 3.3005,
+      "step": 30450
+    },
+    {
+      "epoch": 9.220125786163521,
+      "grad_norm": 0.3630408048629761,
+      "learning_rate": 0.00032389645776566755,
+      "loss": 3.3213,
+      "step": 30500
+    },
+    {
+      "epoch": 9.235244315432995,
+      "grad_norm": 0.3702227473258972,
+      "learning_rate": 0.0003234423251589464,
+      "loss": 3.3157,
+      "step": 30550
+    },
+    {
+      "epoch": 9.250362844702467,
+      "grad_norm": 0.3907235860824585,
+      "learning_rate": 0.0003229881925522252,
+      "loss": 3.3162,
+      "step": 30600
+    },
+    {
+      "epoch": 9.26548137397194,
+      "grad_norm": 0.35727185010910034,
+      "learning_rate": 0.0003225340599455041,
+      "loss": 3.3156,
+      "step": 30650
+    },
+    {
+      "epoch": 9.280599903241413,
+      "grad_norm": 0.3772910237312317,
+      "learning_rate": 0.0003220799273387829,
+      "loss": 3.3272,
+      "step": 30700
+    },
+    {
+      "epoch": 9.295718432510885,
+      "grad_norm": 0.38522136211395264,
+      "learning_rate": 0.0003216257947320617,
+      "loss": 3.3221,
+      "step": 30750
+    },
+    {
+      "epoch": 9.310836961780359,
+      "grad_norm": 0.37482699751853943,
+      "learning_rate": 0.0003211716621253406,
+      "loss": 3.3235,
+      "step": 30800
+    },
+    {
+      "epoch": 9.32595549104983,
+      "grad_norm": 0.3574056029319763,
+      "learning_rate": 0.0003207175295186194,
+      "loss": 3.3391,
+      "step": 30850
+    },
+    {
+      "epoch": 9.341074020319303,
+      "grad_norm": 0.37357887625694275,
+      "learning_rate": 0.00032026339691189825,
+      "loss": 3.3356,
+      "step": 30900
+    },
+    {
+      "epoch": 9.356192549588776,
+      "grad_norm": 0.3630390763282776,
+      "learning_rate": 0.0003198092643051771,
+      "loss": 3.3334,
+      "step": 30950
+    },
+    {
+      "epoch": 9.371311078858248,
+      "grad_norm": 0.3472421169281006,
+      "learning_rate": 0.00031935513169845597,
+      "loss": 3.3281,
+      "step": 31000
+    },
+    {
+      "epoch": 9.371311078858248,
+      "eval_accuracy": 0.3679117102194405,
+      "eval_loss": 3.584568977355957,
+      "eval_runtime": 182.5198,
+      "eval_samples_per_second": 91.19,
+      "eval_steps_per_second": 5.703,
+      "step": 31000
+    },
+    {
+      "epoch": 9.386429608127722,
+      "grad_norm": 0.3653644919395447,
+      "learning_rate": 0.00031890099909173474,
+      "loss": 3.3286,
+      "step": 31050
+    },
+    {
+      "epoch": 9.401548137397194,
+      "grad_norm": 0.3632447123527527,
+      "learning_rate": 0.0003184468664850136,
+      "loss": 3.3314,
+      "step": 31100
+    },
+    {
+      "epoch": 9.416666666666666,
+      "grad_norm": 0.37565043568611145,
+      "learning_rate": 0.0003179927338782924,
+      "loss": 3.3458,
+      "step": 31150
+    },
+    {
+      "epoch": 9.43178519593614,
+      "grad_norm": 0.3509560525417328,
+      "learning_rate": 0.0003175386012715713,
+      "loss": 3.3354,
+      "step": 31200
+    },
+    {
+      "epoch": 9.446903725205612,
+      "grad_norm": 0.38907599449157715,
+      "learning_rate": 0.0003170844686648501,
+      "loss": 3.3462,
+      "step": 31250
+    },
+    {
+      "epoch": 9.462022254475084,
+      "grad_norm": 0.3524637222290039,
+      "learning_rate": 0.00031663033605812895,
+      "loss": 3.3425,
+      "step": 31300
+    },
+    {
+      "epoch": 9.477140783744558,
+      "grad_norm": 0.3674623370170593,
+      "learning_rate": 0.00031617620345140773,
+      "loss": 3.3356,
+      "step": 31350
+    },
+    {
+      "epoch": 9.49225931301403,
+      "grad_norm": 0.3574652373790741,
+      "learning_rate": 0.0003157220708446866,
+      "loss": 3.3486,
+      "step": 31400
+    },
+    {
+      "epoch": 9.507377842283503,
+      "grad_norm": 0.36924028396606445,
+      "learning_rate": 0.00031526793823796544,
+      "loss": 3.3381,
+      "step": 31450
+    },
+    {
+      "epoch": 9.522496371552975,
+      "grad_norm": 0.3748309910297394,
+      "learning_rate": 0.0003148138056312443,
+      "loss": 3.3494,
+      "step": 31500
+    },
+    {
+      "epoch": 9.537614900822447,
+      "grad_norm": 0.35588720440864563,
+      "learning_rate": 0.00031435967302452316,
+      "loss": 3.3437,
+      "step": 31550
+    },
+    {
+      "epoch": 9.552733430091921,
+      "grad_norm": 0.36605000495910645,
+      "learning_rate": 0.000313905540417802,
+      "loss": 3.3367,
+      "step": 31600
+    },
+    {
+      "epoch": 9.567851959361393,
+      "grad_norm": 0.358410120010376,
+      "learning_rate": 0.0003134514078110808,
+      "loss": 3.3494,
+      "step": 31650
+    },
+    {
+      "epoch": 9.582970488630867,
+      "grad_norm": 0.35448575019836426,
+      "learning_rate": 0.0003129972752043596,
+      "loss": 3.3329,
+      "step": 31700
+    },
+    {
+      "epoch": 9.598089017900339,
+      "grad_norm": 0.36444327235221863,
+      "learning_rate": 0.0003125431425976385,
+      "loss": 3.352,
+      "step": 31750
+    },
+    {
+      "epoch": 9.61320754716981,
+      "grad_norm": 0.3727281093597412,
+      "learning_rate": 0.0003120890099909173,
+      "loss": 3.3421,
+      "step": 31800
+    },
+    {
+      "epoch": 9.628326076439285,
+      "grad_norm": 0.3686982989311218,
+      "learning_rate": 0.00031163487738419614,
+      "loss": 3.3528,
+      "step": 31850
+    },
+    {
+      "epoch": 9.643444605708757,
+      "grad_norm": 0.3870965838432312,
+      "learning_rate": 0.000311180744777475,
+      "loss": 3.3532,
+      "step": 31900
+    },
+    {
+      "epoch": 9.658563134978229,
+      "grad_norm": 0.3674597144126892,
+      "learning_rate": 0.00031072661217075386,
+      "loss": 3.3457,
+      "step": 31950
+    },
+    {
+      "epoch": 9.673681664247702,
+      "grad_norm": 0.3584219813346863,
+      "learning_rate": 0.0003102724795640327,
+      "loss": 3.3594,
+      "step": 32000
+    },
+    {
+      "epoch": 9.673681664247702,
+      "eval_accuracy": 0.36860141534169444,
+      "eval_loss": 3.5766985416412354,
+      "eval_runtime": 182.1545,
+      "eval_samples_per_second": 91.373,
+      "eval_steps_per_second": 5.715,
+      "step": 32000
+    },
+    {
+      "epoch": 9.688800193517174,
+      "grad_norm": 0.37279021739959717,
+      "learning_rate": 0.00030981834695731147,
+      "loss": 3.3629,
+      "step": 32050
+    },
+    {
+      "epoch": 9.703918722786648,
+      "grad_norm": 0.35782763361930847,
+      "learning_rate": 0.00030936421435059035,
+      "loss": 3.3531,
+      "step": 32100
+    },
+    {
+      "epoch": 9.71903725205612,
+      "grad_norm": 0.35830315947532654,
+      "learning_rate": 0.0003089100817438692,
+      "loss": 3.3649,
+      "step": 32150
+    },
+    {
+      "epoch": 9.734155781325592,
+      "grad_norm": 0.3557308614253998,
+      "learning_rate": 0.000308455949137148,
+      "loss": 3.3552,
+      "step": 32200
+    },
+    {
+      "epoch": 9.749274310595066,
+      "grad_norm": 0.36622321605682373,
+      "learning_rate": 0.00030800181653042684,
+      "loss": 3.3478,
+      "step": 32250
+    },
+    {
+      "epoch": 9.764392839864538,
+      "grad_norm": 0.38028189539909363,
+      "learning_rate": 0.00030754768392370573,
+      "loss": 3.3565,
+      "step": 32300
+    },
+    {
+      "epoch": 9.77951136913401,
+      "grad_norm": 0.36501839756965637,
+      "learning_rate": 0.00030709355131698456,
+      "loss": 3.3522,
+      "step": 32350
+    },
+    {
+      "epoch": 9.794629898403484,
+      "grad_norm": 0.3978039026260376,
+      "learning_rate": 0.00030663941871026334,
+      "loss": 3.3553,
+      "step": 32400
+    },
+    {
+      "epoch": 9.809748427672956,
+      "grad_norm": 0.36402684450149536,
+      "learning_rate": 0.00030618528610354217,
+      "loss": 3.3593,
+      "step": 32450
+    },
+    {
+      "epoch": 9.82486695694243,
+      "grad_norm": 0.38092276453971863,
+      "learning_rate": 0.00030573115349682105,
+      "loss": 3.3608,
+      "step": 32500
+    },
+    {
+      "epoch": 9.839985486211901,
+      "grad_norm": 0.3872188329696655,
+      "learning_rate": 0.0003052770208900999,
+      "loss": 3.3612,
+      "step": 32550
+    },
+    {
+      "epoch": 9.855104015481373,
+      "grad_norm": 0.3871103525161743,
+      "learning_rate": 0.0003048228882833787,
+      "loss": 3.353,
+      "step": 32600
+    },
+    {
+      "epoch": 9.870222544750847,
+      "grad_norm": 0.3618019223213196,
+      "learning_rate": 0.00030436875567665754,
+      "loss": 3.3555,
+      "step": 32650
+    },
+    {
+      "epoch": 9.88534107402032,
+      "grad_norm": 0.3572273552417755,
+      "learning_rate": 0.00030391462306993643,
+      "loss": 3.359,
+      "step": 32700
+    },
+    {
+      "epoch": 9.900459603289793,
+      "grad_norm": 0.3581383228302002,
+      "learning_rate": 0.00030346049046321526,
+      "loss": 3.361,
+      "step": 32750
+    },
+    {
+      "epoch": 9.915578132559265,
+      "grad_norm": 0.3652154505252838,
+      "learning_rate": 0.00030300635785649404,
+      "loss": 3.3456,
+      "step": 32800
+    },
+    {
+      "epoch": 9.930696661828737,
+      "grad_norm": 0.3734304904937744,
+      "learning_rate": 0.0003025522252497729,
+      "loss": 3.3664,
+      "step": 32850
+    },
+    {
+      "epoch": 9.94581519109821,
+      "grad_norm": 0.3810529112815857,
+      "learning_rate": 0.00030209809264305175,
+      "loss": 3.3742,
+      "step": 32900
+    },
+    {
+      "epoch": 9.960933720367683,
+      "grad_norm": 0.3573470115661621,
+      "learning_rate": 0.0003016439600363306,
+      "loss": 3.3652,
+      "step": 32950
+    },
