diff --git "a/checkpoint-50000/trainer_state.json" "b/checkpoint-50000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-50000/trainer_state.json"
@@ -0,0 +1,7483 @@
+{
+  "best_metric": 3.403038263320923,
+  "best_model_checkpoint": "/scratch/cl5625/exceptions/models/100M__634/checkpoint-50000",
+  "epoch": 5.390835579514825,
+  "eval_steps": 1000,
+  "global_step": 50000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.005390835579514825,
+      "grad_norm": 1.412644624710083,
+      "learning_rate": 0.0003,
+      "loss": 8.6259,
+      "step": 50
+    },
+    {
+      "epoch": 0.01078167115902965,
+      "grad_norm": 3.4297802448272705,
+      "learning_rate": 0.0006,
+      "loss": 6.925,
+      "step": 100
+    },
+    {
+      "epoch": 0.016172506738544475,
+      "grad_norm": 1.0797241926193237,
+      "learning_rate": 0.0005996762007555315,
+      "loss": 6.4904,
+      "step": 150
+    },
+    {
+      "epoch": 0.0215633423180593,
+      "grad_norm": 1.8681477308273315,
+      "learning_rate": 0.000599352401511063,
+      "loss": 6.2514,
+      "step": 200
+    },
+    {
+      "epoch": 0.026954177897574125,
+      "grad_norm": 1.0239183902740479,
+      "learning_rate": 0.0005990286022665946,
+      "loss": 6.0911,
+      "step": 250
+    },
+    {
+      "epoch": 0.03234501347708895,
+      "grad_norm": 1.5482831001281738,
+      "learning_rate": 0.0005987048030221263,
+      "loss": 5.9845,
+      "step": 300
+    },
+    {
+      "epoch": 0.03773584905660377,
+      "grad_norm": 1.6334092617034912,
+      "learning_rate": 0.0005983810037776578,
+      "loss": 5.875,
+      "step": 350
+    },
+    {
+      "epoch": 0.0431266846361186,
+      "grad_norm": 1.5844093561172485,
+      "learning_rate": 0.0005980572045331894,
+      "loss": 5.7818,
+      "step": 400
+    },
+    {
+      "epoch": 0.04851752021563342,
+      "grad_norm": 1.2689578533172607,
+      "learning_rate": 0.0005977334052887209,
+      "loss": 5.712,
+      "step": 450
+    },
+    {
+      "epoch": 0.05390835579514825,
+      "grad_norm": 1.3288975954055786,
+      "learning_rate": 0.0005974096060442526,
+      "loss": 5.6535,
+      "step": 500
+    },
+    {
+      "epoch": 0.05929919137466307,
+      "grad_norm": 2.041428804397583,
+      "learning_rate": 0.0005970858067997841,
+      "loss": 5.5725,
+      "step": 550
+    },
+    {
+      "epoch": 0.0646900269541779,
+      "grad_norm": 1.4345282316207886,
+      "learning_rate": 0.0005967620075553157,
+      "loss": 5.5081,
+      "step": 600
+    },
+    {
+      "epoch": 0.07008086253369272,
+      "grad_norm": 1.1454274654388428,
+      "learning_rate": 0.0005964382083108472,
+      "loss": 5.4223,
+      "step": 650
+    },
+    {
+      "epoch": 0.07547169811320754,
+      "grad_norm": 1.3169114589691162,
+      "learning_rate": 0.0005961144090663788,
+      "loss": 5.3711,
+      "step": 700
+    },
+    {
+      "epoch": 0.08086253369272237,
+      "grad_norm": 1.3919389247894287,
+      "learning_rate": 0.0005957906098219104,
+      "loss": 5.3003,
+      "step": 750
+    },
+    {
+      "epoch": 0.0862533692722372,
+      "grad_norm": 1.0709481239318848,
+      "learning_rate": 0.0005954668105774419,
+      "loss": 5.2637,
+      "step": 800
+    },
+    {
+      "epoch": 0.09164420485175202,
+      "grad_norm": 0.8944941759109497,
+      "learning_rate": 0.0005951430113329735,
+      "loss": 5.1967,
+      "step": 850
+    },
+    {
+      "epoch": 0.09703504043126684,
+      "grad_norm": 1.1984753608703613,
+      "learning_rate": 0.0005948192120885051,
+      "loss": 5.1709,
+      "step": 900
+    },
+    {
+      "epoch": 0.10242587601078167,
+      "grad_norm": 1.0900869369506836,
+      "learning_rate": 0.0005944954128440366,
+      "loss": 5.1332,
+      "step": 950
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "grad_norm": 1.4634649753570557,
+      "learning_rate": 0.0005941716135995682,
+      "loss": 5.0778,
+      "step": 1000
+    },
+    {
+      "epoch": 0.1078167115902965,
+      "eval_accuracy": 0.22697957684263617,
+      "eval_loss": 5.024423599243164,
+      "eval_runtime": 183.2168,
+      "eval_samples_per_second": 98.304,
+      "eval_steps_per_second": 6.146,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11320754716981132,
+      "grad_norm": 1.0724071264266968,
+      "learning_rate": 0.0005938478143550997,
+      "loss": 5.039,
+      "step": 1050
+    },
+    {
+      "epoch": 0.11859838274932614,
+      "grad_norm": 0.7434335947036743,
+      "learning_rate": 0.0005935240151106314,
+      "loss": 5.0145,
+      "step": 1100
+    },
+    {
+      "epoch": 0.12398921832884097,
+      "grad_norm": 1.1407990455627441,
+      "learning_rate": 0.0005932002158661629,
+      "loss": 4.9884,
+      "step": 1150
+    },
+    {
+      "epoch": 0.1293800539083558,
+      "grad_norm": 1.3579697608947754,
+      "learning_rate": 0.0005928764166216945,
+      "loss": 4.9232,
+      "step": 1200
+    },
+    {
+      "epoch": 0.1347708894878706,
+      "grad_norm": 1.170091986656189,
+      "learning_rate": 0.000592552617377226,
+      "loss": 4.9022,
+      "step": 1250
+    },
+    {
+      "epoch": 0.14016172506738545,
+      "grad_norm": 1.056681513786316,
+      "learning_rate": 0.0005922288181327577,
+      "loss": 4.8796,
+      "step": 1300
+    },
+    {
+      "epoch": 0.14555256064690028,
+      "grad_norm": 0.8261628746986389,
+      "learning_rate": 0.0005919050188882893,
+      "loss": 4.8425,
+      "step": 1350
+    },
+    {
+      "epoch": 0.1509433962264151,
+      "grad_norm": 0.7253502011299133,
+      "learning_rate": 0.0005915812196438207,
+      "loss": 4.836,
+      "step": 1400
+    },
+    {
+      "epoch": 0.15633423180592992,
+      "grad_norm": 1.0881081819534302,
+      "learning_rate": 0.0005912574203993524,
+      "loss": 4.8436,
+      "step": 1450
+    },
+    {
+      "epoch": 0.16172506738544473,
+      "grad_norm": 0.8581550717353821,
+      "learning_rate": 0.0005909336211548839,
+      "loss": 4.8009,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16711590296495957,
+      "grad_norm": 0.9875918030738831,
+      "learning_rate": 0.0005906098219104155,
+      "loss": 4.7541,
+      "step": 1550
+    },
+    {
+      "epoch": 0.1725067385444744,
+      "grad_norm": 0.8188138008117676,
+      "learning_rate": 0.000590286022665947,
+      "loss": 4.744,
+      "step": 1600
+    },
+    {
+      "epoch": 0.1778975741239892,
+      "grad_norm": 0.8326888084411621,
+      "learning_rate": 0.0005899622234214787,
+      "loss": 4.7177,
+      "step": 1650
+    },
+    {
+      "epoch": 0.18328840970350405,
+      "grad_norm": 1.0679171085357666,
+      "learning_rate": 0.0005896384241770102,
+      "loss": 4.6933,
+      "step": 1700
+    },
+    {
+      "epoch": 0.18867924528301888,
+      "grad_norm": 0.8685047626495361,
+      "learning_rate": 0.0005893146249325418,
+      "loss": 4.7025,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1940700808625337,
+      "grad_norm": 0.8494108319282532,
+      "learning_rate": 0.0005889908256880733,
+      "loss": 4.6703,
+      "step": 1800
+    },
+    {
+      "epoch": 0.19946091644204852,
+      "grad_norm": 1.175525426864624,
+      "learning_rate": 0.0005886670264436049,
+      "loss": 4.6366,
+      "step": 1850
+    },
+    {
+      "epoch": 0.20485175202156333,
+      "grad_norm": 1.0531870126724243,
+      "learning_rate": 0.0005883432271991365,
+      "loss": 4.6263,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21024258760107817,
+      "grad_norm": 0.9467228651046753,
+      "learning_rate": 0.0005880194279546681,
+      "loss": 4.6254,
+      "step": 1950
+    },
+    {
+      "epoch": 0.215633423180593,
+      "grad_norm": 0.9411425590515137,
+      "learning_rate": 0.0005876956287101996,
+      "loss": 4.5738,
+      "step": 2000
+    },
+    {
+      "epoch": 0.215633423180593,
+      "eval_accuracy": 0.271918085284042,
+      "eval_loss": 4.5008015632629395,
+      "eval_runtime": 181.4309,
+      "eval_samples_per_second": 99.272,
+      "eval_steps_per_second": 6.206,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2210242587601078,
+      "grad_norm": 0.920215368270874,
+      "learning_rate": 0.0005873718294657312,
+      "loss": 4.5559,
+      "step": 2050
+    },
+    {
+      "epoch": 0.22641509433962265,
+      "grad_norm": 0.7012114524841309,
+      "learning_rate": 0.0005870480302212628,
+      "loss": 4.5445,
+      "step": 2100
+    },
+    {
+      "epoch": 0.23180592991913745,
+      "grad_norm": 0.9538075923919678,
+      "learning_rate": 0.0005867242309767943,
+      "loss": 4.5259,
+      "step": 2150
+    },
+    {
+      "epoch": 0.2371967654986523,
+      "grad_norm": 0.7847324013710022,
+      "learning_rate": 0.0005864004317323259,
+      "loss": 4.4889,
+      "step": 2200
+    },
+    {
+      "epoch": 0.24258760107816713,
+      "grad_norm": 1.0440484285354614,
+      "learning_rate": 0.0005860766324878575,
+      "loss": 4.504,
+      "step": 2250
+    },
+    {
+      "epoch": 0.24797843665768193,
+      "grad_norm": 0.782093346118927,
+      "learning_rate": 0.000585752833243389,
+      "loss": 4.4768,
+      "step": 2300
+    },
+    {
+      "epoch": 0.25336927223719674,
+      "grad_norm": 0.9644035696983337,
+      "learning_rate": 0.0005854290339989206,
+      "loss": 4.4522,
+      "step": 2350
+    },
+    {
+      "epoch": 0.2587601078167116,
+      "grad_norm": 1.0412849187850952,
+      "learning_rate": 0.0005851052347544521,
+      "loss": 4.4489,
+      "step": 2400
+    },
+    {
+      "epoch": 0.2641509433962264,
+      "grad_norm": 0.7987921237945557,
+      "learning_rate": 0.0005847814355099838,
+      "loss": 4.4304,
+      "step": 2450
+    },
+    {
+      "epoch": 0.2695417789757412,
+      "grad_norm": 0.8646295070648193,
+      "learning_rate": 0.0005844576362655154,
+      "loss": 4.4196,
+      "step": 2500
+    },
+    {
+      "epoch": 0.2749326145552561,
+      "grad_norm": 1.0314178466796875,
+      "learning_rate": 0.0005841338370210469,
+      "loss": 4.4017,
+      "step": 2550
+    },
+    {
+      "epoch": 0.2803234501347709,
+      "grad_norm": 0.9287024140357971,
+      "learning_rate": 0.0005838100377765785,
+      "loss": 4.3729,
+      "step": 2600
+    },
+    {
+      "epoch": 0.2857142857142857,
+      "grad_norm": 0.8492061495780945,
+      "learning_rate": 0.0005834862385321101,
+      "loss": 4.37,
+      "step": 2650
+    },
+    {
+      "epoch": 0.29110512129380056,
+      "grad_norm": 0.7161704301834106,
+      "learning_rate": 0.0005831624392876417,
+      "loss": 4.3608,
+      "step": 2700
+    },
+    {