+    {
+      "epoch": 9.976052249637155,
+      "grad_norm": 0.3516159951686859,
+      "learning_rate": 0.0003011898274296094,
+      "loss": 3.3589,
+      "step": 33000
+    },
+    {
+      "epoch": 9.976052249637155,
+      "eval_accuracy": 0.36912145723663636,
+      "eval_loss": 3.5680832862854004,
+      "eval_runtime": 182.3106,
+      "eval_samples_per_second": 91.295,
+      "eval_steps_per_second": 5.71,
+      "step": 33000
+    },
+    {
+      "epoch": 9.991170778906628,
+      "grad_norm": 0.35868018865585327,
+      "learning_rate": 0.0003007356948228883,
+      "loss": 3.3603,
+      "step": 33050
+    },
+    {
+      "epoch": 10.006047411707788,
+      "grad_norm": 0.386972576379776,
+      "learning_rate": 0.00030028156221616713,
+      "loss": 3.3157,
+      "step": 33100
+    },
+    {
+      "epoch": 10.021165940977262,
+      "grad_norm": 0.37675729393959045,
+      "learning_rate": 0.0002998274296094459,
+      "loss": 3.2594,
+      "step": 33150
+    },
+    {
+      "epoch": 10.036284470246734,
+      "grad_norm": 0.3801371157169342,
+      "learning_rate": 0.0002993732970027248,
+      "loss": 3.2623,
+      "step": 33200
+    },
+    {
+      "epoch": 10.051402999516206,
+      "grad_norm": 0.38906607031822205,
+      "learning_rate": 0.0002989191643960036,
+      "loss": 3.264,
+      "step": 33250
+    },
+    {
+      "epoch": 10.06652152878568,
+      "grad_norm": 0.38562434911727905,
+      "learning_rate": 0.00029846503178928245,
+      "loss": 3.2688,
+      "step": 33300
+    },
+    {
+      "epoch": 10.081640058055152,
+      "grad_norm": 0.37735775113105774,
+      "learning_rate": 0.0002980108991825613,
+      "loss": 3.2636,
+      "step": 33350
+    },
+    {
+      "epoch": 10.096758587324626,
+      "grad_norm": 0.3712925314903259,
+      "learning_rate": 0.0002975567665758401,
+      "loss": 3.2732,
+      "step": 33400
+    },
+    {
+      "epoch": 10.111877116594098,
+      "grad_norm": 0.37154871225357056,
+      "learning_rate": 0.00029710263396911894,
+      "loss": 3.2768,
+      "step": 33450
+    },
+    {
+      "epoch": 10.12699564586357,
+      "grad_norm": 0.3917189836502075,
+      "learning_rate": 0.0002966485013623978,
+      "loss": 3.2761,
+      "step": 33500
+    },
+    {
+      "epoch": 10.142114175133043,
+      "grad_norm": 0.3951186239719391,
+      "learning_rate": 0.00029619436875567666,
+      "loss": 3.2847,
+      "step": 33550
+    },
+    {
+      "epoch": 10.157232704402515,
+      "grad_norm": 0.3763997256755829,
+      "learning_rate": 0.0002957402361489555,
+      "loss": 3.2954,
+      "step": 33600
+    },
+    {
+      "epoch": 10.17235123367199,
+      "grad_norm": 0.3606552183628082,
+      "learning_rate": 0.0002952861035422343,
+      "loss": 3.2845,
+      "step": 33650
+    },
+    {
+      "epoch": 10.187469762941461,
+      "grad_norm": 0.37124159932136536,
+      "learning_rate": 0.00029483197093551315,
+      "loss": 3.2954,
+      "step": 33700
+    },
+    {
+      "epoch": 10.202588292210933,
+      "grad_norm": 0.3961988389492035,
+      "learning_rate": 0.000294377838328792,
+      "loss": 3.28,
+      "step": 33750
+    },
+    {
+      "epoch": 10.217706821480407,
+      "grad_norm": 0.3912711441516876,
+      "learning_rate": 0.0002939237057220708,
+      "loss": 3.2891,
+      "step": 33800
+    },
+    {
+      "epoch": 10.232825350749879,
+      "grad_norm": 0.3611540198326111,
+      "learning_rate": 0.00029346957311534964,
+      "loss": 3.2862,
+      "step": 33850
+    },
+    {
+      "epoch": 10.247943880019351,
+      "grad_norm": 0.3812921345233917,
+      "learning_rate": 0.0002930154405086285,
+      "loss": 3.2914,
+      "step": 33900
+    },
+    {
+      "epoch": 10.263062409288825,
+      "grad_norm": 0.3808566927909851,
+      "learning_rate": 0.00029256130790190736,
+      "loss": 3.3022,
+      "step": 33950
+    },
+    {
+      "epoch": 10.278180938558297,
+      "grad_norm": 0.3736206889152527,
+      "learning_rate": 0.00029210717529518614,
+      "loss": 3.2886,
+      "step": 34000
+    },
+    {
+      "epoch": 10.278180938558297,
+      "eval_accuracy": 0.36839671424761944,
+      "eval_loss": 3.5810015201568604,
+      "eval_runtime": 182.0653,
+      "eval_samples_per_second": 91.418,
+      "eval_steps_per_second": 5.718,
+      "step": 34000
+    },
+    {
+      "epoch": 10.29329946782777,
+      "grad_norm": 0.36434870958328247,
+      "learning_rate": 0.000291653042688465,
+      "loss": 3.2954,
+      "step": 34050
+    },
+    {
+      "epoch": 10.308417997097242,
+      "grad_norm": 0.38643673062324524,
+      "learning_rate": 0.00029119891008174385,
+      "loss": 3.2943,
+      "step": 34100
+    },
+    {
+      "epoch": 10.323536526366714,
+      "grad_norm": 0.38649046421051025,
+      "learning_rate": 0.0002907447774750227,
+      "loss": 3.3119,
+      "step": 34150
+    },
+    {
+      "epoch": 10.338655055636188,
+      "grad_norm": 0.3665493130683899,
+      "learning_rate": 0.0002902906448683015,
+      "loss": 3.3019,
+      "step": 34200
+    },
+    {
+      "epoch": 10.35377358490566,
+      "grad_norm": 0.3971233069896698,
+      "learning_rate": 0.00028983651226158034,
+      "loss": 3.2944,
+      "step": 34250
+    },
+    {
+      "epoch": 10.368892114175132,
+      "grad_norm": 0.3803875148296356,
+      "learning_rate": 0.00028938237965485923,
+      "loss": 3.3025,
+      "step": 34300
+    },
+    {
+      "epoch": 10.384010643444606,
+      "grad_norm": 0.3761271834373474,
+      "learning_rate": 0.000288928247048138,
+      "loss": 3.2975,
+      "step": 34350
+    },
+    {
+      "epoch": 10.399129172714078,
+      "grad_norm": 0.3758532702922821,
+      "learning_rate": 0.0002884741144414169,
+      "loss": 3.3038,
+      "step": 34400
+    },
+    {
+      "epoch": 10.414247701983552,
+      "grad_norm": 0.3741230368614197,
+      "learning_rate": 0.0002880199818346957,
+      "loss": 3.2986,
+      "step": 34450
+    },
+    {
+      "epoch": 10.429366231253024,
+      "grad_norm": 0.371698260307312,
+      "learning_rate": 0.00028756584922797455,
+      "loss": 3.3093,
+      "step": 34500
+    },
+    {
+      "epoch": 10.444484760522496,
+      "grad_norm": 0.3635827898979187,
+      "learning_rate": 0.0002871117166212534,
+      "loss": 3.3104,
+      "step": 34550
+    },
+    {
+      "epoch": 10.45960328979197,
+      "grad_norm": 0.3689658045768738,
+      "learning_rate": 0.0002866575840145322,
+      "loss": 3.3083,
+      "step": 34600
+    },
+    {
+      "epoch": 10.474721819061442,
+      "grad_norm": 0.37572813034057617,
+      "learning_rate": 0.00028620345140781104,
+      "loss": 3.3092,
+      "step": 34650
+    },
+    {
+      "epoch": 10.489840348330915,
+      "grad_norm": 0.37176594138145447,
+      "learning_rate": 0.0002857493188010899,
+      "loss": 3.3139,
+      "step": 34700
+    },
+    {
+      "epoch": 10.504958877600387,
+      "grad_norm": 0.38343092799186707,
+      "learning_rate": 0.0002852951861943687,
+      "loss": 3.3042,
+      "step": 34750
+    },
+    {
+      "epoch": 10.52007740686986,
+      "grad_norm": 0.3622838258743286,
+      "learning_rate": 0.0002848410535876476,
+      "loss": 3.3099,
+      "step": 34800
+    },
+    {
+      "epoch": 10.535195936139333,
+      "grad_norm": 0.3808203339576721,
+      "learning_rate": 0.00028438692098092637,
+      "loss": 3.3134,
+      "step": 34850
+    },
+    {
+      "epoch": 10.550314465408805,
+      "grad_norm": 0.3763934373855591,
+      "learning_rate": 0.00028393278837420525,
+      "loss": 3.3119,
+      "step": 34900
+    },
+    {
+      "epoch": 10.565432994678277,
+      "grad_norm": 0.39781829714775085,
+      "learning_rate": 0.0002834786557674841,
+      "loss": 3.3192,
+      "step": 34950
+    },
+    {
+      "epoch": 10.58055152394775,
+      "grad_norm": 0.3610644042491913,
+      "learning_rate": 0.0002830245231607629,
+      "loss": 3.3164,
+      "step": 35000
+    },
+    {
+      "epoch": 10.58055152394775,
+      "eval_accuracy": 0.3692526728718964,
+      "eval_loss": 3.5734996795654297,
+      "eval_runtime": 182.1137,
+      "eval_samples_per_second": 91.393,
+      "eval_steps_per_second": 5.716,
+      "step": 35000
+    },
+    {
+      "epoch": 10.595670053217223,
+      "grad_norm": 0.3807249665260315,
+      "learning_rate": 0.00028257039055404174,
+      "loss": 3.3209,
+      "step": 35050
+    },
+    {
+      "epoch": 10.610788582486695,
+      "grad_norm": 0.3662697672843933,
+      "learning_rate": 0.0002821162579473206,
+      "loss": 3.3197,
+      "step": 35100
+    },
+    {
+      "epoch": 10.625907111756169,
+      "grad_norm": 0.3839268386363983,
+      "learning_rate": 0.00028166212534059946,
+      "loss": 3.3253,
+      "step": 35150
+    },
+    {
+      "epoch": 10.64102564102564,
+      "grad_norm": 0.41075611114501953,
+      "learning_rate": 0.00028120799273387824,
+      "loss": 3.3404,
+      "step": 35200
+    },
+    {
+      "epoch": 10.656144170295114,
+      "grad_norm": 0.37889009714126587,
+      "learning_rate": 0.0002807538601271571,
+      "loss": 3.3136,
+      "step": 35250
+    },
+    {
+      "epoch": 10.671262699564586,
+      "grad_norm": 0.37822359800338745,
+      "learning_rate": 0.00028029972752043595,
+      "loss": 3.3198,
+      "step": 35300
+    },
+    {
+      "epoch": 10.686381228834058,
+      "grad_norm": 0.3874446749687195,
+      "learning_rate": 0.0002798455949137148,
+      "loss": 3.3341,
+      "step": 35350
+    },
+    {
+      "epoch": 10.701499758103532,
+      "grad_norm": 0.36894306540489197,
+      "learning_rate": 0.0002793914623069936,
+      "loss": 3.3199,
+      "step": 35400
+    },
+    {
+      "epoch": 10.716618287373004,
+      "grad_norm": 0.377370148897171,
+      "learning_rate": 0.00027893732970027244,
+      "loss": 3.3164,
+      "step": 35450
+    },
+    {
+      "epoch": 10.731736816642478,
+      "grad_norm": 0.4032292366027832,
+      "learning_rate": 0.0002784831970935513,
+      "loss": 3.3203,
+      "step": 35500
+    },
+    {
+      "epoch": 10.74685534591195,
+      "grad_norm": 0.38730841875076294,
+      "learning_rate": 0.00027802906448683016,
+      "loss": 3.3291,
+      "step": 35550
+    },
+    {
+      "epoch": 10.761973875181422,
+      "grad_norm": 0.3680964708328247,
+      "learning_rate": 0.00027757493188010894,
+      "loss": 3.3402,
+      "step": 35600
+    },
+    {
+      "epoch": 10.777092404450896,
+      "grad_norm": 0.3801586925983429,
+      "learning_rate": 0.0002771207992733878,
+      "loss": 3.3207,
+      "step": 35650
+    },
+    {
+      "epoch": 10.792210933720368,
+      "grad_norm": 0.3695020079612732,
+      "learning_rate": 0.00027666666666666665,
+      "loss": 3.3398,
+      "step": 35700
+    },
+    {
+      "epoch": 10.80732946298984,
+      "grad_norm": 0.37439194321632385,
+      "learning_rate": 0.0002762125340599455,
+      "loss": 3.3204,
+      "step": 35750
+    },
+    {
+      "epoch": 10.822447992259313,
+      "grad_norm": 0.38966503739356995,
+      "learning_rate": 0.0002757584014532243,
+      "loss": 3.321,
+      "step": 35800
+    },
+    {
+      "epoch": 10.837566521528785,
+      "grad_norm": 0.3790731430053711,
+      "learning_rate": 0.00027530426884650314,
+      "loss": 3.3157,
+      "step": 35850
+    },
+    {
+      "epoch": 10.852685050798259,
+      "grad_norm": 0.3855128586292267,
+      "learning_rate": 0.00027485013623978203,
+      "loss": 3.3352,
+      "step": 35900
+    },
+    {
+      "epoch": 10.867803580067731,
+      "grad_norm": 0.36516350507736206,
+      "learning_rate": 0.0002743960036330608,
+      "loss": 3.3265,
+      "step": 35950
+    },
+    {
+      "epoch": 10.882922109337203,
+      "grad_norm": 0.3645329177379608,
+      "learning_rate": 0.0002739418710263397,
+      "loss": 3.333,
+      "step": 36000
+    },
+    {
+      "epoch": 10.882922109337203,
+      "eval_accuracy": 0.3697523739918383,
+      "eval_loss": 3.5634424686431885,
+      "eval_runtime": 182.1916,
+      "eval_samples_per_second": 91.354,
+      "eval_steps_per_second": 5.714,
+      "step": 36000
+    },
+    {
+      "epoch": 10.898040638606677,
+      "grad_norm": 0.3843611180782318,
+      "learning_rate": 0.0002734877384196185,
+      "loss": 3.3418,
+      "step": 36050
+    },
+    {
+      "epoch": 10.913159167876149,
+      "grad_norm": 0.3862791061401367,
+      "learning_rate": 0.00027303360581289735,
+      "loss": 3.3377,
+      "step": 36100
+    },
+    {
+      "epoch": 10.92827769714562,
+      "grad_norm": 0.39486852288246155,
+      "learning_rate": 0.0002725794732061762,
+      "loss": 3.3234,
+      "step": 36150
+    },
+    {
+      "epoch": 10.943396226415095,
+      "grad_norm": 0.38032087683677673,
+      "learning_rate": 0.000272125340599455,
+      "loss": 3.3271,
+      "step": 36200
+    },
+    {
+      "epoch": 10.958514755684567,
+      "grad_norm": 0.3659936189651489,
+      "learning_rate": 0.00027167120799273384,
+      "loss": 3.3357,
+      "step": 36250
+    },
+    {
+      "epoch": 10.97363328495404,
+      "grad_norm": 0.37582048773765564,
+      "learning_rate": 0.0002712170753860127,
+      "loss": 3.3281,
+      "step": 36300
+    },
+    {
+      "epoch": 10.988751814223512,
+      "grad_norm": 0.3983675241470337,
+      "learning_rate": 0.0002707629427792915,
+      "loss": 3.3304,
+      "step": 36350
+    },
+    {
+      "epoch": 11.003628447024674,
+      "grad_norm": 0.37397173047065735,
+      "learning_rate": 0.0002703088101725704,
+      "loss": 3.3006,
+      "step": 36400
+    },
+    {
+      "epoch": 11.018746976294146,
+      "grad_norm": 0.3784228265285492,
+      "learning_rate": 0.0002698546775658492,
+      "loss": 3.2259,
+      "step": 36450
+    },
+    {
+      "epoch": 11.033865505563618,
+      "grad_norm": 0.3985145390033722,
+      "learning_rate": 0.00026940054495912805,
+      "loss": 3.2336,
+      "step": 36500
+    },
+    {
+      "epoch": 11.048984034833092,
+      "grad_norm": 0.3934805989265442,
+      "learning_rate": 0.0002689464123524069,
+      "loss": 3.2294,
+      "step": 36550
+    },
+    {
+      "epoch": 11.064102564102564,
+      "grad_norm": 0.3866509199142456,
+      "learning_rate": 0.0002684922797456857,
+      "loss": 3.2465,
+      "step": 36600
+    },
+    {
+      "epoch": 11.079221093372038,
+      "grad_norm": 0.38315460085868835,
+      "learning_rate": 0.00026803814713896454,
+      "loss": 3.2519,
+      "step": 36650
+    },
+    {
+      "epoch": 11.09433962264151,
+      "grad_norm": 0.37440481781959534,
+      "learning_rate": 0.0002675840145322434,
+      "loss": 3.2416,
+      "step": 36700
+    },
+    {
+      "epoch": 11.109458151910982,
+      "grad_norm": 0.3933398425579071,
+      "learning_rate": 0.00026712988192552226,
+      "loss": 3.2557,
+      "step": 36750
+    },
+    {
+      "epoch": 11.124576681180455,
+      "grad_norm": 0.3803589940071106,
+      "learning_rate": 0.00026667574931880104,
+      "loss": 3.252,
+      "step": 36800
+    },
+    {
+      "epoch": 11.139695210449927,
+      "grad_norm": 0.38710469007492065,
+      "learning_rate": 0.0002662216167120799,
+      "loss": 3.2331,
+      "step": 36850
+    },
+    {
+      "epoch": 11.1548137397194,
+      "grad_norm": 0.38195425271987915,
+      "learning_rate": 0.00026576748410535875,
+      "loss": 3.2622,
+      "step": 36900
+    },
+    {
+      "epoch": 11.169932268988873,
+      "grad_norm": 0.3998967409133911,
+      "learning_rate": 0.0002653133514986376,
+      "loss": 3.2556,
+      "step": 36950
+    },
+    {
+      "epoch": 11.185050798258345,
+      "grad_norm": 0.398590624332428,
+      "learning_rate": 0.0002648592188919164,
+      "loss": 3.2511,
+      "step": 37000
+    },
+    {
+      "epoch": 11.185050798258345,
+      "eval_accuracy": 0.3694328004285437,
+      "eval_loss": 3.575971841812134,
+      "eval_runtime": 181.6699,
+      "eval_samples_per_second": 91.617,
+      "eval_steps_per_second": 5.73,
+      "step": 37000
+    },
+    {
+      "epoch": 11.200169327527819,
+      "grad_norm": 0.40366700291633606,
+      "learning_rate": 0.00026440508628519524,
+      "loss": 3.26,
+      "step": 37050
+    },
+    {
+      "epoch": 11.215287856797291,
+      "grad_norm": 0.39497584104537964,
+      "learning_rate": 0.0002639509536784741,
+      "loss": 3.2659,
+      "step": 37100
+    },
+    {
+      "epoch": 11.230406386066763,
+      "grad_norm": 0.39283204078674316,
+      "learning_rate": 0.0002634968210717529,
+      "loss": 3.2634,
+      "step": 37150
+    },
+    {
+      "epoch": 11.245524915336237,
+      "grad_norm": 0.39168092608451843,
+      "learning_rate": 0.0002630426884650318,
+      "loss": 3.2622,
+      "step": 37200
+    },
+    {
+      "epoch": 11.260643444605709,
+      "grad_norm": 0.37125763297080994,
+      "learning_rate": 0.0002625885558583106,
+      "loss": 3.2655,
+      "step": 37250
+    },
+    {
+      "epoch": 11.27576197387518,
+      "grad_norm": 0.3719472289085388,
+      "learning_rate": 0.00026213442325158945,
+      "loss": 3.2694,
+      "step": 37300
+    },
+    {
+      "epoch": 11.290880503144654,
+      "grad_norm": 0.39197173714637756,
+      "learning_rate": 0.0002616802906448683,
+      "loss": 3.2616,
+      "step": 37350
+    },
+    {
+      "epoch": 11.305999032414126,
+      "grad_norm": 0.3874538540840149,
+      "learning_rate": 0.0002612261580381471,
+      "loss": 3.2726,
+      "step": 37400
+    },
+    {
+      "epoch": 11.3211175616836,
+      "grad_norm": 0.39033427834510803,
+      "learning_rate": 0.00026077202543142594,
+      "loss": 3.2733,
+      "step": 37450
+    },
+    {
+      "epoch": 11.336236090953072,
+      "grad_norm": 0.38132426142692566,
+      "learning_rate": 0.0002603178928247048,
+      "loss": 3.2841,
+      "step": 37500
+    },
+    {
+      "epoch": 11.351354620222544,
+      "grad_norm": 0.39073213934898376,
+      "learning_rate": 0.0002598637602179836,
+      "loss": 3.2753,
+      "step": 37550
+    },
+    {
+      "epoch": 11.366473149492018,
+      "grad_norm": 0.4000060558319092,
+      "learning_rate": 0.0002594096276112625,
+      "loss": 3.2635,
+      "step": 37600
+    },
+    {
+      "epoch": 11.38159167876149,
+      "grad_norm": 0.39490681886672974,
+      "learning_rate": 0.00025895549500454127,
+      "loss": 3.2764,
+      "step": 37650
+    },
+    {
+      "epoch": 11.396710208030962,
+      "grad_norm": 0.4136489927768707,
+      "learning_rate": 0.00025850136239782015,
+      "loss": 3.2757,
+      "step": 37700
+    },
+    {
+      "epoch": 11.411828737300436,
+      "grad_norm": 0.3896493911743164,
+      "learning_rate": 0.000258047229791099,
+      "loss": 3.2851,
+      "step": 37750
+    },
+    {
+      "epoch": 11.426947266569908,
+      "grad_norm": 0.3810581862926483,
+      "learning_rate": 0.0002575930971843778,
+      "loss": 3.285,
+      "step": 37800
+    },
+    {
+      "epoch": 11.442065795839381,
+      "grad_norm": 0.38600221276283264,
+      "learning_rate": 0.00025713896457765664,
+      "loss": 3.2936,
+      "step": 37850
+    },
+    {
+      "epoch": 11.457184325108853,
+      "grad_norm": 0.3670991361141205,