+      "epoch": 0.29649595687331537,
+      "grad_norm": 1.0816575288772583,
+      "learning_rate": 0.0005828386400431731,
+      "loss": 4.3705,
+      "step": 2750
+    },
+    {
+      "epoch": 0.3018867924528302,
+      "grad_norm": 0.8127713799476624,
+      "learning_rate": 0.0005825148407987048,
+      "loss": 4.3681,
+      "step": 2800
+    },
+    {
+      "epoch": 0.30727762803234504,
+      "grad_norm": 0.9167231321334839,
+      "learning_rate": 0.0005821910415542363,
+      "loss": 4.33,
+      "step": 2850
+    },
+    {
+      "epoch": 0.31266846361185985,
+      "grad_norm": 0.8297504782676697,
+      "learning_rate": 0.0005818672423097679,
+      "loss": 4.3137,
+      "step": 2900
+    },
+    {
+      "epoch": 0.31805929919137466,
+      "grad_norm": 0.9186894297599792,
+      "learning_rate": 0.0005815434430652994,
+      "loss": 4.3441,
+      "step": 2950
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "grad_norm": 0.7359763979911804,
+      "learning_rate": 0.0005812196438208311,
+      "loss": 4.309,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32345013477088946,
+      "eval_accuracy": 0.29948495271589304,
+      "eval_loss": 4.228354454040527,
+      "eval_runtime": 183.1945,
+      "eval_samples_per_second": 98.316,
+      "eval_steps_per_second": 6.146,
+      "step": 3000
+    },
+    {
+      "epoch": 0.3288409703504043,
+      "grad_norm": 0.7310630679130554,
+      "learning_rate": 0.0005808958445763626,
+      "loss": 4.3024,
+      "step": 3050
+    },
+    {
+      "epoch": 0.33423180592991913,
+      "grad_norm": 0.8260939717292786,
+      "learning_rate": 0.0005805720453318942,
+      "loss": 4.2831,
+      "step": 3100
+    },
+    {
+      "epoch": 0.33962264150943394,
+      "grad_norm": 0.7478753924369812,
+      "learning_rate": 0.0005802482460874257,
+      "loss": 4.2722,
+      "step": 3150
+    },
+    {
+      "epoch": 0.3450134770889488,
+      "grad_norm": 0.8138642907142639,
+      "learning_rate": 0.0005799244468429573,
+      "loss": 4.2816,
+      "step": 3200
+    },
+    {
+      "epoch": 0.3504043126684636,
+      "grad_norm": 0.7284408211708069,
+      "learning_rate": 0.0005796006475984889,
+      "loss": 4.2495,
+      "step": 3250
+    },
+    {
+      "epoch": 0.3557951482479784,
+      "grad_norm": 0.6755571365356445,
+      "learning_rate": 0.0005792768483540205,
+      "loss": 4.2595,
+      "step": 3300
+    },
+    {
+      "epoch": 0.3611859838274933,
+      "grad_norm": 0.9542193412780762,
+      "learning_rate": 0.000578953049109552,
+      "loss": 4.2538,
+      "step": 3350
+    },
+    {
+      "epoch": 0.3665768194070081,
+      "grad_norm": 0.6383764743804932,
+      "learning_rate": 0.0005786292498650836,
+      "loss": 4.2319,
+      "step": 3400
+    },
+    {
+      "epoch": 0.3719676549865229,
+      "grad_norm": 0.8126682639122009,
+      "learning_rate": 0.0005783054506206152,
+      "loss": 4.2272,
+      "step": 3450
+    },
+    {
+      "epoch": 0.37735849056603776,
+      "grad_norm": 0.7132463455200195,
+      "learning_rate": 0.0005779816513761467,
+      "loss": 4.2186,
+      "step": 3500
+    },
+    {
+      "epoch": 0.38274932614555257,
+      "grad_norm": 0.818424642086029,
+      "learning_rate": 0.0005776578521316782,
+      "loss": 4.1982,
+      "step": 3550
+    },
+    {
+      "epoch": 0.3881401617250674,
+      "grad_norm": 0.6916500926017761,
+      "learning_rate": 0.0005773340528872099,
+      "loss": 4.2057,
+      "step": 3600
+    },
+    {
+      "epoch": 0.3935309973045822,
+      "grad_norm": 0.6858584880828857,
+      "learning_rate": 0.0005770102536427414,
+      "loss": 4.2202,
+      "step": 3650
+    },
+    {
+      "epoch": 0.39892183288409705,
+      "grad_norm": 0.7574723362922668,
+      "learning_rate": 0.000576686454398273,
+      "loss": 4.2073,
+      "step": 3700
+    },
+    {
+      "epoch": 0.40431266846361186,
+      "grad_norm": 0.7107515931129456,
+      "learning_rate": 0.0005763626551538045,
+      "loss": 4.1939,
+      "step": 3750
+    },
+    {
+      "epoch": 0.40970350404312667,
+      "grad_norm": 0.7424683570861816,
+      "learning_rate": 0.0005760388559093362,
+      "loss": 4.1827,
+      "step": 3800
+    },
+    {
+      "epoch": 0.41509433962264153,
+      "grad_norm": 0.6541684865951538,
+      "learning_rate": 0.0005757150566648678,
+      "loss": 4.1782,
+      "step": 3850
+    },
+    {
+      "epoch": 0.42048517520215634,
+      "grad_norm": 0.797566294670105,
+      "learning_rate": 0.0005753912574203993,
+      "loss": 4.17,
+      "step": 3900
+    },
+    {
+      "epoch": 0.42587601078167114,
+      "grad_norm": 0.5999880433082581,
+      "learning_rate": 0.0005750674581759309,
+      "loss": 4.1651,
+      "step": 3950
+    },
+    {
+      "epoch": 0.431266846361186,
+      "grad_norm": 0.8264360427856445,
+      "learning_rate": 0.0005747436589314624,
+      "loss": 4.1524,
+      "step": 4000
+    },
+    {
+      "epoch": 0.431266846361186,
+      "eval_accuracy": 0.312240258480874,
+      "eval_loss": 4.087605953216553,
+      "eval_runtime": 183.2024,
+      "eval_samples_per_second": 98.312,
+      "eval_steps_per_second": 6.146,
+      "step": 4000
+    },
+    {
+      "epoch": 0.4366576819407008,
+      "grad_norm": 0.7332233786582947,
+      "learning_rate": 0.0005744198596869941,
+      "loss": 4.1654,
+      "step": 4050
+    },
+    {
+      "epoch": 0.4420485175202156,
+      "grad_norm": 0.5775137543678284,
+      "learning_rate": 0.0005740960604425255,
+      "loss": 4.158,
+      "step": 4100
+    },
+    {
+      "epoch": 0.4474393530997305,
+      "grad_norm": 0.8009674549102783,
+      "learning_rate": 0.0005737722611980572,
+      "loss": 4.1392,
+      "step": 4150
+    },
+    {
+      "epoch": 0.4528301886792453,
+      "grad_norm": 0.6772514581680298,
+      "learning_rate": 0.0005734484619535887,
+      "loss": 4.1186,
+      "step": 4200
+    },
+    {
+      "epoch": 0.4582210242587601,
+      "grad_norm": 0.6068300604820251,
+      "learning_rate": 0.0005731246627091203,
+      "loss": 4.1518,
+      "step": 4250
+    },
+    {
+      "epoch": 0.4636118598382749,
+      "grad_norm": 0.6603842973709106,
+      "learning_rate": 0.0005728008634646518,
+      "loss": 4.1228,
+      "step": 4300
+    },
+    {
+      "epoch": 0.46900269541778977,
+      "grad_norm": 0.6870344877243042,
+      "learning_rate": 0.0005724770642201835,
+      "loss": 4.1306,
+      "step": 4350
+    },
+    {
+      "epoch": 0.4743935309973046,
+      "grad_norm": 0.8701305985450745,
+      "learning_rate": 0.000572153264975715,
+      "loss": 4.1236,
+      "step": 4400
+    },
+    {
+      "epoch": 0.4797843665768194,
+      "grad_norm": 0.646145224571228,
+      "learning_rate": 0.0005718294657312466,
+      "loss": 4.0979,
+      "step": 4450
+    },
+    {
+      "epoch": 0.48517520215633425,
+      "grad_norm": 0.5847651362419128,
+      "learning_rate": 0.0005715056664867781,
+      "loss": 4.0943,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49056603773584906,
+      "grad_norm": 0.7485958933830261,
+      "learning_rate": 0.0005711818672423097,
+      "loss": 4.0872,
+      "step": 4550
+    },
+    {
+      "epoch": 0.49595687331536387,
+      "grad_norm": 0.6113314628601074,
+      "learning_rate": 0.0005708580679978413,
+      "loss": 4.0971,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5013477088948787,
+      "grad_norm": 0.814666748046875,
+      "learning_rate": 0.0005705342687533729,
+      "loss": 4.088,
+      "step": 4650
+    },
+    {
+      "epoch": 0.5067385444743935,
+      "grad_norm": 0.6813623905181885,
+      "learning_rate": 0.0005702104695089044,
+      "loss": 4.0697,
+      "step": 4700
+    },
+    {
+      "epoch": 0.5121293800539084,
+      "grad_norm": 0.7049392461776733,
+      "learning_rate": 0.000569886670264436,
+      "loss": 4.083,
+      "step": 4750
+    },
+    {
+      "epoch": 0.5175202156334232,
+      "grad_norm": 0.6689280867576599,
+      "learning_rate": 0.0005695628710199675,
+      "loss": 4.0704,
+      "step": 4800
+    },
+    {
+      "epoch": 0.522911051212938,
+      "grad_norm": 0.5872611403465271,
+      "learning_rate": 0.0005692390717754991,
+      "loss": 4.0659,
+      "step": 4850
+    },
+    {
+      "epoch": 0.5283018867924528,
+      "grad_norm": 0.7064571976661682,
+      "learning_rate": 0.0005689152725310306,
+      "loss": 4.0586,
+      "step": 4900
+    },
+    {
+      "epoch": 0.5336927223719676,
+      "grad_norm": 0.7021653652191162,
+      "learning_rate": 0.0005685914732865623,
+      "loss": 4.0619,
+      "step": 4950
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "grad_norm": 0.6203348636627197,
+      "learning_rate": 0.0005682676740420939,
+      "loss": 4.0714,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5390835579514824,
+      "eval_accuracy": 0.32160700663358477,
+      "eval_loss": 3.9901158809661865,
+      "eval_runtime": 183.2916,
+      "eval_samples_per_second": 98.264,
+      "eval_steps_per_second": 6.143,
+      "step": 5000
+    },
+    {
+      "epoch": 0.5444743935309974,
+      "grad_norm": 0.7498524188995361,
+      "learning_rate": 0.0005679438747976254,
+      "loss": 4.0504,
+      "step": 5050
+    },
+    {
+      "epoch": 0.5498652291105122,
+      "grad_norm": 0.6415616869926453,
+      "learning_rate": 0.000567620075553157,
+      "loss": 4.045,
+      "step": 5100
+    },
+    {
+      "epoch": 0.555256064690027,
+      "grad_norm": 0.68656986951828,
+      "learning_rate": 0.0005672962763086886,
+      "loss": 4.0469,
+      "step": 5150
+    },
+    {
+      "epoch": 0.5606469002695418,
+      "grad_norm": 0.6866742968559265,
+      "learning_rate": 0.0005669724770642202,
+      "loss": 4.0368,
+      "step": 5200
+    },
+    {
+      "epoch": 0.5660377358490566,
+      "grad_norm": 0.6404621601104736,
+      "learning_rate": 0.0005666486778197517,
+      "loss": 4.0323,
+      "step": 5250
+    },
+    {
+      "epoch": 0.5714285714285714,
+      "grad_norm": 0.6272525191307068,
+      "learning_rate": 0.0005663248785752833,
+      "loss": 4.0431,
+      "step": 5300
+    },
+    {
+      "epoch": 0.5768194070080862,
+      "grad_norm": 0.6219531297683716,
+      "learning_rate": 0.0005660010793308148,
+      "loss": 4.0283,
+      "step": 5350
+    },
+    {
+      "epoch": 0.5822102425876011,
+      "grad_norm": 0.5969902276992798,
+      "learning_rate": 0.0005656772800863465,
+      "loss": 4.0455,
+      "step": 5400
+    },
+    {
+      "epoch": 0.5876010781671159,
+      "grad_norm": 0.7162837982177734,
+      "learning_rate": 0.0005653534808418779,