+      "learning_rate": 0.0002566848319709355,
+      "loss": 3.2807,
+      "step": 37900
+    },
+    {
+      "epoch": 11.472302854378325,
+      "grad_norm": 0.3892036974430084,
+      "learning_rate": 0.00025623069936421436,
+      "loss": 3.2759,
+      "step": 37950
+    },
+    {
+      "epoch": 11.4874213836478,
+      "grad_norm": 0.4121161997318268,
+      "learning_rate": 0.00025577656675749314,
+      "loss": 3.2867,
+      "step": 38000
+    },
+    {
+      "epoch": 11.4874213836478,
+      "eval_accuracy": 0.3699749467494971,
+      "eval_loss": 3.5709855556488037,
+      "eval_runtime": 181.4502,
+      "eval_samples_per_second": 91.728,
+      "eval_steps_per_second": 5.737,
+      "step": 38000
+    },
+    {
+      "epoch": 11.502539912917271,
+      "grad_norm": 0.3651583790779114,
+      "learning_rate": 0.000255322434150772,
+      "loss": 3.2857,
+      "step": 38050
+    },
+    {
+      "epoch": 11.517658442186743,
+      "grad_norm": 0.39698708057403564,
+      "learning_rate": 0.00025486830154405085,
+      "loss": 3.2755,
+      "step": 38100
+    },
+    {
+      "epoch": 11.532776971456217,
+      "grad_norm": 0.3846209943294525,
+      "learning_rate": 0.0002544141689373297,
+      "loss": 3.2946,
+      "step": 38150
+    },
+    {
+      "epoch": 11.547895500725689,
+      "grad_norm": 0.395131915807724,
+      "learning_rate": 0.0002539600363306085,
+      "loss": 3.2884,
+      "step": 38200
+    },
+    {
+      "epoch": 11.563014029995163,
+      "grad_norm": 0.4034663736820221,
+      "learning_rate": 0.00025350590372388734,
+      "loss": 3.2797,
+      "step": 38250
+    },
+    {
+      "epoch": 11.578132559264635,
+      "grad_norm": 0.3774363398551941,
+      "learning_rate": 0.0002530517711171662,
+      "loss": 3.2811,
+      "step": 38300
+    },
+    {
+      "epoch": 11.593251088534107,
+      "grad_norm": 0.386628121137619,
+      "learning_rate": 0.00025259763851044506,
+      "loss": 3.2884,
+      "step": 38350
+    },
+    {
+      "epoch": 11.60836961780358,
+      "grad_norm": 0.39379242062568665,
+      "learning_rate": 0.00025214350590372384,
+      "loss": 3.2942,
+      "step": 38400
+    },
+    {
+      "epoch": 11.623488147073052,
+      "grad_norm": 0.403767466545105,
+      "learning_rate": 0.0002516893732970027,
+      "loss": 3.2802,
+      "step": 38450
+    },
+    {
+      "epoch": 11.638606676342526,
+      "grad_norm": 0.37925538420677185,
+      "learning_rate": 0.00025123524069028155,
+      "loss": 3.2897,
+      "step": 38500
+    },
+    {
+      "epoch": 11.653725205611998,
+      "grad_norm": 0.38747406005859375,
+      "learning_rate": 0.0002507811080835604,
+      "loss": 3.2969,
+      "step": 38550
+    },
+    {
+      "epoch": 11.66884373488147,
+      "grad_norm": 0.39923471212387085,
+      "learning_rate": 0.0002503269754768392,
+      "loss": 3.2988,
+      "step": 38600
+    },
+    {
+      "epoch": 11.683962264150944,
+      "grad_norm": 0.3835216164588928,
+      "learning_rate": 0.00024987284287011804,
+      "loss": 3.2925,
+      "step": 38650
+    },
+    {
+      "epoch": 11.699080793420416,
+      "grad_norm": 0.37910664081573486,
+      "learning_rate": 0.00024941871026339693,
+      "loss": 3.2974,
+      "step": 38700
+    },
+    {
+      "epoch": 11.714199322689888,
+      "grad_norm": 0.3880641758441925,
+      "learning_rate": 0.0002489645776566757,
+      "loss": 3.2975,
+      "step": 38750
+    },
+    {
+      "epoch": 11.729317851959362,
+      "grad_norm": 0.4259456694126129,
+      "learning_rate": 0.0002485104450499546,
+      "loss": 3.3015,
+      "step": 38800
+    },
+    {
+      "epoch": 11.744436381228834,
+      "grad_norm": 0.387437105178833,
+      "learning_rate": 0.0002480563124432334,
+      "loss": 3.3009,
+      "step": 38850
+    },
+    {
+      "epoch": 11.759554910498307,
+      "grad_norm": 0.39015689492225647,
+      "learning_rate": 0.00024760217983651225,
+      "loss": 3.2999,
+      "step": 38900
+    },
+    {
+      "epoch": 11.77467343976778,
+      "grad_norm": 0.41513824462890625,
+      "learning_rate": 0.0002471480472297911,
+      "loss": 3.2961,
+      "step": 38950
+    },
+    {
+      "epoch": 11.789791969037251,
+      "grad_norm": 0.38026106357574463,
+      "learning_rate": 0.0002466939146230699,
+      "loss": 3.2981,
+      "step": 39000
+    },
+    {
+      "epoch": 11.789791969037251,
+      "eval_accuracy": 0.3703576590190056,
+      "eval_loss": 3.5617599487304688,
+      "eval_runtime": 181.747,
+      "eval_samples_per_second": 91.578,
+      "eval_steps_per_second": 5.728,
+      "step": 39000
+    },
+    {
+      "epoch": 11.804910498306725,
+      "grad_norm": 0.3747030794620514,
+      "learning_rate": 0.00024623978201634874,
+      "loss": 3.3024,
+      "step": 39050
+    },
+    {
+      "epoch": 11.820029027576197,
+      "grad_norm": 0.37710049748420715,
+      "learning_rate": 0.0002457856494096276,
+      "loss": 3.3002,
+      "step": 39100
+    },
+    {
+      "epoch": 11.83514755684567,
+      "grad_norm": 0.3922806978225708,
+      "learning_rate": 0.0002453315168029064,
+      "loss": 3.313,
+      "step": 39150
+    },
+    {
+      "epoch": 11.850266086115143,
+      "grad_norm": 0.4067835509777069,
+      "learning_rate": 0.0002448773841961853,
+      "loss": 3.3049,
+      "step": 39200
+    },
+    {
+      "epoch": 11.865384615384615,
+      "grad_norm": 0.3902236223220825,
+      "learning_rate": 0.00024442325158946407,
+      "loss": 3.2974,
+      "step": 39250
+    },
+    {
+      "epoch": 11.880503144654089,
+      "grad_norm": 0.39979732036590576,
+      "learning_rate": 0.00024396911898274295,
+      "loss": 3.3028,
+      "step": 39300
+    },
+    {
+      "epoch": 11.89562167392356,
+      "grad_norm": 0.3815559446811676,
+      "learning_rate": 0.00024351498637602178,
+      "loss": 3.3034,
+      "step": 39350
+    },
+    {
+      "epoch": 11.910740203193033,
+      "grad_norm": 0.39621683955192566,
+      "learning_rate": 0.0002430608537693006,
+      "loss": 3.3031,
+      "step": 39400
+    },
+    {
+      "epoch": 11.925858732462506,
+      "grad_norm": 0.3927263915538788,
+      "learning_rate": 0.00024260672116257947,
+      "loss": 3.3094,
+      "step": 39450
+    },
+    {
+      "epoch": 11.940977261731978,
+      "grad_norm": 0.386884868144989,
+      "learning_rate": 0.00024215258855585827,
+      "loss": 3.3016,
+      "step": 39500
+    },
+    {
+      "epoch": 11.956095791001452,
+      "grad_norm": 0.3932025134563446,
+      "learning_rate": 0.00024169845594913713,
+      "loss": 3.2988,
+      "step": 39550
+    },
+    {
+      "epoch": 11.971214320270924,
+      "grad_norm": 0.38946929574012756,
+      "learning_rate": 0.00024124432334241596,
+      "loss": 3.2982,
+      "step": 39600
+    },
+    {
+      "epoch": 11.986332849540396,
+      "grad_norm": 0.386614054441452,
+      "learning_rate": 0.00024079019073569482,
+      "loss": 3.3105,
+      "step": 39650
+    },
+    {
+      "epoch": 12.001209482341558,
+      "grad_norm": 0.4196450710296631,
+      "learning_rate": 0.00024033605812897362,
+      "loss": 3.3075,
+      "step": 39700
+    },
+    {
+      "epoch": 12.01632801161103,
+      "grad_norm": 0.38088303804397583,
+      "learning_rate": 0.00023988192552225248,
+      "loss": 3.1953,
+      "step": 39750
+    },
+    {
+      "epoch": 12.031446540880504,
+      "grad_norm": 0.4057537615299225,
+      "learning_rate": 0.0002394277929155313,
+      "loss": 3.2045,
+      "step": 39800
+    },
+    {
+      "epoch": 12.046565070149976,
+      "grad_norm": 0.4013504683971405,
+      "learning_rate": 0.00023897366030881014,
+      "loss": 3.2117,
+      "step": 39850
+    },
+    {
+      "epoch": 12.061683599419448,
+      "grad_norm": 0.4028356075286865,
+      "learning_rate": 0.00023851952770208897,
+      "loss": 3.2161,
+      "step": 39900
+    },
+    {
+      "epoch": 12.076802128688922,
+      "grad_norm": 0.39186498522758484,
+      "learning_rate": 0.00023806539509536783,
+      "loss": 3.2256,
+      "step": 39950
+    },
+    {
+      "epoch": 12.091920657958394,
+      "grad_norm": 0.40833449363708496,
+      "learning_rate": 0.00023761126248864664,
+      "loss": 3.2113,
+      "step": 40000
+    },
+    {
+      "epoch": 12.091920657958394,
+      "eval_accuracy": 0.37015895433836987,
+      "eval_loss": 3.572819471359253,
+      "eval_runtime": 181.5667,
+      "eval_samples_per_second": 91.669,
+      "eval_steps_per_second": 5.733,
+      "step": 40000
+    },
+    {
+      "epoch": 12.107039187227867,
+      "grad_norm": 0.39534837007522583,
+      "learning_rate": 0.0002371571298819255,
+      "loss": 3.2207,
+      "step": 40050
+    },
+    {
+      "epoch": 12.12215771649734,
+      "grad_norm": 0.4102253317832947,
+      "learning_rate": 0.00023670299727520435,
+      "loss": 3.21,
+      "step": 40100
+    },
+    {
+      "epoch": 12.137276245766811,
+      "grad_norm": 0.41309070587158203,
+      "learning_rate": 0.00023624886466848318,
+      "loss": 3.2275,
+      "step": 40150
+    },
+    {
+      "epoch": 12.152394775036285,
+      "grad_norm": 0.3820253908634186,
+      "learning_rate": 0.00023579473206176204,
+      "loss": 3.2241,
+      "step": 40200
+    },
+    {
+      "epoch": 12.167513304305757,
+      "grad_norm": 0.4166340231895447,
+      "learning_rate": 0.00023534059945504084,
+      "loss": 3.2359,
+      "step": 40250
+    },
+    {
+      "epoch": 12.182631833575229,
+      "grad_norm": 0.38069966435432434,
+      "learning_rate": 0.0002348864668483197,
+      "loss": 3.2177,
+      "step": 40300
+    },
+    {
+      "epoch": 12.197750362844703,
+      "grad_norm": 0.4127219617366791,
+      "learning_rate": 0.00023443233424159853,
+      "loss": 3.2377,
+      "step": 40350
+    },
+    {
+      "epoch": 12.212868892114175,
+      "grad_norm": 0.3831850588321686,
+      "learning_rate": 0.00023397820163487736,
+      "loss": 3.2367,
+      "step": 40400
+    },
+    {
+      "epoch": 12.227987421383649,
+      "grad_norm": 0.3869633078575134,
+      "learning_rate": 0.0002335240690281562,
+      "loss": 3.2443,
+      "step": 40450
+    },
+    {
+      "epoch": 12.24310595065312,
+      "grad_norm": 0.4016641676425934,
+      "learning_rate": 0.00023306993642143505,
+      "loss": 3.2473,
+      "step": 40500
+    },
+    {
+      "epoch": 12.258224479922593,
+      "grad_norm": 0.4102655053138733,
+      "learning_rate": 0.00023261580381471385,
+      "loss": 3.2412,
+      "step": 40550
+    },
+    {
+      "epoch": 12.273343009192066,
+      "grad_norm": 0.396590918302536,
+      "learning_rate": 0.0002321616712079927,
+      "loss": 3.2485,
+      "step": 40600
+    },
+    {
+      "epoch": 12.288461538461538,
+      "grad_norm": 0.39611926674842834,
+      "learning_rate": 0.00023170753860127154,
+      "loss": 3.2371,
+      "step": 40650
+    },
+    {
+      "epoch": 12.30358006773101,
+      "grad_norm": 0.3962503969669342,
+      "learning_rate": 0.0002312534059945504,
+      "loss": 3.2564,
+      "step": 40700
+    },
+    {
+      "epoch": 12.318698597000484,
+      "grad_norm": 0.3942914605140686,
+      "learning_rate": 0.00023079927338782923,
+      "loss": 3.2442,
+      "step": 40750
+    },
+    {
+      "epoch": 12.333817126269956,
+      "grad_norm": 0.3998517096042633,
+      "learning_rate": 0.00023034514078110806,
+      "loss": 3.2515,
+      "step": 40800
+    },
+    {
+      "epoch": 12.34893565553943,
+      "grad_norm": 0.3914017081260681,
+      "learning_rate": 0.00022989100817438692,
+      "loss": 3.2522,
+      "step": 40850
+    },
+    {
+      "epoch": 12.364054184808902,
+      "grad_norm": 0.3937617242336273,
+      "learning_rate": 0.00022943687556766572,
+      "loss": 3.2534,
+      "step": 40900
+    },
+    {
+      "epoch": 12.379172714078374,
+      "grad_norm": 0.39187556505203247,
+      "learning_rate": 0.00022898274296094458,
+      "loss": 3.2531,
+      "step": 40950
+    },
+    {
+      "epoch": 12.394291243347848,
+      "grad_norm": 0.40456128120422363,
+      "learning_rate": 0.0002285286103542234,
+      "loss": 3.2584,
+      "step": 41000
+    },
+    {
+      "epoch": 12.394291243347848,
+      "eval_accuracy": 0.370464418693572,
+      "eval_loss": 3.568147659301758,
+      "eval_runtime": 181.5665,
+      "eval_samples_per_second": 91.669,
+      "eval_steps_per_second": 5.733,
+      "step": 41000
+    },
+    {
+      "epoch": 12.40940977261732,
+      "grad_norm": 0.4108814597129822,
+      "learning_rate": 0.00022807447774750227,
+      "loss": 3.2602,
+      "step": 41050
+    },
+    {
+      "epoch": 12.424528301886792,
+      "grad_norm": 0.40934261679649353,
+      "learning_rate": 0.00022762034514078107,
+      "loss": 3.2471,
+      "step": 41100
+    },
+    {
+      "epoch": 12.439646831156265,
+      "grad_norm": 0.4114815592765808,
+      "learning_rate": 0.00022716621253405993,
+      "loss": 3.2563,
+      "step": 41150
+    },
+    {
+      "epoch": 12.454765360425737,
+      "grad_norm": 0.41049736738204956,
+      "learning_rate": 0.00022671207992733876,
+      "loss": 3.2481,
+      "step": 41200
+    },
+    {
+      "epoch": 12.469883889695211,
+      "grad_norm": 0.40271687507629395,
+      "learning_rate": 0.0002262579473206176,
+      "loss": 3.2655,
+      "step": 41250
+    },
+    {
+      "epoch": 12.485002418964683,
+      "grad_norm": 0.39302605390548706,
+      "learning_rate": 0.00022580381471389642,
+      "loss": 3.2559,
+      "step": 41300
+    },
+    {
+      "epoch": 12.500120948234155,
+      "grad_norm": 0.4222753643989563,
+      "learning_rate": 0.00022534968210717528,
+      "loss": 3.2646,
+      "step": 41350
+    },
+    {
+      "epoch": 12.515239477503629,
+      "grad_norm": 0.3878752887248993,
+      "learning_rate": 0.00022489554950045408,
+      "loss": 3.2597,
+      "step": 41400
+    },
+    {
+      "epoch": 12.5303580067731,
+      "grad_norm": 0.3960876762866974,
+      "learning_rate": 0.00022444141689373294,
+      "loss": 3.2597,
+      "step": 41450
+    },
+    {
+      "epoch": 12.545476536042575,
+      "grad_norm": 0.3885483741760254,
+      "learning_rate": 0.0002239872842870118,
+      "loss": 3.2664,
+      "step": 41500
+    },
+    {
+      "epoch": 12.560595065312047,
+      "grad_norm": 0.37746551632881165,
+      "learning_rate": 0.00022353315168029063,
+      "loss": 3.27,
+      "step": 41550
+    },
+    {
+      "epoch": 12.575713594581519,
+      "grad_norm": 0.4066789150238037,
+      "learning_rate": 0.0002230790190735695,
+      "loss": 3.262,
+      "step": 41600
+    },
+    {
+      "epoch": 12.590832123850992,
+      "grad_norm": 0.4039228558540344,
+      "learning_rate": 0.0002226248864668483,
+      "loss": 3.2678,
+      "step": 41650
+    },
+    {
+      "epoch": 12.605950653120464,
+      "grad_norm": 0.3782307803630829,
+      "learning_rate": 0.00022217075386012715,
+      "loss": 3.272,
+      "step": 41700
+    },
+    {
+      "epoch": 12.621069182389936,
+      "grad_norm": 0.40933120250701904,
+      "learning_rate": 0.00022171662125340598,
+      "loss": 3.271,
+      "step": 41750
+    },
+    {
+      "epoch": 12.63618771165941,
+      "grad_norm": 0.422473669052124,
+      "learning_rate": 0.0002212624886466848,
+      "loss": 3.2552,
+      "step": 41800
+    },
+    {
+      "epoch": 12.651306240928882,
+      "grad_norm": 0.39664652943611145,
+      "learning_rate": 0.00022080835603996364,
+      "loss": 3.2603,
+      "step": 41850
+    },
+    {
+      "epoch": 12.666424770198356,
+      "grad_norm": 0.4060044586658478,
+      "learning_rate": 0.0002203542234332425,
+      "loss": 3.268,
+      "step": 41900
+    },
+    {
+      "epoch": 12.681543299467828,
+      "grad_norm": 0.4117172956466675,
+      "learning_rate": 0.0002199000908265213,
+      "loss": 3.2788,
+      "step": 41950
+    },
+    {
+      "epoch": 12.6966618287373,
+      "grad_norm": 0.3968527317047119,
+      "learning_rate": 0.00021944595821980016,
+      "loss": 3.2717,
+      "step": 42000
+    },
+    {
+      "epoch": 12.6966618287373,
+      "eval_accuracy": 0.37127358177767555,
+      "eval_loss": 3.560342788696289,
+      "eval_runtime": 181.731,
+      "eval_samples_per_second": 91.586,
+      "eval_steps_per_second": 5.728,
+      "step": 42000
+    },
+    {
+      "epoch": 12.711780358006774,
+      "grad_norm": 0.3838765025138855,
+      "learning_rate": 0.000218991825613079,
+      "loss": 3.269,
+      "step": 42050
+    },
+    {
+      "epoch": 12.726898887276246,
+      "grad_norm": 0.40154194831848145,
+      "learning_rate": 0.00021853769300635785,
+      "loss": 3.2805,
+      "step": 42100
+    },
+    {
+      "epoch": 12.742017416545718,
+      "grad_norm": 0.41265416145324707,
+      "learning_rate": 0.00021808356039963665,
+      "loss": 3.2806,
+      "step": 42150
+    },
+    {
+      "epoch": 12.757135945815191,
+      "grad_norm": 0.40578603744506836,
+      "learning_rate": 0.0002176294277929155,
+      "loss": 3.2676,
+      "step": 42200
+    },
+    {
+      "epoch": 12.772254475084663,
+      "grad_norm": 0.4076383113861084,
+      "learning_rate": 0.00021717529518619437,
+      "loss": 3.273,
+      "step": 42250
+    },
+    {
+      "epoch": 12.787373004354137,
+      "grad_norm": 0.38173574209213257,
+      "learning_rate": 0.00021672116257947317,
+      "loss": 3.2639,
+      "step": 42300
+    },
+    {
+      "epoch": 12.80249153362361,
+      "grad_norm": 0.4017915427684784,
+      "learning_rate": 0.00021626702997275203,
+      "loss": 3.2709,
+      "step": 42350
+    },
+    {
+      "epoch": 12.817610062893081,
+      "grad_norm": 0.400341659784317,
+      "learning_rate": 0.00021581289736603086,
+      "loss": 3.2604,
+      "step": 42400
+    },
+    {
+      "epoch": 12.832728592162555,
+      "grad_norm": 0.40442341566085815,
+      "learning_rate": 0.00021535876475930972,
+      "loss": 3.2677,
+      "step": 42450
+    },
+    {
+      "epoch": 12.847847121432027,
+      "grad_norm": 0.39479678869247437,
+      "learning_rate": 0.00021490463215258852,
+      "loss": 3.2707,
+      "step": 42500
+    },
+    {
+      "epoch": 12.8629656507015,
+      "grad_norm": 0.3941473364830017,
+      "learning_rate": 0.00021445049954586738,
+      "loss": 3.2603,
+      "step": 42550
+    },
+    {
+      "epoch": 12.878084179970973,
+      "grad_norm": 0.3957080841064453,
+      "learning_rate": 0.0002139963669391462,
+      "loss": 3.2705,
+      "step": 42600
+    },
+    {
+      "epoch": 12.893202709240445,
+      "grad_norm": 0.3981188237667084,
+      "learning_rate": 0.00021354223433242504,
+      "loss": 3.2811,
+      "step": 42650
+    },
+    {
+      "epoch": 12.908321238509918,
+      "grad_norm": 0.40485623478889465,
+      "learning_rate": 0.00021308810172570387,
+      "loss": 3.2726,
+      "step": 42700
+    },
+    {
+      "epoch": 12.92343976777939,
+      "grad_norm": 0.42019230127334595,
+      "learning_rate": 0.00021263396911898273,
+      "loss": 3.273,
+      "step": 42750
+    },
+    {
+      "epoch": 12.938558297048862,
+      "grad_norm": 0.40033140778541565,
+      "learning_rate": 0.00021217983651226153,
+      "loss": 3.2837,