+      "loss": 4.0187,
+      "step": 5450
+    },
+    {
+      "epoch": 0.5929919137466307,
+      "grad_norm": 0.5840233564376831,
+      "learning_rate": 0.0005650296815974096,
+      "loss": 4.0264,
+      "step": 5500
+    },
+    {
+      "epoch": 0.5983827493261455,
+      "grad_norm": 0.5966022610664368,
+      "learning_rate": 0.0005647058823529411,
+      "loss": 4.0154,
+      "step": 5550
+    },
+    {
+      "epoch": 0.6037735849056604,
+      "grad_norm": 0.5551064610481262,
+      "learning_rate": 0.0005643820831084727,
+      "loss": 4.0069,
+      "step": 5600
+    },
+    {
+      "epoch": 0.6091644204851752,
+      "grad_norm": 0.7473669052124023,
+      "learning_rate": 0.0005640582838640042,
+      "loss": 4.0204,
+      "step": 5650
+    },
+    {
+      "epoch": 0.6145552560646901,
+      "grad_norm": 0.6089016795158386,
+      "learning_rate": 0.0005637344846195358,
+      "loss": 4.0132,
+      "step": 5700
+    },
+    {
+      "epoch": 0.6199460916442049,
+      "grad_norm": 0.5742376446723938,
+      "learning_rate": 0.0005634106853750674,
+      "loss": 4.02,
+      "step": 5750
+    },
+    {
+      "epoch": 0.6253369272237197,
+      "grad_norm": 0.857323169708252,
+      "learning_rate": 0.000563086886130599,
+      "loss": 3.9884,
+      "step": 5800
+    },
+    {
+      "epoch": 0.6307277628032345,
+      "grad_norm": 0.6711622476577759,
+      "learning_rate": 0.0005627630868861305,
+      "loss": 3.9889,
+      "step": 5850
+    },
+    {
+      "epoch": 0.6361185983827493,
+      "grad_norm": 0.6008835434913635,
+      "learning_rate": 0.0005624392876416621,
+      "loss": 3.9763,
+      "step": 5900
+    },
+    {
+      "epoch": 0.6415094339622641,
+      "grad_norm": 0.6592042446136475,
+      "learning_rate": 0.0005621154883971937,
+      "loss": 3.9816,
+      "step": 5950
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "grad_norm": 0.6811545491218567,
+      "learning_rate": 0.0005617916891527253,
+      "loss": 3.9631,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6469002695417789,
+      "eval_accuracy": 0.328109556010618,
+      "eval_loss": 3.9180023670196533,
+      "eval_runtime": 183.5563,
+      "eval_samples_per_second": 98.122,
+      "eval_steps_per_second": 6.134,
+      "step": 6000
+    },
+    {
+      "epoch": 0.6522911051212938,
+      "grad_norm": 0.838979959487915,
+      "learning_rate": 0.0005614678899082568,
+      "loss": 3.9977,
+      "step": 6050
+    },
+    {
+      "epoch": 0.6576819407008087,
+      "grad_norm": 0.6069373488426208,
+      "learning_rate": 0.0005611440906637884,
+      "loss": 3.9844,
+      "step": 6100
+    },
+    {
+      "epoch": 0.6630727762803235,
+      "grad_norm": 0.6294558048248291,
+      "learning_rate": 0.00056082029141932,
+      "loss": 3.9769,
+      "step": 6150
+    },
+    {
+      "epoch": 0.6684636118598383,
+      "grad_norm": 0.6527539491653442,
+      "learning_rate": 0.0005604964921748515,
+      "loss": 3.9875,
+      "step": 6200
+    },
+    {
+      "epoch": 0.6738544474393531,
+      "grad_norm": 0.6204699873924255,
+      "learning_rate": 0.000560172692930383,
+      "loss": 3.949,
+      "step": 6250
+    },
+    {
+      "epoch": 0.6792452830188679,
+      "grad_norm": 0.6665420532226562,
+      "learning_rate": 0.0005598488936859147,
+      "loss": 3.96,
+      "step": 6300
+    },
+    {
+      "epoch": 0.6846361185983828,
+      "grad_norm": 0.6489077806472778,
+      "learning_rate": 0.0005595250944414463,
+      "loss": 3.9534,
+      "step": 6350
+    },
+    {
+      "epoch": 0.6900269541778976,
+      "grad_norm": 0.6623448729515076,
+      "learning_rate": 0.0005592012951969778,
+      "loss": 3.9454,
+      "step": 6400
+    },
+    {
+      "epoch": 0.6954177897574124,
+      "grad_norm": 0.7035852670669556,
+      "learning_rate": 0.0005588774959525094,
+      "loss": 3.9722,
+      "step": 6450
+    },
+    {
+      "epoch": 0.7008086253369272,
+      "grad_norm": 0.5686920881271362,
+      "learning_rate": 0.000558553696708041,
+      "loss": 3.9525,
+      "step": 6500
+    },
+    {
+      "epoch": 0.706199460916442,
+      "grad_norm": 0.5780633091926575,
+      "learning_rate": 0.0005582298974635726,
+      "loss": 3.9525,
+      "step": 6550
+    },
+    {
+      "epoch": 0.7115902964959568,
+      "grad_norm": 0.6254565715789795,
+      "learning_rate": 0.0005579060982191041,
+      "loss": 3.9494,
+      "step": 6600
+    },
+    {
+      "epoch": 0.7169811320754716,
+      "grad_norm": 0.6464234590530396,
+      "learning_rate": 0.0005575822989746357,
+      "loss": 3.9432,
+      "step": 6650
+    },
+    {
+      "epoch": 0.7223719676549866,
+      "grad_norm": 0.5895872712135315,
+      "learning_rate": 0.0005572584997301672,
+      "loss": 3.9378,
+      "step": 6700
+    },
+    {
+      "epoch": 0.7277628032345014,
+      "grad_norm": 0.6661088466644287,
+      "learning_rate": 0.0005569347004856989,
+      "loss": 3.9405,
+      "step": 6750
+    },
+    {
+      "epoch": 0.7331536388140162,
+      "grad_norm": 0.5486308932304382,
+      "learning_rate": 0.0005566109012412303,
+      "loss": 3.9499,
+      "step": 6800
+    },
+    {
+      "epoch": 0.738544474393531,
+      "grad_norm": 0.5206322073936462,
+      "learning_rate": 0.000556287101996762,
+      "loss": 3.9308,
+      "step": 6850
+    },
+    {
+      "epoch": 0.7439353099730458,
+      "grad_norm": 0.60262531042099,
+      "learning_rate": 0.0005559633027522935,
+      "loss": 3.9279,
+      "step": 6900
+    },
+    {
+      "epoch": 0.7493261455525606,
+      "grad_norm": 0.6978471875190735,
+      "learning_rate": 0.0005556395035078251,
+      "loss": 3.9152,
+      "step": 6950
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "grad_norm": 0.5400425791740417,
+      "learning_rate": 0.0005553157042633566,
+      "loss": 3.9285,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7547169811320755,
+      "eval_accuracy": 0.33344778104183126,
+      "eval_loss": 3.8615846633911133,
+      "eval_runtime": 182.602,
+      "eval_samples_per_second": 98.635,
+      "eval_steps_per_second": 6.166,
+      "step": 7000
+    },
+    {
+      "epoch": 0.7601078167115903,
+      "grad_norm": 0.5189043283462524,
+      "learning_rate": 0.0005549919050188882,
+      "loss": 3.9302,
+      "step": 7050
+    },
+    {
+      "epoch": 0.7654986522911051,
+      "grad_norm": 0.5751085877418518,
+      "learning_rate": 0.0005546681057744198,
+      "loss": 3.9332,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77088948787062,
+      "grad_norm": 0.6791032552719116,
+      "learning_rate": 0.0005543443065299514,
+      "loss": 3.9073,
+      "step": 7150
+    },
+    {
+      "epoch": 0.7762803234501348,
+      "grad_norm": 0.7200894951820374,
+      "learning_rate": 0.000554020507285483,
+      "loss": 3.9193,
+      "step": 7200
+    },
+    {
+      "epoch": 0.7816711590296496,
+      "grad_norm": 0.6101612448692322,
+      "learning_rate": 0.0005536967080410145,
+      "loss": 3.919,
+      "step": 7250
+    },
+    {
+      "epoch": 0.7870619946091644,
+      "grad_norm": 0.5997413396835327,
+      "learning_rate": 0.0005533729087965462,
+      "loss": 3.9079,
+      "step": 7300
+    },
+    {
+      "epoch": 0.7924528301886793,
+      "grad_norm": 0.5748898386955261,
+      "learning_rate": 0.0005530491095520777,
+      "loss": 3.9162,
+      "step": 7350
+    },
+    {
+      "epoch": 0.7978436657681941,
+      "grad_norm": 0.6359212398529053,
+      "learning_rate": 0.0005527253103076093,
+      "loss": 3.8807,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8032345013477089,
+      "grad_norm": 0.6088876724243164,
+      "learning_rate": 0.0005524015110631408,
+      "loss": 3.8902,
+      "step": 7450
+    },
+    {
+      "epoch": 0.8086253369272237,
+      "grad_norm": 0.6242630481719971,
+      "learning_rate": 0.0005520777118186724,
+      "loss": 3.9035,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8140161725067385,
+      "grad_norm": 0.622336208820343,
+      "learning_rate": 0.0005517539125742039,
+      "loss": 3.8986,
+      "step": 7550
+    },
+    {
+      "epoch": 0.8194070080862533,
+      "grad_norm": 0.5424439311027527,
+      "learning_rate": 0.0005514301133297355,
+      "loss": 3.9166,
+      "step": 7600
+    },
+    {
+      "epoch": 0.8247978436657682,
+      "grad_norm": 0.5808700323104858,
+      "learning_rate": 0.0005511063140852671,
+      "loss": 3.9005,
+      "step": 7650
+    },
+    {
+      "epoch": 0.8301886792452831,
+      "grad_norm": 0.5555144548416138,
+      "learning_rate": 0.0005507825148407987,
+      "loss": 3.8722,
+      "step": 7700
+    },
+    {
+      "epoch": 0.8355795148247979,
+      "grad_norm": 0.7033872604370117,
+      "learning_rate": 0.0005504587155963302,
+      "loss": 3.8852,
+      "step": 7750
+    },
+    {
+      "epoch": 0.8409703504043127,
+      "grad_norm": 0.5577759742736816,
+      "learning_rate": 0.0005501349163518618,
+      "loss": 3.9008,
+      "step": 7800
+    },
+    {
+      "epoch": 0.8463611859838275,
+      "grad_norm": Infinity,
+      "learning_rate": 0.0005498175930922827,
+      "loss": 3.888,
+      "step": 7850
+    },
+    {
+      "epoch": 0.8517520215633423,
+      "grad_norm": 0.6767401099205017,
+      "learning_rate": 0.0005494937938478143,
+      "loss": 3.8863,
+      "step": 7900
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.5402886271476746,
+      "learning_rate": 0.0005491699946033459,
+      "loss": 3.8885,
+      "step": 7950
+    },
+    {
+      "epoch": 0.862533692722372,
+      "grad_norm": 0.5724102258682251,
+      "learning_rate": 0.0005488461953588775,
+      "loss": 3.8676,
+      "step": 8000
+    },
+    {
+      "epoch": 0.862533692722372,
+      "eval_accuracy": 0.3375052058314874,
+      "eval_loss": 3.816683292388916,
+      "eval_runtime": 181.6977,
+      "eval_samples_per_second": 99.126,
+      "eval_steps_per_second": 6.197,
+      "step": 8000
+    },
+    {
+      "epoch": 0.8679245283018868,
+      "grad_norm": 0.5436832308769226,
+      "learning_rate": 0.000548522396114409,
+      "loss": 3.8838,
+      "step": 8050
+    },
+    {
+      "epoch": 0.8733153638814016,
+      "grad_norm": 0.6710783243179321,
+      "learning_rate": 0.0005481985968699406,
+      "loss": 3.8623,
+      "step": 8100
+    },
+    {
+      "epoch": 0.8787061994609164,
+      "grad_norm": 0.5671558380126953,
+      "learning_rate": 0.0005478747976254721,
+      "loss": 3.8746,
+      "step": 8150
+    },
+    {
+      "epoch": 0.8840970350404312,
+      "grad_norm": 0.6092495322227478,
+      "learning_rate": 0.0005475509983810037,