+      "step": 42800
+    },
+    {
+      "epoch": 12.953676826318336,
+      "grad_norm": 0.3891710638999939,
+      "learning_rate": 0.0002117257039055404,
+      "loss": 3.2718,
+      "step": 42850
+    },
+    {
+      "epoch": 12.968795355587808,
+      "grad_norm": 0.38506442308425903,
+      "learning_rate": 0.00021127157129881925,
+      "loss": 3.2775,
+      "step": 42900
+    },
+    {
+      "epoch": 12.98391388485728,
+      "grad_norm": 0.38258466124534607,
+      "learning_rate": 0.00021081743869209808,
+      "loss": 3.2817,
+      "step": 42950
+    },
+    {
+      "epoch": 12.999032414126754,
+      "grad_norm": 0.3776659667491913,
+      "learning_rate": 0.00021036330608537694,
+      "loss": 3.2805,
+      "step": 43000
+    },
+    {
+      "epoch": 12.999032414126754,
+      "eval_accuracy": 0.3715297814812881,
+      "eval_loss": 3.5555002689361572,
+      "eval_runtime": 181.8803,
+      "eval_samples_per_second": 91.511,
+      "eval_steps_per_second": 5.724,
+      "step": 43000
+    },
+    {
+      "epoch": 13.013909046927916,
+      "grad_norm": 0.3910834789276123,
+      "learning_rate": 0.00020990917347865574,
+      "loss": 3.1917,
+      "step": 43050
+    },
+    {
+      "epoch": 13.029027576197388,
+      "grad_norm": 0.3958907425403595,
+      "learning_rate": 0.0002094550408719346,
+      "loss": 3.188,
+      "step": 43100
+    },
+    {
+      "epoch": 13.04414610546686,
+      "grad_norm": 0.3842446506023407,
+      "learning_rate": 0.00020900090826521343,
+      "loss": 3.175,
+      "step": 43150
+    },
+    {
+      "epoch": 13.059264634736333,
+      "grad_norm": 0.40903329849243164,
+      "learning_rate": 0.00020854677565849226,
+      "loss": 3.188,
+      "step": 43200
+    },
+    {
+      "epoch": 13.074383164005805,
+      "grad_norm": 0.42529699206352234,
+      "learning_rate": 0.0002080926430517711,
+      "loss": 3.1999,
+      "step": 43250
+    },
+    {
+      "epoch": 13.089501693275277,
+      "grad_norm": 0.38891202211380005,
+      "learning_rate": 0.00020763851044504995,
+      "loss": 3.1974,
+      "step": 43300
+    },
+    {
+      "epoch": 13.104620222544751,
+      "grad_norm": 0.4020247459411621,
+      "learning_rate": 0.00020718437783832875,
+      "loss": 3.2018,
+      "step": 43350
+    },
+    {
+      "epoch": 13.119738751814223,
+      "grad_norm": 0.4016630947589874,
+      "learning_rate": 0.0002067302452316076,
+      "loss": 3.206,
+      "step": 43400
+    },
+    {
+      "epoch": 13.134857281083697,
+      "grad_norm": 0.4027242958545685,
+      "learning_rate": 0.00020627611262488644,
+      "loss": 3.1896,
+      "step": 43450
+    },
+    {
+      "epoch": 13.149975810353169,
+      "grad_norm": 0.4166494607925415,
+      "learning_rate": 0.0002058219800181653,
+      "loss": 3.215,
+      "step": 43500
+    },
+    {
+      "epoch": 13.165094339622641,
+      "grad_norm": 0.42354124784469604,
+      "learning_rate": 0.0002053678474114441,
+      "loss": 3.2164,
+      "step": 43550
+    },
+    {
+      "epoch": 13.180212868892115,
+      "grad_norm": 0.387747198343277,
+      "learning_rate": 0.00020491371480472296,
+      "loss": 3.2104,
+      "step": 43600
+    },
+    {
+      "epoch": 13.195331398161587,
+      "grad_norm": 0.411479115486145,
+      "learning_rate": 0.00020445958219800182,
+      "loss": 3.2221,
+      "step": 43650
+    },
+    {
+      "epoch": 13.210449927431059,
+      "grad_norm": 0.42029666900634766,
+      "learning_rate": 0.00020400544959128062,
+      "loss": 3.2243,
+      "step": 43700
+    },
+    {
+      "epoch": 13.225568456700532,
+      "grad_norm": 0.42034053802490234,
+      "learning_rate": 0.00020355131698455948,
+      "loss": 3.2211,
+      "step": 43750
+    },
+    {
+      "epoch": 13.240686985970004,
+      "grad_norm": 0.4205058217048645,
+      "learning_rate": 0.0002030971843778383,
+      "loss": 3.2086,
+      "step": 43800
+    },
+    {
+      "epoch": 13.255805515239478,
+      "grad_norm": 0.4254988431930542,
+      "learning_rate": 0.00020264305177111717,
+      "loss": 3.2186,
+      "step": 43850
+    },
+    {
+      "epoch": 13.27092404450895,
+      "grad_norm": 0.43356308341026306,
+      "learning_rate": 0.00020218891916439597,
+      "loss": 3.2186,
+      "step": 43900
+    },
+    {
+      "epoch": 13.286042573778422,
+      "grad_norm": 0.39849865436553955,
+      "learning_rate": 0.00020173478655767483,
+      "loss": 3.2152,
+      "step": 43950
+    },
+    {
+      "epoch": 13.301161103047896,
+      "grad_norm": 0.4155410826206207,
+      "learning_rate": 0.00020128065395095366,
+      "loss": 3.2124,
+      "step": 44000
+    },
+    {
+      "epoch": 13.301161103047896,
+      "eval_accuracy": 0.37124994885412066,
+      "eval_loss": 3.567697048187256,
+      "eval_runtime": 181.5157,
+      "eval_samples_per_second": 91.695,
+      "eval_steps_per_second": 5.735,
+      "step": 44000
+    },
+    {
+      "epoch": 13.316279632317368,
+      "grad_norm": 0.3929089605808258,
+      "learning_rate": 0.0002008265213442325,
+      "loss": 3.2163,
+      "step": 44050
+    },
+    {
+      "epoch": 13.33139816158684,
+      "grad_norm": 0.4119015336036682,
+      "learning_rate": 0.00020037238873751132,
+      "loss": 3.2243,
+      "step": 44100
+    },
+    {
+      "epoch": 13.346516690856314,
+      "grad_norm": 0.3972527086734772,
+      "learning_rate": 0.00019991825613079018,
+      "loss": 3.2297,
+      "step": 44150
+    },
+    {
+      "epoch": 13.361635220125786,
+      "grad_norm": 0.4144071042537689,
+      "learning_rate": 0.00019946412352406898,
+      "loss": 3.2162,
+      "step": 44200
+    },
+    {
+      "epoch": 13.37675374939526,
+      "grad_norm": 0.39286085963249207,
+      "learning_rate": 0.00019900999091734784,
+      "loss": 3.2202,
+      "step": 44250
+    },
+    {
+      "epoch": 13.391872278664732,
+      "grad_norm": 0.39850661158561707,
+      "learning_rate": 0.00019855585831062667,
+      "loss": 3.2285,
+      "step": 44300
+    },
+    {
+      "epoch": 13.406990807934204,
+      "grad_norm": 0.3939213454723358,
+      "learning_rate": 0.00019810172570390553,
+      "loss": 3.2397,
+      "step": 44350
+    },
+    {
+      "epoch": 13.422109337203677,
+      "grad_norm": 0.4019733965396881,
+      "learning_rate": 0.0001976475930971844,
+      "loss": 3.2131,
+      "step": 44400
+    },
+    {
+      "epoch": 13.43722786647315,
+      "grad_norm": 0.40081989765167236,
+      "learning_rate": 0.0001971934604904632,
+      "loss": 3.2349,
+      "step": 44450
+    },
+    {
+      "epoch": 13.452346395742623,
+      "grad_norm": 0.41404154896736145,
+      "learning_rate": 0.00019673932788374205,
+      "loss": 3.2303,
+      "step": 44500
+    },
+    {
+      "epoch": 13.467464925012095,
+      "grad_norm": 0.4119343161582947,
+      "learning_rate": 0.00019628519527702088,
+      "loss": 3.2266,
+      "step": 44550
+    },
+    {
+      "epoch": 13.482583454281567,
+      "grad_norm": 0.41343453526496887,
+      "learning_rate": 0.0001958310626702997,
+      "loss": 3.2294,
+      "step": 44600
+    },
+    {
+      "epoch": 13.49770198355104,
+      "grad_norm": 0.3903872072696686,
+      "learning_rate": 0.00019537693006357854,
+      "loss": 3.223,
+      "step": 44650
+    },
+    {
+      "epoch": 13.512820512820513,
+      "grad_norm": 0.41756418347358704,
+      "learning_rate": 0.0001949227974568574,
+      "loss": 3.2336,
+      "step": 44700
+    },
+    {
+      "epoch": 13.527939042089985,
+      "grad_norm": 0.40313833951950073,
+      "learning_rate": 0.0001944686648501362,
+      "loss": 3.2332,
+      "step": 44750
+    },
+    {
+      "epoch": 13.543057571359459,
+      "grad_norm": 0.40880855917930603,
+      "learning_rate": 0.00019401453224341506,
+      "loss": 3.237,
+      "step": 44800
+    },
+    {
+      "epoch": 13.55817610062893,
+      "grad_norm": 0.41109326481819153,
+      "learning_rate": 0.0001935603996366939,
+      "loss": 3.2484,
+      "step": 44850
+    },
+    {
+      "epoch": 13.573294629898404,
+      "grad_norm": 0.4063083231449127,
+      "learning_rate": 0.00019310626702997275,
+      "loss": 3.2367,
+      "step": 44900
+    },
+    {
+      "epoch": 13.588413159167876,
+      "grad_norm": 0.39859169721603394,
+      "learning_rate": 0.00019265213442325155,
+      "loss": 3.2337,
+      "step": 44950
+    },
+    {
+      "epoch": 13.603531688437348,
+      "grad_norm": 0.39923831820487976,
+      "learning_rate": 0.0001921980018165304,
+      "loss": 3.2352,
+      "step": 45000
+    },
+    {
+      "epoch": 13.603531688437348,
+      "eval_accuracy": 0.371590686229554,
+      "eval_loss": 3.5596277713775635,
+      "eval_runtime": 182.4311,
+      "eval_samples_per_second": 91.234,
+      "eval_steps_per_second": 5.706,
+      "step": 45000
+    },
+    {
+      "epoch": 13.618650217706822,
+      "grad_norm": 0.40481671690940857,
+      "learning_rate": 0.00019174386920980924,
+      "loss": 3.2446,
+      "step": 45050
+    },
+    {
+      "epoch": 13.633768746976294,
+      "grad_norm": 0.4066201448440552,
+      "learning_rate": 0.00019128973660308807,
+      "loss": 3.2385,
+      "step": 45100
+    },
+    {
+      "epoch": 13.648887276245766,
+      "grad_norm": 0.411175012588501,
+      "learning_rate": 0.00019083560399636693,
+      "loss": 3.2362,
+      "step": 45150
+    },
+    {
+      "epoch": 13.66400580551524,
+      "grad_norm": 0.4065161645412445,
+      "learning_rate": 0.00019038147138964576,
+      "loss": 3.2415,
+      "step": 45200
+    },
+    {
+      "epoch": 13.679124334784712,
+      "grad_norm": 0.41124144196510315,
+      "learning_rate": 0.00018992733878292462,
+      "loss": 3.2396,
+      "step": 45250
+    },
+    {
+      "epoch": 13.694242864054186,
+      "grad_norm": 0.4290773272514343,
+      "learning_rate": 0.00018947320617620342,
+      "loss": 3.2601,
+      "step": 45300
+    },
+    {
+      "epoch": 13.709361393323658,
+      "grad_norm": 0.4111163020133972,
+      "learning_rate": 0.00018901907356948228,
+      "loss": 3.241,
+      "step": 45350
+    },
+    {
+      "epoch": 13.72447992259313,
+      "grad_norm": 0.40063321590423584,
+      "learning_rate": 0.0001885649409627611,
+      "loss": 3.2305,
+      "step": 45400
+    },
+    {
+      "epoch": 13.739598451862603,
+      "grad_norm": 0.38352343440055847,
+      "learning_rate": 0.00018811080835603994,
+      "loss": 3.2383,
+      "step": 45450
+    },
+    {
+      "epoch": 13.754716981132075,
+      "grad_norm": 0.43132641911506653,
+      "learning_rate": 0.00018765667574931877,
+      "loss": 3.2537,
+      "step": 45500
+    },
+    {
+      "epoch": 13.769835510401549,
+      "grad_norm": 0.409151166677475,
+      "learning_rate": 0.00018720254314259763,
+      "loss": 3.2405,
+      "step": 45550
+    },
+    {
+      "epoch": 13.784954039671021,
+      "grad_norm": 0.4174506366252899,
+      "learning_rate": 0.00018674841053587643,
+      "loss": 3.2594,
+      "step": 45600
+    },
+    {
+      "epoch": 13.800072568940493,
+      "grad_norm": 0.40947169065475464,
+      "learning_rate": 0.0001862942779291553,
+      "loss": 3.248,
+      "step": 45650
+    },
+    {
+      "epoch": 13.815191098209967,
+      "grad_norm": 0.4033055603504181,
+      "learning_rate": 0.00018584014532243412,
+      "loss": 3.2331,
+      "step": 45700
+    },
+    {
+      "epoch": 13.830309627479439,
+      "grad_norm": 0.4000418484210968,
+      "learning_rate": 0.00018538601271571298,
+      "loss": 3.249,
+      "step": 45750
+    },
+    {
+      "epoch": 13.84542815674891,
+      "grad_norm": 0.4021860957145691,
+      "learning_rate": 0.00018493188010899184,
+      "loss": 3.2478,
+      "step": 45800
+    },
+    {
+      "epoch": 13.860546686018385,
+      "grad_norm": 0.3957579731941223,
+      "learning_rate": 0.00018447774750227064,
+      "loss": 3.2536,
+      "step": 45850
+    },
+    {
+      "epoch": 13.875665215287857,
+      "grad_norm": 0.43215376138687134,
+      "learning_rate": 0.0001840236148955495,
+      "loss": 3.247,
+      "step": 45900
+    },
+    {
+      "epoch": 13.890783744557329,
+      "grad_norm": 0.4118666350841522,
+      "learning_rate": 0.00018356948228882833,
+      "loss": 3.2431,
+      "step": 45950
+    },
+    {
+      "epoch": 13.905902273826802,
+      "grad_norm": 0.4115840494632721,
+      "learning_rate": 0.00018311534968210716,
+      "loss": 3.2477,
+      "step": 46000
+    },
+    {
+      "epoch": 13.905902273826802,
+      "eval_accuracy": 0.3719852737492069,
+      "eval_loss": 3.5575146675109863,
+      "eval_runtime": 182.5723,
+      "eval_samples_per_second": 91.164,
+      "eval_steps_per_second": 5.702,
+      "step": 46000
+    },
+    {
+      "epoch": 13.921020803096274,
+      "grad_norm": 0.4212381839752197,
+      "learning_rate": 0.000182661217075386,
+      "loss": 3.2476,
+      "step": 46050
+    },
+    {
+      "epoch": 13.936139332365748,
+      "grad_norm": 0.4087257385253906,
+      "learning_rate": 0.00018220708446866485,
+      "loss": 3.2438,
+      "step": 46100
+    },
+    {
+      "epoch": 13.95125786163522,
+      "grad_norm": 0.39609295129776,
+      "learning_rate": 0.00018175295186194365,
+      "loss": 3.252,
+      "step": 46150
+    },
+    {
+      "epoch": 13.966376390904692,
+      "grad_norm": 0.38680288195610046,
+      "learning_rate": 0.0001812988192552225,
+      "loss": 3.2479,
+      "step": 46200
+    },
+    {
+      "epoch": 13.981494920174166,
+      "grad_norm": 0.39530307054519653,
+      "learning_rate": 0.00018084468664850134,
+      "loss": 3.2696,
+      "step": 46250
+    },
+    {
+      "epoch": 13.996613449443638,
+      "grad_norm": 0.40767961740493774,
+      "learning_rate": 0.0001803905540417802,
+      "loss": 3.2543,
+      "step": 46300
+    },
+    {
+      "epoch": 14.0114900822448,
+      "grad_norm": 0.4131290018558502,
+      "learning_rate": 0.000179936421435059,
+      "loss": 3.1936,
+      "step": 46350
+    },
+    {
+      "epoch": 14.026608611514272,
+      "grad_norm": 0.40820956230163574,
+      "learning_rate": 0.00017948228882833786,
+      "loss": 3.1675,
+      "step": 46400
+    },
+    {
+      "epoch": 14.041727140783745,
+      "grad_norm": 0.4055199921131134,
+      "learning_rate": 0.0001790281562216167,
+      "loss": 3.1683,
+      "step": 46450
+    },
+    {
+      "epoch": 14.056845670053217,
+      "grad_norm": 0.41949304938316345,
+      "learning_rate": 0.00017857402361489552,
+      "loss": 3.1716,
+      "step": 46500
+    },
+    {
+      "epoch": 14.07196419932269,
+      "grad_norm": 0.4142007529735565,
+      "learning_rate": 0.00017811989100817438,
+      "loss": 3.1763,
+      "step": 46550
+    },
+    {
+      "epoch": 14.087082728592163,
+      "grad_norm": 0.42433109879493713,
+      "learning_rate": 0.0001776657584014532,
+      "loss": 3.1737,
+      "step": 46600
+    },
+    {
+      "epoch": 14.102201257861635,
+      "grad_norm": 0.4084174931049347,
+      "learning_rate": 0.00017721162579473207,
+      "loss": 3.1795,
+      "step": 46650
+    },
+    {
+      "epoch": 14.117319787131107,
+      "grad_norm": 0.42154616117477417,
+      "learning_rate": 0.00017675749318801087,
+      "loss": 3.1796,
+      "step": 46700
+    },
+    {
+      "epoch": 14.132438316400581,
+      "grad_norm": 0.42639973759651184,
+      "learning_rate": 0.00017630336058128973,
+      "loss": 3.186,
+      "step": 46750
+    },
+    {
+      "epoch": 14.147556845670053,
+      "grad_norm": 0.40734776854515076,
+      "learning_rate": 0.00017584922797456856,
+      "loss": 3.1896,
+      "step": 46800
+    },
+    {
+      "epoch": 14.162675374939527,
+      "grad_norm": 0.44012895226478577,
+      "learning_rate": 0.0001753950953678474,
+      "loss": 3.1707,
+      "step": 46850
+    },
+    {
+      "epoch": 14.177793904208999,
+      "grad_norm": 0.42036205530166626,
+      "learning_rate": 0.00017494096276112622,
+      "loss": 3.1837,
+      "step": 46900
+    },
+    {
+      "epoch": 14.19291243347847,
+      "grad_norm": 0.4042738676071167,
+      "learning_rate": 0.00017448683015440508,
+      "loss": 3.1915,
+      "step": 46950
+    },
+    {
+      "epoch": 14.208030962747944,
+      "grad_norm": 0.41336682438850403,
+      "learning_rate": 0.00017403269754768388,
+      "loss": 3.1954,
+      "step": 47000
+    },
+    {
+      "epoch": 14.208030962747944,
+      "eval_accuracy": 0.37157375517984303,
+      "eval_loss": 3.564302682876587,
+      "eval_runtime": 182.5131,
+      "eval_samples_per_second": 91.193,
+      "eval_steps_per_second": 5.704,
+      "step": 47000
+    },
+    {
+      "epoch": 14.223149492017416,
+      "grad_norm": 0.46255195140838623,
+      "learning_rate": 0.00017357856494096274,
+      "loss": 3.1851,
+      "step": 47050
+    },
+    {
+      "epoch": 14.238268021286888,
+      "grad_norm": 0.4109150469303131,
+      "learning_rate": 0.00017312443233424157,
+      "loss": 3.1951,
+      "step": 47100
+    },
+    {
+      "epoch": 14.253386550556362,
+      "grad_norm": 0.4143703877925873,
+      "learning_rate": 0.00017267029972752043,
+      "loss": 3.2023,
+      "step": 47150
+    },
+    {
+      "epoch": 14.268505079825834,
+      "grad_norm": 0.41615933179855347,
+      "learning_rate": 0.00017221616712079923,
+      "loss": 3.196,
+      "step": 47200
+    },
+    {
+      "epoch": 14.283623609095308,
+      "grad_norm": 0.43715471029281616,
+      "learning_rate": 0.0001717620345140781,
+      "loss": 3.2061,
+      "step": 47250
+    },
+    {
+      "epoch": 14.29874213836478,
+      "grad_norm": 0.4121893048286438,
+      "learning_rate": 0.00017130790190735695,
+      "loss": 3.2027,
+      "step": 47300
+    },
+    {
+      "epoch": 14.313860667634252,
+      "grad_norm": 0.424890398979187,
+      "learning_rate": 0.00017085376930063578,
+      "loss": 3.1913,
+      "step": 47350
+    },
+    {
+      "epoch": 14.328979196903726,
+      "grad_norm": 0.4319377839565277,
+      "learning_rate": 0.0001703996366939146,
+      "loss": 3.2066,
+      "step": 47400
+    },
+    {
+      "epoch": 14.344097726173198,
+      "grad_norm": 0.4028719663619995,
+      "learning_rate": 0.00016994550408719344,
+      "loss": 3.1982,
+      "step": 47450
+    },
+    {
+      "epoch": 14.359216255442671,
+      "grad_norm": 0.4185102880001068,
+      "learning_rate": 0.0001694913714804723,
+      "loss": 3.1978,
+      "step": 47500
+    },
+    {
+      "epoch": 14.374334784712143,
+      "grad_norm": 0.4216921627521515,
+      "learning_rate": 0.0001690372388737511,
+      "loss": 3.1973,
+      "step": 47550
+    },
+    {
+      "epoch": 14.389453313981615,
+      "grad_norm": 0.40805599093437195,
+      "learning_rate": 0.00016858310626702996,
+      "loss": 3.1978,
+      "step": 47600
+    },
+    {
+      "epoch": 14.40457184325109,
+      "grad_norm": 0.42335882782936096,
+      "learning_rate": 0.0001681289736603088,
+      "loss": 3.199,
+      "step": 47650
+    },
+    {
+      "epoch": 14.419690372520561,
+      "grad_norm": 0.4133867621421814,
+      "learning_rate": 0.00016767484105358765,
+      "loss": 3.2183,
+      "step": 47700
+    },
+    {