+      "loss": 3.8635,
+      "step": 8200
+    },
+    {
+      "epoch": 0.889487870619946,
+      "grad_norm": 0.6280654072761536,
+      "learning_rate": 0.0005472271991365352,
+      "loss": 3.8772,
+      "step": 8250
+    },
+    {
+      "epoch": 0.894878706199461,
+      "grad_norm": 0.5402874946594238,
+      "learning_rate": 0.0005469033998920669,
+      "loss": 3.8525,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9002695417789758,
+      "grad_norm": 0.5597426295280457,
+      "learning_rate": 0.0005465796006475984,
+      "loss": 3.8685,
+      "step": 8350
+    },
+    {
+      "epoch": 0.9056603773584906,
+      "grad_norm": 0.5542333126068115,
+      "learning_rate": 0.00054625580140313,
+      "loss": 3.8677,
+      "step": 8400
+    },
+    {
+      "epoch": 0.9110512129380054,
+      "grad_norm": 0.5573087930679321,
+      "learning_rate": 0.0005459320021586615,
+      "loss": 3.8402,
+      "step": 8450
+    },
+    {
+      "epoch": 0.9164420485175202,
+      "grad_norm": 0.5278663039207458,
+      "learning_rate": 0.0005456082029141932,
+      "loss": 3.8521,
+      "step": 8500
+    },
+    {
+      "epoch": 0.921832884097035,
+      "grad_norm": 0.6010989546775818,
+      "learning_rate": 0.0005452844036697248,
+      "loss": 3.8536,
+      "step": 8550
+    },
+    {
+      "epoch": 0.9272237196765498,
+      "grad_norm": 0.6041896939277649,
+      "learning_rate": 0.0005449606044252563,
+      "loss": 3.8816,
+      "step": 8600
+    },
+    {
+      "epoch": 0.9326145552560647,
+      "grad_norm": 0.6230564117431641,
+      "learning_rate": 0.0005446368051807879,
+      "loss": 3.8547,
+      "step": 8650
+    },
+    {
+      "epoch": 0.9380053908355795,
+      "grad_norm": 0.6061603426933289,
+      "learning_rate": 0.0005443130059363194,
+      "loss": 3.8612,
+      "step": 8700
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.532725989818573,
+      "learning_rate": 0.0005439892066918511,
+      "loss": 3.8453,
+      "step": 8750
+    },
+    {
+      "epoch": 0.9487870619946092,
+      "grad_norm": 0.6650099158287048,
+      "learning_rate": 0.0005436654074473825,
+      "loss": 3.845,
+      "step": 8800
+    },
+    {
+      "epoch": 0.954177897574124,
+      "grad_norm": 0.5258676409721375,
+      "learning_rate": 0.0005433416082029142,
+      "loss": 3.8474,
+      "step": 8850
+    },
+    {
+      "epoch": 0.9595687331536388,
+      "grad_norm": 0.5918166637420654,
+      "learning_rate": 0.0005430178089584457,
+      "loss": 3.8476,
+      "step": 8900
+    },
+    {
+      "epoch": 0.9649595687331537,
+      "grad_norm": 0.7267642617225647,
+      "learning_rate": 0.0005426940097139773,
+      "loss": 3.8489,
+      "step": 8950
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "grad_norm": 0.639377772808075,
+      "learning_rate": 0.0005423702104695088,
+      "loss": 3.8648,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9703504043126685,
+      "eval_accuracy": 0.3408996306996996,
+      "eval_loss": 3.779827356338501,
+      "eval_runtime": 181.5426,
+      "eval_samples_per_second": 99.211,
+      "eval_steps_per_second": 6.202,
+      "step": 9000
+    },
+    {
+      "epoch": 0.9757412398921833,
+      "grad_norm": 0.6780883073806763,
+      "learning_rate": 0.0005420464112250404,
+      "loss": 3.8448,
+      "step": 9050
+    },
+    {
+      "epoch": 0.9811320754716981,
+      "grad_norm": 0.5915326476097107,
+      "learning_rate": 0.000541722611980572,
+      "loss": 3.8335,
+      "step": 9100
+    },
+    {
+      "epoch": 0.9865229110512129,
+      "grad_norm": 0.5880796313285828,
+      "learning_rate": 0.0005413988127361036,
+      "loss": 3.8434,
+      "step": 9150
+    },
+    {
+      "epoch": 0.9919137466307277,
+      "grad_norm": 0.551231861114502,
+      "learning_rate": 0.0005410750134916351,
+      "loss": 3.8469,
+      "step": 9200
+    },
+    {
+      "epoch": 0.9973045822102425,
+      "grad_norm": 0.542789101600647,
+      "learning_rate": 0.0005407512142471667,
+      "loss": 3.8382,
+      "step": 9250
+    },
+    {
+      "epoch": 1.0026954177897573,
+      "grad_norm": 0.5782141089439392,
+      "learning_rate": 0.0005404274150026983,
+      "loss": 3.8143,
+      "step": 9300
+    },
+    {
+      "epoch": 1.0080862533692723,
+      "grad_norm": 0.585417628288269,
+      "learning_rate": 0.0005401036157582299,
+      "loss": 3.782,
+      "step": 9350
+    },
+    {
+      "epoch": 1.013477088948787,
+      "grad_norm": 0.5786333680152893,
+      "learning_rate": 0.0005397798165137614,
+      "loss": 3.7919,
+      "step": 9400
+    },
+    {
+      "epoch": 1.0188679245283019,
+      "grad_norm": 0.5752071738243103,
+      "learning_rate": 0.000539456017269293,
+      "loss": 3.7809,
+      "step": 9450
+    },
+    {
+      "epoch": 1.0242587601078168,
+      "grad_norm": 0.5453478693962097,
+      "learning_rate": 0.0005391322180248245,
+      "loss": 3.7775,
+      "step": 9500
+    },
+    {
+      "epoch": 1.0296495956873315,
+      "grad_norm": 0.5964511632919312,
+      "learning_rate": 0.0005388084187803561,
+      "loss": 3.7881,
+      "step": 9550
+    },
+    {
+      "epoch": 1.0350404312668464,
+      "grad_norm": 0.6188067197799683,
+      "learning_rate": 0.0005384846195358876,
+      "loss": 3.7672,
+      "step": 9600
+    },
+    {
+      "epoch": 1.0404312668463611,
+      "grad_norm": 0.5592512488365173,
+      "learning_rate": 0.0005381608202914193,
+      "loss": 3.7615,
+      "step": 9650
+    },
+    {
+      "epoch": 1.045822102425876,
+      "grad_norm": 0.5626137256622314,
+      "learning_rate": 0.0005378370210469509,
+      "loss": 3.7723,
+      "step": 9700
+    },
+    {
+      "epoch": 1.0512129380053907,
+      "grad_norm": 0.6109785437583923,
+      "learning_rate": 0.0005375132218024824,
+      "loss": 3.7644,
+      "step": 9750
+    },
+    {
+      "epoch": 1.0566037735849056,
+      "grad_norm": 0.6150305271148682,
+      "learning_rate": 0.000537189422558014,
+      "loss": 3.7717,
+      "step": 9800
+    },
+    {
+      "epoch": 1.0619946091644206,
+      "grad_norm": 0.5786086320877075,
+      "learning_rate": 0.0005368656233135455,
+      "loss": 3.7537,
+      "step": 9850
+    },
+    {
+      "epoch": 1.0673854447439353,
+      "grad_norm": 0.5935583710670471,
+      "learning_rate": 0.0005365418240690772,
+      "loss": 3.7646,
+      "step": 9900
+    },
+    {
+      "epoch": 1.0727762803234502,
+      "grad_norm": 0.5563629269599915,
+      "learning_rate": 0.0005362180248246087,
+      "loss": 3.7625,
+      "step": 9950
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "grad_norm": 0.5583340525627136,
+      "learning_rate": 0.0005358942255801403,
+      "loss": 3.7661,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0781671159029649,
+      "eval_accuracy": 0.3447591985806456,
+      "eval_loss": 3.7474043369293213,
+      "eval_runtime": 179.3333,
+      "eval_samples_per_second": 100.433,
+      "eval_steps_per_second": 6.279,
+      "step": 10000
+    },
+    {
+      "epoch": 1.0835579514824798,
+      "grad_norm": 0.5278469324111938,
+      "learning_rate": 0.0005355704263356718,
+      "loss": 3.7572,
+      "step": 10050
+    },
+    {
+      "epoch": 1.0889487870619945,
+      "grad_norm": 0.599793553352356,
+      "learning_rate": 0.0005352466270912035,
+      "loss": 3.7482,
+      "step": 10100
+    },
+    {
+      "epoch": 1.0943396226415094,
+      "grad_norm": 0.594103991985321,
+      "learning_rate": 0.000534922827846735,
+      "loss": 3.7811,
+      "step": 10150
+    },
+    {
+      "epoch": 1.0997304582210243,
+      "grad_norm": 0.539659321308136,
+      "learning_rate": 0.0005345990286022666,
+      "loss": 3.7808,
+      "step": 10200
+    },
+    {
+      "epoch": 1.105121293800539,
+      "grad_norm": 0.5949695110321045,
+      "learning_rate": 0.0005342752293577981,
+      "loss": 3.7483,
+      "step": 10250
+    },
+    {
+      "epoch": 1.110512129380054,
+      "grad_norm": 0.6354559659957886,
+      "learning_rate": 0.0005339514301133297,
+      "loss": 3.7743,
+      "step": 10300
+    },
+    {
+      "epoch": 1.1159029649595686,
+      "grad_norm": 0.5751685500144958,
+      "learning_rate": 0.0005336276308688612,
+      "loss": 3.7727,
+      "step": 10350
+    },
+    {
+      "epoch": 1.1212938005390836,
+      "grad_norm": 0.6152482032775879,
+      "learning_rate": 0.0005333038316243928,
+      "loss": 3.7604,
+      "step": 10400
+    },
+    {
+      "epoch": 1.1266846361185983,
+      "grad_norm": 0.6259979009628296,
+      "learning_rate": 0.0005329800323799244,
+      "loss": 3.7242,
+      "step": 10450
+    },
+    {
+      "epoch": 1.1320754716981132,
+      "grad_norm": 0.5735751390457153,
+      "learning_rate": 0.000532656233135456,
+      "loss": 3.75,
+      "step": 10500
+    },
+    {
+      "epoch": 1.137466307277628,
+      "grad_norm": 0.5335408449172974,
+      "learning_rate": 0.0005323324338909875,
+      "loss": 3.7497,
+      "step": 10550
+    },
+    {
+      "epoch": 1.1428571428571428,
+      "grad_norm": 0.5535331964492798,
+      "learning_rate": 0.0005320086346465191,
+      "loss": 3.757,
+      "step": 10600
+    },
+    {
+      "epoch": 1.1482479784366577,
+      "grad_norm": 0.5518361926078796,
+      "learning_rate": 0.0005316848354020507,
+      "loss": 3.7635,
+      "step": 10650
+    },
+    {
+      "epoch": 1.1536388140161726,
+      "grad_norm": 0.6065233945846558,
+      "learning_rate": 0.0005313610361575823,
+      "loss": 3.7482,
+      "step": 10700
+    },
+    {
+      "epoch": 1.1590296495956873,
+      "grad_norm": 0.5510832667350769,
+      "learning_rate": 0.0005310372369131138,
+      "loss": 3.7464,
+      "step": 10750
+    },
+    {
+      "epoch": 1.1644204851752022,
+      "grad_norm": 0.5745865702629089,
+      "learning_rate": 0.0005307134376686454,
+      "loss": 3.7237,
+      "step": 10800
+    },
+    {
+      "epoch": 1.169811320754717,
+      "grad_norm": 0.6756680011749268,
+      "learning_rate": 0.000530389638424177,
+      "loss": 3.7627,
+      "step": 10850
+    },
+    {
+      "epoch": 1.1752021563342319,
+      "grad_norm": 0.6508898138999939,
+      "learning_rate": 0.0005300658391797085,
+      "loss": 3.7616,
+      "step": 10900
+    },
+    {
+      "epoch": 1.1805929919137466,
+      "grad_norm": 0.6372632384300232,
+      "learning_rate": 0.00052974203993524,
+      "loss": 3.7205,
+      "step": 10950
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "grad_norm": 0.584235668182373,
+      "learning_rate": 0.0005294182406907717,
+      "loss": 3.7415,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1859838274932615,
+      "eval_accuracy": 0.3469530090385078,
+      "eval_loss": 3.7233991622924805,
+      "eval_runtime": 179.2943,
+      "eval_samples_per_second": 100.455,
+      "eval_steps_per_second": 6.28,
+      "step": 11000
+    },
+    {
+      "epoch": 1.1913746630727764,
+      "grad_norm": 0.5607488751411438,
+      "learning_rate": 0.0005290944414463033,
+      "loss": 3.7504,
+      "step": 11050
+    },
+    {
+      "epoch": 1.196765498652291,
+      "grad_norm": 0.6091164946556091,
+      "learning_rate": 0.0005287706422018348,
+      "loss": 3.744,
+      "step": 11100
+    },
+    {
+      "epoch": 1.202156334231806,
+      "grad_norm": 0.6206035614013672,
+      "learning_rate": 0.0005284468429573664,
+      "loss": 3.7378,
+      "step": 11150
+    },
+    {
+      "epoch": 1.2075471698113207,
+      "grad_norm": 0.6528010964393616,
+      "learning_rate": 0.0005281230437128979,
+      "loss": 3.7509,
+      "step": 11200
+    },
+    {
+      "epoch": 1.2129380053908356,
+      "grad_norm": 0.520330011844635,
+      "learning_rate": 0.0005278057204533189,
+      "loss": 3.7608,
+      "step": 11250
+    },
+    {
+      "epoch": 1.2183288409703503,
+      "grad_norm": 0.6227236390113831,
+      "learning_rate": 0.0005274819212088505,
+      "loss": 3.7286,
+      "step": 11300
+    },
+    {
+      "epoch": 1.2237196765498652,
+      "grad_norm": 0.5682984590530396,
+      "learning_rate": 0.0005271581219643821,
+      "loss": 3.7505,
+      "step": 11350
+    },
+    {
+      "epoch": 1.2291105121293802,
+      "grad_norm": 0.6123554706573486,
+      "learning_rate": 0.0005268343227199136,
+      "loss": 3.7508,
+      "step": 11400
+    },
+    {
+      "epoch": 1.2345013477088949,
+      "grad_norm": 0.507030189037323,
+      "learning_rate": 0.0005265105234754452,
+      "loss": 3.7553,
+      "step": 11450
+    },
+    {
+      "epoch": 1.2398921832884098,
+      "grad_norm": 0.5759048461914062,
+      "learning_rate": 0.0005261867242309767,
+      "loss": 3.7192,
+      "step": 11500
+    },
+    {
+      "epoch": 1.2452830188679245,
+      "grad_norm": 0.5370573997497559,
+      "learning_rate": 0.0005258629249865083,
+      "loss": 3.7409,
+      "step": 11550
+    },
+    {
+      "epoch": 1.2506738544474394,
+      "grad_norm": 0.5597508549690247,
+      "learning_rate": 0.0005255391257420398,
+      "loss": 3.739,
+      "step": 11600
+    },
+    {
+      "epoch": 1.256064690026954,
+      "grad_norm": 0.6163648366928101,
+      "learning_rate": 0.0005252153264975715,
+      "loss": 3.7364,
+      "step": 11650
+    },
+    {
+      "epoch": 1.261455525606469,
+      "grad_norm": 0.561890721321106,
+      "learning_rate": 0.000524891527253103,
+      "loss": 3.7371,
+      "step": 11700
+    },
+    {
+      "epoch": 1.266846361185984,
+      "grad_norm": 0.6164953708648682,
+      "learning_rate": 0.0005245677280086346,
+      "loss": 3.7437,
+      "step": 11750
+    },
+    {
+      "epoch": 1.2722371967654986,
+      "grad_norm": 0.5438469648361206,
+      "learning_rate": 0.0005242439287641661,
+      "loss": 3.7317,
+      "step": 11800
+    },
+    {
+      "epoch": 1.2776280323450135,
+      "grad_norm": 0.5996424555778503,
+      "learning_rate": 0.0005239201295196978,
+      "loss": 3.7139,
+      "step": 11850
+    },
+    {
+      "epoch": 1.2830188679245282,
+      "grad_norm": 0.5443153381347656,
+      "learning_rate": 0.0005235963302752293,
+      "loss": 3.7189,
+      "step": 11900
+    },
+    {
+      "epoch": 1.2884097035040432,
+      "grad_norm": 0.5056083798408508,
+      "learning_rate": 0.0005232725310307609,
+      "loss": 3.7286,
+      "step": 11950
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "grad_norm": 0.5346047878265381,
+      "learning_rate": 0.0005229487317862924,
+      "loss": 3.7241,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2938005390835579,
+      "eval_accuracy": 0.3486271327339533,
+      "eval_loss": 3.701392889022827,
+      "eval_runtime": 179.3961,
+      "eval_samples_per_second": 100.398,
+      "eval_steps_per_second": 6.277,
+      "step": 12000
+    },
+    {
+      "epoch": 1.2991913746630728,
+      "grad_norm": 0.561946451663971,
+      "learning_rate": 0.000522624932541824,
+      "loss": 3.7065,
+      "step": 12050
+    },
+    {
+      "epoch": 1.3045822102425877,
+      "grad_norm": 0.6094053387641907,
+      "learning_rate": 0.0005223011332973557,
+      "loss": 3.7432,
+      "step": 12100
+    },
+    {
+      "epoch": 1.3099730458221024,
+      "grad_norm": 0.548875093460083,
+      "learning_rate": 0.0005219773340528872,
+      "loss": 3.7251,
+      "step": 12150
+    },
+    {
+      "epoch": 1.3153638814016173,
+      "grad_norm": 0.5728870630264282,
+      "learning_rate": 0.0005216535348084188,
+      "loss": 3.7107,
+      "step": 12200
+    },
+    {
+      "epoch": 1.320754716981132,
+      "grad_norm": 0.6108008623123169,
+      "learning_rate": 0.0005213297355639503,
+      "loss": 3.7369,
+      "step": 12250
+    },
+    {
+      "epoch": 1.326145552560647,
+      "grad_norm": 0.5879302024841309,
+      "learning_rate": 0.0005210059363194819,
+      "loss": 3.7043,
+      "step": 12300
+    },
+    {
+      "epoch": 1.3315363881401616,
+      "grad_norm": 0.550482988357544,
+      "learning_rate": 0.0005206821370750134,
+      "loss": 3.7288,
+      "step": 12350
+    },
+    {
+      "epoch": 1.3369272237196765,
+      "grad_norm": 0.5876588821411133,
+      "learning_rate": 0.000520358337830545,
+      "loss": 3.7,
+      "step": 12400
+    },
+    {
+      "epoch": 1.3423180592991915,
+      "grad_norm": 0.5548244118690491,
+      "learning_rate": 0.0005200345385860766,
+      "loss": 3.7202,
+      "step": 12450
+    },
+    {
+      "epoch": 1.3477088948787062,
+      "grad_norm": 0.6892051696777344,
+      "learning_rate": 0.0005197107393416082,
+      "loss": 3.7287,
+      "step": 12500
+    },
+    {
+      "epoch": 1.353099730458221,
+      "grad_norm": 0.5671435594558716,
+      "learning_rate": 0.0005193869400971397,
+      "loss": 3.7361,
+      "step": 12550
+    },
+    {
+      "epoch": 1.3584905660377358,
+      "grad_norm": 0.5692296624183655,
+      "learning_rate": 0.0005190631408526713,
+      "loss": 3.7137,
+      "step": 12600
+    },
+    {
+      "epoch": 1.3638814016172507,
+      "grad_norm": 0.5560773611068726,
+      "learning_rate": 0.0005187393416082029,
+      "loss": 3.7092,
+      "step": 12650
+    },
+    {
+      "epoch": 1.3692722371967654,
+      "grad_norm": 0.5214106440544128,
+      "learning_rate": 0.0005184155423637345,
+      "loss": 3.7076,
+      "step": 12700
+    },
+    {
+      "epoch": 1.3746630727762803,
+      "grad_norm": 0.5432198643684387,
+      "learning_rate": 0.000518091743119266,
+      "loss": 3.7224,
+      "step": 12750
+    },
+    {
+      "epoch": 1.3800539083557952,
+      "grad_norm": 0.5707585215568542,
+      "learning_rate": 0.0005177679438747976,
+      "loss": 3.7136,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38544474393531,
+      "grad_norm": 0.583604633808136,
+      "learning_rate": 0.0005174441446303291,
+      "loss": 3.7356,
+      "step": 12850
+    },
+    {
+      "epoch": 1.3908355795148248,
+      "grad_norm": 0.567251443862915,
+      "learning_rate": 0.0005171203453858607,
+      "loss": 3.7341,
+      "step": 12900
+    },
+    {
+      "epoch": 1.3962264150943398,
+      "grad_norm": 0.5615735054016113,
+      "learning_rate": 0.0005167965461413922,
+      "loss": 3.7331,
+      "step": 12950
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "grad_norm": 0.6066398024559021,
+      "learning_rate": 0.0005164727468969239,
+      "loss": 3.715,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4016172506738545,
+      "eval_accuracy": 0.3512098118773035,
+      "eval_loss": 3.67993426322937,
+      "eval_runtime": 181.5275,
+      "eval_samples_per_second": 99.219,
+      "eval_steps_per_second": 6.203,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4070080862533692,
+      "grad_norm": 0.6397793292999268,
+      "learning_rate": 0.0005161489476524554,
+      "loss": 3.6967,
+      "step": 13050
+    },
+    {
+      "epoch": 1.412398921832884,
+      "grad_norm": 0.6502065062522888,
+      "learning_rate": 0.000515825148407987,
+      "loss": 3.7131,
+      "step": 13100
+    },
+    {
+      "epoch": 1.417789757412399,
+      "grad_norm": 0.5458613038063049,
+      "learning_rate": 0.0005155013491635185,
+      "loss": 3.7147,
+      "step": 13150
+    },
+    {
+      "epoch": 1.4231805929919137,
+      "grad_norm": 0.5654726028442383,
+      "learning_rate": 0.0005151775499190501,
+      "loss": 3.7392,
+      "step": 13200
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.5382741689682007,
+      "learning_rate": 0.0005148537506745818,
+      "loss": 3.6966,
+      "step": 13250
+    },
+    {
+      "epoch": 1.4339622641509435,
+      "grad_norm": 0.5260798931121826,
+      "learning_rate": 0.0005145299514301133,
+      "loss": 3.703,
+      "step": 13300
+    },
+    {
+      "epoch": 1.4393530997304582,
+      "grad_norm": 0.5853448510169983,
+      "learning_rate": 0.0005142061521856449,
+      "loss": 3.684,
+      "step": 13350
+    },
+    {
+      "epoch": 1.444743935309973,
+      "grad_norm": 0.538550615310669,
+      "learning_rate": 0.0005138823529411764,
+      "loss": 3.7219,
+      "step": 13400
+    },
+    {
+      "epoch": 1.4501347708894878,
+      "grad_norm": 0.596662700176239,
+      "learning_rate": 0.0005135585536967081,
+      "loss": 3.7194,
+      "step": 13450
+    },
+    {
+      "epoch": 1.4555256064690028,
+      "grad_norm": 0.6186038851737976,
+      "learning_rate": 0.0005132347544522396,
+      "loss": 3.7033,
+      "step": 13500
+    },
+    {
+      "epoch": 1.4609164420485174,
+      "grad_norm": 0.542559802532196,
+      "learning_rate": 0.0005129109552077712,
+      "loss": 3.7194,
+      "step": 13550
+    },
+    {
+      "epoch": 1.4663072776280324,
+      "grad_norm": 0.5352370738983154,
+      "learning_rate": 0.0005125871559633027,
+      "loss": 3.7046,
+      "step": 13600
+    },
+    {
+      "epoch": 1.4716981132075473,
+      "grad_norm": 0.5471335649490356,
+      "learning_rate": 0.0005122633567188343,
+      "loss": 3.7091,
+      "step": 13650
+    },
+    {
+      "epoch": 1.477088948787062,
+      "grad_norm": 0.6399145126342773,
+      "learning_rate": 0.0005119395574743658,
+      "loss": 3.7208,
+      "step": 13700
+    },
+    {
+      "epoch": 1.482479784366577,
+      "grad_norm": 0.5687584280967712,
+      "learning_rate": 0.0005116222342147868,