+      "epoch": 14.434808901790033,
+      "grad_norm": 0.41599327325820923,
+      "learning_rate": 0.00016722070844686645,
+      "loss": 3.1966,
+      "step": 47750
+    },
+    {
+      "epoch": 14.449927431059507,
+      "grad_norm": 0.41563186049461365,
+      "learning_rate": 0.0001667665758401453,
+      "loss": 3.1937,
+      "step": 47800
+    },
+    {
+      "epoch": 14.465045960328979,
+      "grad_norm": 0.39862963557243347,
+      "learning_rate": 0.00016631244323342414,
+      "loss": 3.1962,
+      "step": 47850
+    },
+    {
+      "epoch": 14.480164489598453,
+      "grad_norm": 0.4591251015663147,
+      "learning_rate": 0.00016585831062670297,
+      "loss": 3.1961,
+      "step": 47900
+    },
+    {
+      "epoch": 14.495283018867925,
+      "grad_norm": 0.4092547595500946,
+      "learning_rate": 0.00016540417801998183,
+      "loss": 3.214,
+      "step": 47950
+    },
+    {
+      "epoch": 14.510401548137397,
+      "grad_norm": 0.4241653084754944,
+      "learning_rate": 0.00016495004541326066,
+      "loss": 3.2191,
+      "step": 48000
+    },
+    {
+      "epoch": 14.510401548137397,
+      "eval_accuracy": 0.37222407209617214,
+      "eval_loss": 3.557905673980713,
+      "eval_runtime": 182.5474,
+      "eval_samples_per_second": 91.176,
+      "eval_steps_per_second": 5.703,
+      "step": 48000
+    },
+    {
+      "epoch": 14.52552007740687,
+      "grad_norm": 0.4303036034107208,
+      "learning_rate": 0.00016449591280653952,
+      "loss": 3.2149,
+      "step": 48050
+    },
+    {
+      "epoch": 14.540638606676342,
+      "grad_norm": 0.4225316047668457,
+      "learning_rate": 0.00016404178019981832,
+      "loss": 3.2069,
+      "step": 48100
+    },
+    {
+      "epoch": 14.555757135945814,
+      "grad_norm": 0.44241517782211304,
+      "learning_rate": 0.00016358764759309718,
+      "loss": 3.2143,
+      "step": 48150
+    },
+    {
+      "epoch": 14.570875665215288,
+      "grad_norm": 0.44027650356292725,
+      "learning_rate": 0.000163133514986376,
+      "loss": 3.2134,
+      "step": 48200
+    },
+    {
+      "epoch": 14.58599419448476,
+      "grad_norm": 0.40961533784866333,
+      "learning_rate": 0.00016267938237965484,
+      "loss": 3.22,
+      "step": 48250
+    },
+    {
+      "epoch": 14.601112723754234,
+      "grad_norm": 0.4150550663471222,
+      "learning_rate": 0.00016222524977293367,
+      "loss": 3.2215,
+      "step": 48300
+    },
+    {
+      "epoch": 14.616231253023706,
+      "grad_norm": 0.41550400853157043,
+      "learning_rate": 0.00016177111716621253,
+      "loss": 3.2093,
+      "step": 48350
+    },
+    {
+      "epoch": 14.631349782293178,
+      "grad_norm": 0.4034069776535034,
+      "learning_rate": 0.00016131698455949133,
+      "loss": 3.2205,
+      "step": 48400
+    },
+    {
+      "epoch": 14.646468311562652,
+      "grad_norm": 0.43780672550201416,
+      "learning_rate": 0.0001608628519527702,
+      "loss": 3.2103,
+      "step": 48450
+    },
+    {
+      "epoch": 14.661586840832124,
+      "grad_norm": 0.41662082076072693,
+      "learning_rate": 0.00016040871934604902,
+      "loss": 3.2207,
+      "step": 48500
+    },
+    {
+      "epoch": 14.676705370101596,
+      "grad_norm": 0.41818660497665405,
+      "learning_rate": 0.00015995458673932788,
+      "loss": 3.219,
+      "step": 48550
+    },
+    {
+      "epoch": 14.69182389937107,
+      "grad_norm": 0.4019031524658203,
+      "learning_rate": 0.00015950045413260668,
+      "loss": 3.2011,
+      "step": 48600
+    },
+    {
+      "epoch": 14.706942428640541,
+      "grad_norm": 0.4318902790546417,
+      "learning_rate": 0.00015904632152588554,
+      "loss": 3.2212,
+      "step": 48650
+    },
+    {
+      "epoch": 14.722060957910015,
+      "grad_norm": 0.4129679203033447,
+      "learning_rate": 0.0001585921889191644,
+      "loss": 3.2094,
+      "step": 48700
+    },
+    {
+      "epoch": 14.737179487179487,
+      "grad_norm": 0.4268816411495209,
+      "learning_rate": 0.00015813805631244323,
+      "loss": 3.2194,
+      "step": 48750
+    },
+    {
+      "epoch": 14.75229801644896,
+      "grad_norm": 0.43164098262786865,
+      "learning_rate": 0.00015768392370572206,
+      "loss": 3.2072,
+      "step": 48800
+    },
+    {
+      "epoch": 14.767416545718433,
+      "grad_norm": 0.4305291771888733,
+      "learning_rate": 0.0001572297910990009,
+      "loss": 3.2239,
+      "step": 48850
+    },
+    {
+      "epoch": 14.782535074987905,
+      "grad_norm": 0.4099324941635132,
+      "learning_rate": 0.00015677565849227975,
+      "loss": 3.2257,
+      "step": 48900
+    },
+    {
+      "epoch": 14.797653604257377,
+      "grad_norm": 0.431211918592453,
+      "learning_rate": 0.00015632152588555855,
+      "loss": 3.2173,
+      "step": 48950
+    },
+    {
+      "epoch": 14.81277213352685,
+      "grad_norm": 0.43717148900032043,
+      "learning_rate": 0.0001558673932788374,
+      "loss": 3.2221,
+      "step": 49000
+    },
+    {
+      "epoch": 14.81277213352685,
+      "eval_accuracy": 0.37258820724169217,
+      "eval_loss": 3.5516107082366943,
+      "eval_runtime": 182.6239,
+      "eval_samples_per_second": 91.138,
+      "eval_steps_per_second": 5.7,
+      "step": 49000
+    },
+    {
+      "epoch": 14.827890662796323,
+      "grad_norm": 0.4287441372871399,
+      "learning_rate": 0.00015541326067211624,
+      "loss": 3.2187,
+      "step": 49050
+    },
+    {
+      "epoch": 14.843009192065796,
+      "grad_norm": 0.39929869771003723,
+      "learning_rate": 0.0001549591280653951,
+      "loss": 3.2176,
+      "step": 49100
+    },
+    {
+      "epoch": 14.858127721335268,
+      "grad_norm": 0.43470391631126404,
+      "learning_rate": 0.0001545049954586739,
+      "loss": 3.2304,
+      "step": 49150
+    },
+    {
+      "epoch": 14.87324625060474,
+      "grad_norm": 0.4141775667667389,
+      "learning_rate": 0.00015405086285195276,
+      "loss": 3.2258,
+      "step": 49200
+    },
+    {
+      "epoch": 14.888364779874214,
+      "grad_norm": 0.4020072817802429,
+      "learning_rate": 0.0001535967302452316,
+      "loss": 3.2275,
+      "step": 49250
+    },
+    {
+      "epoch": 14.903483309143686,
+      "grad_norm": 0.42552366852760315,
+      "learning_rate": 0.00015314259763851042,
+      "loss": 3.2308,
+      "step": 49300
+    },
+    {
+      "epoch": 14.91860183841316,
+      "grad_norm": 0.41669708490371704,
+      "learning_rate": 0.00015268846503178925,
+      "loss": 3.219,
+      "step": 49350
+    },
+    {
+      "epoch": 14.933720367682632,
+      "grad_norm": 0.43189725279808044,
+      "learning_rate": 0.0001522343324250681,
+      "loss": 3.2263,
+      "step": 49400
+    },
+    {
+      "epoch": 14.948838896952104,
+      "grad_norm": 0.4100758135318756,
+      "learning_rate": 0.00015178019981834697,
+      "loss": 3.2267,
+      "step": 49450
+    },
+    {
+      "epoch": 14.963957426221578,
+      "grad_norm": 0.41336727142333984,
+      "learning_rate": 0.00015132606721162577,
+      "loss": 3.2247,
+      "step": 49500
+    },
+    {
+      "epoch": 14.97907595549105,
+      "grad_norm": 0.4141419529914856,
+      "learning_rate": 0.00015087193460490463,
+      "loss": 3.213,
+      "step": 49550
+    },
+    {
+      "epoch": 14.994194484760522,
+      "grad_norm": 0.4345361888408661,
+      "learning_rate": 0.00015041780199818346,
+      "loss": 3.2258,
+      "step": 49600
+    },
+    {
+      "epoch": 15.009071117561684,
+      "grad_norm": 0.40050098299980164,
+      "learning_rate": 0.0001499636693914623,
+      "loss": 3.182,
+      "step": 49650
+    },
+    {
+      "epoch": 15.024189646831156,
+      "grad_norm": 0.41687700152397156,
+      "learning_rate": 0.00014950953678474112,
+      "loss": 3.1643,
+      "step": 49700
+    },
+    {
+      "epoch": 15.03930817610063,
+      "grad_norm": 0.43859630823135376,
+      "learning_rate": 0.00014905540417801998,
+      "loss": 3.1614,
+      "step": 49750
+    },
+    {
+      "epoch": 15.054426705370101,
+      "grad_norm": 0.4150136709213257,
+      "learning_rate": 0.0001486012715712988,
+      "loss": 3.1513,
+      "step": 49800
+    },
+    {
+      "epoch": 15.069545234639575,
+      "grad_norm": 0.41403576731681824,
+      "learning_rate": 0.00014814713896457764,
+      "loss": 3.1426,
+      "step": 49850
+    },
+    {
+      "epoch": 15.084663763909047,
+      "grad_norm": 0.4175975024700165,
+      "learning_rate": 0.0001476930063578565,
+      "loss": 3.1457,
+      "step": 49900
+    },
+    {
+      "epoch": 15.099782293178519,
+      "grad_norm": 0.4216543734073639,
+      "learning_rate": 0.00014723887375113533,
+      "loss": 3.1489,
+      "step": 49950
+    },
+    {
+      "epoch": 15.114900822447993,
+      "grad_norm": 0.4189589321613312,
+      "learning_rate": 0.00014678474114441416,
+      "loss": 3.1676,
+      "step": 50000
+    },
+    {
+      "epoch": 15.114900822447993,
+      "eval_accuracy": 0.3722348891557097,
+      "eval_loss": 3.562861442565918,
+      "eval_runtime": 182.4146,
+      "eval_samples_per_second": 91.243,
+      "eval_steps_per_second": 5.707,
+      "step": 50000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 66160,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "EarlyStoppingCallback": {
+      "args": {
+        "early_stopping_patience": 20,
+        "early_stopping_threshold": 0.0
+      },
+      "attributes": {
+        "early_stopping_patience_counter": 1
+      }
+    },
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.04491728764928e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}