+      "loss": 3.7,
+      "step": 13750
+    },
+    {
+      "epoch": 1.4878706199460916,
+      "grad_norm": 0.5827769041061401,
+      "learning_rate": 0.0005112984349703183,
+      "loss": 3.7004,
+      "step": 13800
+    },
+    {
+      "epoch": 1.4932614555256065,
+      "grad_norm": 0.5126985907554626,
+      "learning_rate": 0.00051097463572585,
+      "loss": 3.6955,
+      "step": 13850
+    },
+    {
+      "epoch": 1.4986522911051212,
+      "grad_norm": 0.6048992872238159,
+      "learning_rate": 0.0005106508364813815,
+      "loss": 3.6853,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5040431266846361,
+      "grad_norm": 0.5209577679634094,
+      "learning_rate": 0.0005103270372369131,
+      "loss": 3.7061,
+      "step": 13950
+    },
+    {
+      "epoch": 1.509433962264151,
+      "grad_norm": 0.558080792427063,
+      "learning_rate": 0.0005100032379924446,
+      "loss": 3.6982,
+      "step": 14000
+    },
+    {
+      "epoch": 1.509433962264151,
+      "eval_accuracy": 0.35338667248466665,
+      "eval_loss": 3.6556079387664795,
+      "eval_runtime": 181.0746,
+      "eval_samples_per_second": 99.467,
+      "eval_steps_per_second": 6.218,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5148247978436657,
+      "grad_norm": 0.5223732590675354,
+      "learning_rate": 0.0005096794387479762,
+      "loss": 3.6743,
+      "step": 14050
+    },
+    {
+      "epoch": 1.5202156334231804,
+      "grad_norm": 0.5266443490982056,
+      "learning_rate": 0.0005093556395035078,
+      "loss": 3.7049,
+      "step": 14100
+    },
+    {
+      "epoch": 1.5256064690026954,
+      "grad_norm": 0.5410740971565247,
+      "learning_rate": 0.0005090318402590394,
+      "loss": 3.6871,
+      "step": 14150
+    },
+    {
+      "epoch": 1.5309973045822103,
+      "grad_norm": 0.5198672413825989,
+      "learning_rate": 0.0005087080410145709,
+      "loss": 3.6848,
+      "step": 14200
+    },
+    {
+      "epoch": 1.536388140161725,
+      "grad_norm": 0.5388085246086121,
+      "learning_rate": 0.0005083842417701025,
+      "loss": 3.6925,
+      "step": 14250
+    },
+    {
+      "epoch": 1.54177897574124,
+      "grad_norm": 0.5680968761444092,
+      "learning_rate": 0.000508060442525634,
+      "loss": 3.7015,
+      "step": 14300
+    },
+    {
+      "epoch": 1.5471698113207548,
+      "grad_norm": 0.529816746711731,
+      "learning_rate": 0.0005077366432811656,
+      "loss": 3.7089,
+      "step": 14350
+    },
+    {
+      "epoch": 1.5525606469002695,
+      "grad_norm": 0.588886022567749,
+      "learning_rate": 0.0005074128440366971,
+      "loss": 3.6898,
+      "step": 14400
+    },
+    {
+      "epoch": 1.5579514824797842,
+      "grad_norm": 0.53922438621521,
+      "learning_rate": 0.0005070890447922288,
+      "loss": 3.7101,
+      "step": 14450
+    },
+    {
+      "epoch": 1.5633423180592994,
+      "grad_norm": 0.5846419930458069,
+      "learning_rate": 0.0005067652455477604,
+      "loss": 3.6752,
+      "step": 14500
+    },
+    {
+      "epoch": 1.568733153638814,
+      "grad_norm": 0.5448037981987,
+      "learning_rate": 0.0005064414463032919,
+      "loss": 3.6868,
+      "step": 14550
+    },
+    {
+      "epoch": 1.5741239892183287,
+      "grad_norm": 0.5100300908088684,
+      "learning_rate": 0.0005061176470588235,
+      "loss": 3.6773,
+      "step": 14600
+    },
+    {
+      "epoch": 1.5795148247978437,
+      "grad_norm": 0.5662325620651245,
+      "learning_rate": 0.0005057938478143551,
+      "loss": 3.693,
+      "step": 14650
+    },
+    {
+      "epoch": 1.5849056603773586,
+      "grad_norm": 0.5295085906982422,
+      "learning_rate": 0.0005054700485698867,
+      "loss": 3.6999,
+      "step": 14700
+    },
+    {
+      "epoch": 1.5902964959568733,
+      "grad_norm": 0.6152496337890625,
+      "learning_rate": 0.0005051462493254182,
+      "loss": 3.6639,
+      "step": 14750
+    },
+    {
+      "epoch": 1.595687331536388,
+      "grad_norm": 0.5873838067054749,
+      "learning_rate": 0.0005048224500809498,
+      "loss": 3.6879,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6010781671159031,
+      "grad_norm": 0.5197713375091553,
+      "learning_rate": 0.0005044986508364813,
+      "loss": 3.6784,
+      "step": 14850
+    },
+    {
+      "epoch": 1.6064690026954178,
+      "grad_norm": 0.5193688869476318,
+      "learning_rate": 0.0005041748515920129,
+      "loss": 3.6756,
+      "step": 14900
+    },
+    {
+      "epoch": 1.6118598382749325,
+      "grad_norm": 0.5456000566482544,
+      "learning_rate": 0.0005038510523475444,
+      "loss": 3.6745,
+      "step": 14950
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "grad_norm": 0.565951943397522,
+      "learning_rate": 0.0005035272531030761,
+      "loss": 3.6829,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6172506738544474,
+      "eval_accuracy": 0.3553477484568302,
+      "eval_loss": 3.6376187801361084,
+      "eval_runtime": 181.6027,
+      "eval_samples_per_second": 99.178,
+      "eval_steps_per_second": 6.2,
+      "step": 15000
+    },
+    {
+      "epoch": 1.6226415094339623,
+      "grad_norm": 0.5930358171463013,
+      "learning_rate": 0.0005032034538586076,
+      "loss": 3.6899,
+      "step": 15050
+    },
+    {
+      "epoch": 1.628032345013477,
+      "grad_norm": 0.5067706108093262,
+      "learning_rate": 0.0005028796546141392,
+      "loss": 3.6659,
+      "step": 15100
+    },
+    {
+      "epoch": 1.633423180592992,
+      "grad_norm": 0.5201679468154907,
+      "learning_rate": 0.0005025558553696707,
+      "loss": 3.6864,
+      "step": 15150
+    },
+    {
+      "epoch": 1.6388140161725069,
+      "grad_norm": 0.5754110217094421,
+      "learning_rate": 0.0005022320561252023,
+      "loss": 3.6606,
+      "step": 15200
+    },
+    {
+      "epoch": 1.6442048517520216,
+      "grad_norm": 0.592934250831604,
+      "learning_rate": 0.0005019082568807339,
+      "loss": 3.7035,
+      "step": 15250
+    },
+    {
+      "epoch": 1.6495956873315363,
+      "grad_norm": 0.5429046154022217,
+      "learning_rate": 0.0005015844576362655,
+      "loss": 3.687,
+      "step": 15300
+    },
+    {
+      "epoch": 1.6549865229110512,
+      "grad_norm": 0.5521410703659058,
+      "learning_rate": 0.000501260658391797,
+      "loss": 3.6814,
+      "step": 15350
+    },
+    {
+      "epoch": 1.6603773584905661,
+      "grad_norm": 0.5523940324783325,
+      "learning_rate": 0.0005009368591473286,
+      "loss": 3.671,
+      "step": 15400
+    },
+    {
+      "epoch": 1.6657681940700808,
+      "grad_norm": 0.5145202875137329,
+      "learning_rate": 0.0005006130599028602,
+      "loss": 3.6823,
+      "step": 15450
+    },
+    {
+      "epoch": 1.6711590296495957,
+      "grad_norm": 0.541299045085907,
+      "learning_rate": 0.0005002892606583918,
+      "loss": 3.6941,
+      "step": 15500
+    },
+    {
+      "epoch": 1.6765498652291106,
+      "grad_norm": 0.578633189201355,
+      "learning_rate": 0.0004999654614139233,
+      "loss": 3.6671,
+      "step": 15550
+    },
+    {
+      "epoch": 1.6819407008086253,
+      "grad_norm": 0.5630112290382385,
+      "learning_rate": 0.0004996416621694549,
+      "loss": 3.6661,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68733153638814,
+      "grad_norm": 0.546072244644165,
+      "learning_rate": 0.0004993178629249864,
+      "loss": 3.663,
+      "step": 15650
+    },
+    {
+      "epoch": 1.692722371967655,
+      "grad_norm": 0.6002097725868225,
+      "learning_rate": 0.000498994063680518,
+      "loss": 3.6783,
+      "step": 15700
+    },
+    {
+      "epoch": 1.6981132075471699,
+      "grad_norm": 0.6249855756759644,
+      "learning_rate": 0.000498676740420939,
+      "loss": 3.6534,
+      "step": 15750
+    },
+    {
+      "epoch": 1.7035040431266846,
+      "grad_norm": 0.5264080166816711,
+      "learning_rate": 0.0004983529411764705,
+      "loss": 3.6742,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7088948787061995,
+      "grad_norm": 0.5488578081130981,
+      "learning_rate": 0.0004980291419320022,
+      "loss": 3.6645,
+      "step": 15850
+    },
+    {
+      "epoch": 1.7142857142857144,
+      "grad_norm": 0.5747875571250916,
+      "learning_rate": 0.0004977053426875337,
+      "loss": 3.6661,
+      "step": 15900
+    },
+    {
+      "epoch": 1.719676549865229,
+      "grad_norm": 0.5004916191101074,
+      "learning_rate": 0.0004973815434430653,
+      "loss": 3.6556,
+      "step": 15950
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "grad_norm": 0.5628464221954346,
+      "learning_rate": 0.0004970577441985968,
+      "loss": 3.6554,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7250673854447438,
+      "eval_accuracy": 0.3569113721653684,
+      "eval_loss": 3.6237761974334717,
+      "eval_runtime": 181.1991,
+      "eval_samples_per_second": 99.399,
+      "eval_steps_per_second": 6.214,
+      "step": 16000
+    },
+    {
+      "epoch": 1.7304582210242587,
+      "grad_norm": 0.5369266271591187,
+      "learning_rate": 0.0004967339449541284,
+      "loss": 3.6758,
+      "step": 16050
+    },
+    {
+      "epoch": 1.7358490566037736,
+      "grad_norm": 0.575958788394928,
+      "learning_rate": 0.00049641014570966,
+      "loss": 3.6766,
+      "step": 16100
+    },
+    {
+      "epoch": 1.7412398921832883,
+      "grad_norm": 0.6171749234199524,
+      "learning_rate": 0.0004960863464651916,
+      "loss": 3.658,
+      "step": 16150
+    },
+    {
+      "epoch": 1.7466307277628033,
+      "grad_norm": 0.5248631238937378,
+      "learning_rate": 0.0004957625472207231,
+      "loss": 3.6652,
+      "step": 16200
+    },
+    {
+      "epoch": 1.7520215633423182,
+      "grad_norm": 0.5663668513298035,
+      "learning_rate": 0.0004954387479762547,
+      "loss": 3.6645,
+      "step": 16250
+    },
+    {
+      "epoch": 1.7574123989218329,
+      "grad_norm": 0.5405056476593018,
+      "learning_rate": 0.0004951149487317862,
+      "loss": 3.6633,
+      "step": 16300
+    },
+    {
+      "epoch": 1.7628032345013476,
+      "grad_norm": 0.5469680428504944,
+      "learning_rate": 0.0004947911494873178,
+      "loss": 3.6642,
+      "step": 16350
+    },
+    {
+      "epoch": 1.7681940700808625,
+      "grad_norm": 0.5379207134246826,
+      "learning_rate": 0.0004944673502428493,
+      "loss": 3.6666,
+      "step": 16400
+    },
+    {
+      "epoch": 1.7735849056603774,
+      "grad_norm": 0.5439175963401794,
+      "learning_rate": 0.000494143550998381,
+      "loss": 3.6703,
+      "step": 16450
+    },
+    {
+      "epoch": 1.778975741239892,
+      "grad_norm": 0.5585680603981018,
+      "learning_rate": 0.0004938197517539125,
+      "loss": 3.666,
+      "step": 16500
+    },
+    {
+      "epoch": 1.784366576819407,
+      "grad_norm": 0.5458288788795471,
+      "learning_rate": 0.0004934959525094441,
+      "loss": 3.6453,
+      "step": 16550
+    },
+    {
+      "epoch": 1.789757412398922,
+      "grad_norm": 0.5485697388648987,
+      "learning_rate": 0.0004931721532649756,
+      "loss": 3.6609,
+      "step": 16600
+    },
+    {
+      "epoch": 1.7951482479784366,
+      "grad_norm": 0.5749149918556213,
+      "learning_rate": 0.0004928483540205073,
+      "loss": 3.6645,
+      "step": 16650
+    },
+    {
+      "epoch": 1.8005390835579513,
+      "grad_norm": 0.5598737001419067,
+      "learning_rate": 0.0004925245547760388,
+      "loss": 3.6582,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8059299191374663,
+      "grad_norm": 0.5720735788345337,
+      "learning_rate": 0.0004922007555315704,
+      "loss": 3.6479,
+      "step": 16750
+    },
+    {
+      "epoch": 1.8113207547169812,
+      "grad_norm": 0.544169008731842,
+      "learning_rate": 0.000491876956287102,
+      "loss": 3.6539,
+      "step": 16800
+    },
+    {
+      "epoch": 1.8167115902964959,
+      "grad_norm": 0.587104082107544,
+      "learning_rate": 0.0004915531570426335,
+      "loss": 3.6497,
+      "step": 16850
+    },
+    {
+      "epoch": 1.8221024258760108,
+      "grad_norm": 0.556251585483551,
+      "learning_rate": 0.0004912293577981652,
+      "loss": 3.6587,
+      "step": 16900
+    },
+    {
+      "epoch": 1.8274932614555257,
+      "grad_norm": 0.5477184057235718,
+      "learning_rate": 0.0004909055585536966,
+      "loss": 3.6421,
+      "step": 16950
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "grad_norm": 0.5440017580986023,
+      "learning_rate": 0.0004905817593092283,
+      "loss": 3.6345,
+      "step": 17000
+    },
+    {
+      "epoch": 1.8328840970350404,
+      "eval_accuracy": 0.3582564949165117,
+      "eval_loss": 3.606245994567871,
+      "eval_runtime": 181.6324,
+      "eval_samples_per_second": 99.162,
+      "eval_steps_per_second": 6.199,
+      "step": 17000
+    },
+    {
+      "epoch": 1.838274932614555,
+      "grad_norm": 0.5022566318511963,
+      "learning_rate": 0.0004902579600647598,
+      "loss": 3.6459,
+      "step": 17050
+    },
+    {
+      "epoch": 1.8436657681940702,
+      "grad_norm": 0.572801947593689,
+      "learning_rate": 0.0004899341608202914,
+      "loss": 3.646,
+      "step": 17100
+    },
+    {
+      "epoch": 1.849056603773585,
+      "grad_norm": 0.5857950448989868,
+      "learning_rate": 0.0004896103615758229,
+      "loss": 3.6343,
+      "step": 17150
+    },
+    {
+      "epoch": 1.8544474393530996,
+      "grad_norm": 0.529326856136322,
+      "learning_rate": 0.0004892865623313546,
+      "loss": 3.6621,
+      "step": 17200
+    },
+    {
+      "epoch": 1.8598382749326146,
+      "grad_norm": 0.5415377616882324,
+      "learning_rate": 0.0004889627630868861,
+      "loss": 3.6692,
+      "step": 17250
+    },
+    {
+      "epoch": 1.8652291105121295,
+      "grad_norm": 0.5492255091667175,
+      "learning_rate": 0.0004886389638424177,
+      "loss": 3.6534,
+      "step": 17300
+    },
+    {
+      "epoch": 1.8706199460916442,
+      "grad_norm": 0.5427126884460449,
+      "learning_rate": 0.0004883151645979492,
+      "loss": 3.6425,
+      "step": 17350
+    },
+    {
+      "epoch": 1.8760107816711589,
+      "grad_norm": 0.561755359172821,
+      "learning_rate": 0.00048799136535348077,
+      "loss": 3.6583,
+      "step": 17400
+    },
+    {
+      "epoch": 1.881401617250674,
+      "grad_norm": 0.5410698056221008,
+      "learning_rate": 0.0004876675661090124,
+      "loss": 3.6306,
+      "step": 17450
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.542617678642273,
+      "learning_rate": 0.0004873437668645439,
+      "loss": 3.6534,
+      "step": 17500
+    },
+    {
+      "epoch": 1.8921832884097034,
+      "grad_norm": 0.5104948878288269,
+      "learning_rate": 0.00048701996762007553,
+      "loss": 3.6543,
+      "step": 17550
+    },
+    {
+      "epoch": 1.8975741239892183,
+      "grad_norm": 0.6261042356491089,
+      "learning_rate": 0.0004866961683756071,
+      "loss": 3.6425,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9029649595687332,
+      "grad_norm": 0.5958898067474365,
+      "learning_rate": 0.0004863723691311387,
+      "loss": 3.6418,
+      "step": 17650
+    },
+    {
+      "epoch": 1.908355795148248,
+      "grad_norm": 0.5822218656539917,
+      "learning_rate": 0.00048604856988667024,
+      "loss": 3.6586,
+      "step": 17700
+    },
+    {
+      "epoch": 1.9137466307277629,
+      "grad_norm": 0.5634023547172546,
+      "learning_rate": 0.0004857247706422018,
+      "loss": 3.6335,
+      "step": 17750
+    },
+    {
+      "epoch": 1.9191374663072778,
+      "grad_norm": 0.5490560531616211,
+      "learning_rate": 0.00048540744738262274,
+      "loss": 3.6368,
+      "step": 17800
+    },
+    {
+      "epoch": 1.9245283018867925,
+      "grad_norm": 0.5121088027954102,
+      "learning_rate": 0.00048508364813815434,
+      "loss": 3.6266,
+      "step": 17850
+    },
+    {
+      "epoch": 1.9299191374663072,
+      "grad_norm": 0.5225231647491455,
+      "learning_rate": 0.00048475984889368584,
+      "loss": 3.6324,
+      "step": 17900
+    },
+    {
+      "epoch": 1.935309973045822,
+      "grad_norm": 0.5291708111763,
+      "learning_rate": 0.0004844360496492175,
+      "loss": 3.6424,
+      "step": 17950
+    },
+    {
+      "epoch": 1.940700808625337,
+      "grad_norm": 0.5437564849853516,
+      "learning_rate": 0.000484112250404749,
+      "loss": 3.6469,
+      "step": 18000
+    },
+    {
+      "epoch": 1.940700808625337,
+      "eval_accuracy": 0.3598998698447057,
+      "eval_loss": 3.5913586616516113,
+      "eval_runtime": 181.4123,
+      "eval_samples_per_second": 99.282,
+      "eval_steps_per_second": 6.207,
+      "step": 18000
+    },
+    {
+      "epoch": 1.9460916442048517,
+      "grad_norm": 0.5686632394790649,
+      "learning_rate": 0.00048378845116028055,
+      "loss": 3.6235,
+      "step": 18050
+    },
+    {
+      "epoch": 1.9514824797843666,
+      "grad_norm": 0.5239707231521606,
+      "learning_rate": 0.00048346465191581215,
+      "loss": 3.6487,
+      "step": 18100
+    },
+    {
+      "epoch": 1.9568733153638815,
+      "grad_norm": 0.5632966756820679,
+      "learning_rate": 0.0004831408526713437,
+      "loss": 3.642,
+      "step": 18150
+    },
+    {
+      "epoch": 1.9622641509433962,
+      "grad_norm": 0.5901452898979187,
+      "learning_rate": 0.0004828170534268753,
+      "loss": 3.6363,
+      "step": 18200
+    },
+    {
+      "epoch": 1.967654986522911,
+      "grad_norm": 0.5481223464012146,
+      "learning_rate": 0.00048249325418240686,
+      "loss": 3.6498,
+      "step": 18250
+    },
+    {
+      "epoch": 1.9730458221024259,
+      "grad_norm": 0.5510320067405701,
+      "learning_rate": 0.00048216945493793846,
+      "loss": 3.6471,
+      "step": 18300
+    },
+    {
+      "epoch": 1.9784366576819408,
+      "grad_norm": 0.5957186222076416,
+      "learning_rate": 0.00048184565569347,
+      "loss": 3.6295,
+      "step": 18350
+    },
+    {
+      "epoch": 1.9838274932614555,
+      "grad_norm": 0.5391411781311035,
+      "learning_rate": 0.0004815218564490016,
+      "loss": 3.6625,
+      "step": 18400
+    },
+    {
+      "epoch": 1.9892183288409704,
+      "grad_norm": 0.5556368231773376,
+      "learning_rate": 0.00048119805720453317,
+      "loss": 3.6166,
+      "step": 18450
+    },
+    {
+      "epoch": 1.9946091644204853,
+      "grad_norm": 0.6190194487571716,
+      "learning_rate": 0.0004808742579600647,
+      "loss": 3.6231,
+      "step": 18500
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.197678804397583,
+      "learning_rate": 0.0004805504587155963,
+      "loss": 3.6412,
+      "step": 18550
+    },
+    {
+      "epoch": 2.0053908355795147,
+      "grad_norm": 0.5590324401855469,
+      "learning_rate": 0.0004802266594711278,
+      "loss": 3.5499,
+      "step": 18600
+    },
+    {
+      "epoch": 2.01078167115903,
+      "grad_norm": 0.5855118632316589,
+      "learning_rate": 0.0004799028602266594,
+      "loss": 3.5515,
+      "step": 18650
+    },
+    {
+      "epoch": 2.0161725067385445,
+      "grad_norm": 0.5683425664901733,
+      "learning_rate": 0.000479579060982191,
+      "loss": 3.5434,
+      "step": 18700
+    },
+    {
+      "epoch": 2.0215633423180592,
+      "grad_norm": 0.5408341884613037,
+      "learning_rate": 0.0004792552617377226,
+      "loss": 3.5613,
+      "step": 18750
+    },
+    {
+      "epoch": 2.026954177897574,
+      "grad_norm": 0.5905522108078003,
+      "learning_rate": 0.00047893146249325413,
+      "loss": 3.5492,
+      "step": 18800
+    },
+    {
+      "epoch": 2.032345013477089,
+      "grad_norm": 0.5483571887016296,
+      "learning_rate": 0.0004786076632487857,
+      "loss": 3.5309,
+      "step": 18850
+    },
+    {
+      "epoch": 2.0377358490566038,
+      "grad_norm": 0.5604623556137085,
+      "learning_rate": 0.0004782838640043173,
+      "loss": 3.5583,
+      "step": 18900
+    },
+    {
+      "epoch": 2.0431266846361185,
+      "grad_norm": 0.5344943404197693,
+      "learning_rate": 0.00047796006475984883,
+      "loss": 3.5482,
+      "step": 18950
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "grad_norm": 0.5930701494216919,
+      "learning_rate": 0.00047763626551538044,
+      "loss": 3.5639,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0485175202156336,
+      "eval_accuracy": 0.36086101329031256,
+      "eval_loss": 3.584812641143799,
+      "eval_runtime": 181.1798,
+      "eval_samples_per_second": 99.41,
+      "eval_steps_per_second": 6.215,
+      "step": 19000
+    },
+    {
+      "epoch": 2.0539083557951483,
+      "grad_norm": 0.597104012966156,
+      "learning_rate": 0.000477312466270912,
+      "loss": 3.5241,
+      "step": 19050
+    },
+    {
+      "epoch": 2.059299191374663,
+      "grad_norm": 0.5604715943336487,
+      "learning_rate": 0.0004769886670264436,
+      "loss": 3.5646,
+      "step": 19100
+    },
+    {
+      "epoch": 2.0646900269541777,
+      "grad_norm": 0.5276287198066711,
+      "learning_rate": 0.00047666486778197515,
+      "loss": 3.549,
+      "step": 19150
+    },
+    {
+      "epoch": 2.070080862533693,
+      "grad_norm": 0.5561823844909668,
+      "learning_rate": 0.00047634106853750675,
+      "loss": 3.5711,
+      "step": 19200
+    },
+    {
+      "epoch": 2.0754716981132075,
+      "grad_norm": 0.6093335151672363,
+      "learning_rate": 0.00047601726929303825,
+      "loss": 3.5684,
+      "step": 19250
+    },
+    {
+      "epoch": 2.0808625336927222,
+      "grad_norm": 0.588876485824585,
+      "learning_rate": 0.0004756934700485698,
+      "loss": 3.5646,
+      "step": 19300
+    },
+    {
+      "epoch": 2.0862533692722374,
+      "grad_norm": 0.5568276643753052,
+      "learning_rate": 0.0004753696708041014,
+      "loss": 3.5551,
+      "step": 19350
+    },
+    {
+      "epoch": 2.091644204851752,
+      "grad_norm": 0.5360884666442871,
+      "learning_rate": 0.00047504587155963295,
+      "loss": 3.5605,
+      "step": 19400
+    },
+    {
+      "epoch": 2.0970350404312668,
+      "grad_norm": 0.5348809361457825,
+      "learning_rate": 0.00047472207231516456,
+      "loss": 3.5486,
+      "step": 19450
+    },
+    {
+      "epoch": 2.1024258760107815,
+      "grad_norm": 0.6352576017379761,
+      "learning_rate": 0.0004743982730706961,
+      "loss": 3.5433,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1078167115902966,
+      "grad_norm": 0.570564329624176,
+      "learning_rate": 0.0004740744738262277,
+      "loss": 3.5608,
+      "step": 19550
+    },
+    {
+      "epoch": 2.1132075471698113,
+      "grad_norm": 0.5465729832649231,
+      "learning_rate": 0.00047375067458175926,
+      "loss": 3.5578,
+      "step": 19600
+    },
+    {
+      "epoch": 2.118598382749326,
+      "grad_norm": 0.5874576568603516,
+      "learning_rate": 0.00047342687533729087,
+      "loss": 3.5579,
+      "step": 19650
+    },
+    {
+      "epoch": 2.123989218328841,
+      "grad_norm": 0.59136563539505,
+      "learning_rate": 0.0004731030760928224,
+      "loss": 3.5486,
+      "step": 19700
+    },
+    {
+      "epoch": 2.129380053908356,
+      "grad_norm": 0.6302415728569031,
+      "learning_rate": 0.00047277927684835397,
+      "loss": 3.5422,
+      "step": 19750
+    },
+    {
+      "epoch": 2.1347708894878705,
+      "grad_norm": 0.544438898563385,
+      "learning_rate": 0.0004724619535887749,
+      "loss": 3.5742,
+      "step": 19800
+    },
+    {
+      "epoch": 2.1401617250673857,
+      "grad_norm": 0.6034201383590698,
+      "learning_rate": 0.0004721381543443065,
+      "loss": 3.5728,
+      "step": 19850
+    },
+    {
+      "epoch": 2.1455525606469004,
+      "grad_norm": 0.6693742871284485,
+      "learning_rate": 0.000471814355099838,
+      "loss": 3.5634,
+      "step": 19900
+    },
+    {
+      "epoch": 2.150943396226415,
+      "grad_norm": 0.5606766939163208,
+      "learning_rate": 0.0004714905558553697,
+      "loss": 3.5611,
+      "step": 19950
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "grad_norm": 0.600206196308136,
+      "learning_rate": 0.0004711667566109012,
+      "loss": 3.5606,
+      "step": 20000
+    },
+    {
+      "epoch": 2.1563342318059298,
+      "eval_accuracy": 0.36217603919153124,
+      "eval_loss": 3.5728728771209717,
+      "eval_runtime": 181.6212,
+      "eval_samples_per_second": 99.168,
+      "eval_steps_per_second": 6.2,
+      "step": 20000
+    },
+    {
+      "epoch": 2.161725067385445,
+      "grad_norm": 0.5841259956359863,
+      "learning_rate": 0.00047084295736643273,
+      "loss": 3.5593,
+      "step": 20050
+    },
+    {
+      "epoch": 2.1671159029649596,
+      "grad_norm": 0.6095578670501709,
+      "learning_rate": 0.00047051915812196433,
+      "loss": 3.5696,
+      "step": 20100
+    },
+    {
+      "epoch": 2.1725067385444743,
+      "grad_norm": 0.5807099938392639,
+      "learning_rate": 0.0004701953588774959,
+      "loss": 3.546,
+      "step": 20150
+    },
+    {
+      "epoch": 2.177897574123989,
+      "grad_norm": 0.5789870023727417,
+      "learning_rate": 0.0004698715596330275,
+      "loss": 3.5722,
+      "step": 20200
+    },
+    {
+      "epoch": 2.183288409703504,
+      "grad_norm": 0.5594269037246704,
+      "learning_rate": 0.00046954776038855904,
+      "loss": 3.5715,
+      "step": 20250
+    },
+    {
+      "epoch": 2.188679245283019,
+      "grad_norm": 0.643621027469635,
+      "learning_rate": 0.00046922396114409064,
+      "loss": 3.5454,
+      "step": 20300
+    },
+    {
+      "epoch": 2.1940700808625335,
+      "grad_norm": 0.5453854203224182,
+      "learning_rate": 0.0004689001618996222,
+      "loss": 3.5642,
+      "step": 20350
+    },
+    {
+      "epoch": 2.1994609164420487,
+      "grad_norm": 0.5129548907279968,
+      "learning_rate": 0.0004685763626551538,
+      "loss": 3.5621,
+      "step": 20400
+    },
+    {
+      "epoch": 2.2048517520215634,
+      "grad_norm": 0.6147149205207825,
+      "learning_rate": 0.00046825256341068535,
+      "loss": 3.5631,
+      "step": 20450
+    },
+    {
+      "epoch": 2.210242587601078,
+      "grad_norm": 0.5020909905433655,
+      "learning_rate": 0.0004679287641662169,
+      "loss": 3.5656,
+      "step": 20500
+    },
+    {
+      "epoch": 2.215633423180593,
+      "grad_norm": 0.6170671582221985,
+      "learning_rate": 0.0004676049649217485,
+      "loss": 3.5674,
+      "step": 20550
+    },
+    {
+      "epoch": 2.221024258760108,
+      "grad_norm": 0.5955941081047058,
+      "learning_rate": 0.00046728764166216946,
+      "loss": 3.5746,
+      "step": 20600
+    },
+    {
+      "epoch": 2.2264150943396226,
+      "grad_norm": 0.5670434832572937,
+      "learning_rate": 0.00046696384241770095,
+      "loss": 3.5567,
+      "step": 20650
+    },
+    {
+      "epoch": 2.2318059299191373,
+      "grad_norm": 0.570024311542511,
+      "learning_rate": 0.00046664004317323256,
+      "loss": 3.5609,
+      "step": 20700
+    },
+    {
+      "epoch": 2.2371967654986524,
+      "grad_norm": 0.5753904581069946,
+      "learning_rate": 0.0004663162439287641,
+      "loss": 3.563,
+      "step": 20750
+    },
+    {
+      "epoch": 2.242587601078167,
+      "grad_norm": 0.5676912665367126,
+      "learning_rate": 0.00046599244468429566,
+      "loss": 3.5655,
+      "step": 20800
+    },
+    {
+      "epoch": 2.247978436657682,
+      "grad_norm": 0.5626962780952454,
+      "learning_rate": 0.00046566864543982726,
+      "loss": 3.5484,
+      "step": 20850
+    },
+    {
+      "epoch": 2.2533692722371965,
+      "grad_norm": 0.6296677589416504,
+      "learning_rate": 0.0004653448461953588,
+      "loss": 3.5597,
+      "step": 20900
+    },
+    {
+      "epoch": 2.2587601078167117,
+      "grad_norm": 0.59935063123703,
+      "learning_rate": 0.0004650210469508904,
+      "loss": 3.5682,
+      "step": 20950
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "grad_norm": 0.5628940463066101,
+      "learning_rate": 0.00046469724770642197,
+      "loss": 3.5497,
+      "step": 21000
+    },
+    {
+      "epoch": 2.2641509433962264,
+      "eval_accuracy": 0.3636019996912085,
+      "eval_loss": 3.5614914894104004,
+      "eval_runtime": 180.9956,
+      "eval_samples_per_second": 99.511,
+      "eval_steps_per_second": 6.221,
+      "step": 21000
+    },
+    {
+      "epoch": 2.269541778975741,
+      "grad_norm": 0.5957950353622437,
+      "learning_rate": 0.0004643734484619536,
+      "loss": 3.5417,
+      "step": 21050
+    },
+    {
+      "epoch": 2.274932614555256,
+      "grad_norm": 0.5501925349235535,
+      "learning_rate": 0.0004640496492174851,
+      "loss": 3.5728,
+      "step": 21100
+    },
+    {
+      "epoch": 2.280323450134771,
+      "grad_norm": 0.5697695016860962,
+      "learning_rate": 0.00046372584997301673,
+      "loss": 3.5687,
+      "step": 21150
+    },
+    {
+      "epoch": 2.2857142857142856,
+      "grad_norm": 0.5633260607719421,
+      "learning_rate": 0.0004634020507285483,
+      "loss": 3.5589,
+      "step": 21200
+    },
+    {
+      "epoch": 2.2911051212938007,
+      "grad_norm": 0.5334222316741943,
+      "learning_rate": 0.0004630782514840798,
+      "loss": 3.5327,
+      "step": 21250
+    },
+    {
+      "epoch": 2.2964959568733154,
+      "grad_norm": 0.5983673334121704,
+      "learning_rate": 0.0004627544522396114,
+      "loss": 3.5524,
+      "step": 21300
+    },
+    {
+      "epoch": 2.30188679245283,
+      "grad_norm": 0.5501682758331299,
+      "learning_rate": 0.00046243065299514293,
+      "loss": 3.5701,
+      "step": 21350
+    },
+    {
+      "epoch": 2.3072776280323453,
+      "grad_norm": 0.5882470011711121,
+      "learning_rate": 0.00046210685375067454,
+      "loss": 3.546,
+      "step": 21400
+    },
+    {
+      "epoch": 2.31266846361186,
+      "grad_norm": 0.5191405415534973,
+      "learning_rate": 0.0004617830545062061,
+      "loss": 3.5579,
+      "step": 21450
+    },
+    {
+      "epoch": 2.3180592991913747,
+      "grad_norm": 0.5627714395523071,
+      "learning_rate": 0.0004614592552617377,
+      "loss": 3.5624,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3234501347708894,
+      "grad_norm": 0.578151285648346,
+      "learning_rate": 0.00046113545601726924,
+      "loss": 3.5502,
+      "step": 21550
+    },
+    {
+      "epoch": 2.3288409703504045,
+      "grad_norm": 0.5719261169433594,
+      "learning_rate": 0.00046081165677280085,
+      "loss": 3.5608,
+      "step": 21600
+    },
+    {
+      "epoch": 2.334231805929919,
+      "grad_norm": 0.5572363138198853,
+      "learning_rate": 0.0004604878575283324,
+      "loss": 3.5599,
+      "step": 21650
+    },
+    {
+      "epoch": 2.339622641509434,
+      "grad_norm": 0.6042853593826294,
+      "learning_rate": 0.00046016405828386395,
+      "loss": 3.561,
+      "step": 21700
+    },
+    {
+      "epoch": 2.3450134770889486,
+      "grad_norm": 0.5274094343185425,
+      "learning_rate": 0.00045984025903939555,
+      "loss": 3.5605,
+      "step": 21750
+    },
+    {
+      "epoch": 2.3504043126684637,
+      "grad_norm": 0.5416309237480164,
+      "learning_rate": 0.0004595164597949271,
+      "loss": 3.5678,
+      "step": 21800
+    },
+    {
+      "epoch": 2.3557951482479784,
+      "grad_norm": 0.5653731822967529,
+      "learning_rate": 0.0004591926605504587,
+      "loss": 3.5622,
+      "step": 21850
+    },
+    {
+      "epoch": 2.361185983827493,
+      "grad_norm": 0.5564022064208984,
+      "learning_rate": 0.0004588688613059902,
+      "loss": 3.5496,
+      "step": 21900
+    },
+    {
+      "epoch": 2.3665768194070083,
+      "grad_norm": 0.5706598162651062,
+      "learning_rate": 0.00045854506206152186,
+      "loss": 3.5628,
+      "step": 21950
+    },
+    {
+      "epoch": 2.371967654986523,
+      "grad_norm": 0.5681759715080261,
+      "learning_rate": 0.00045822126281705336,
+      "loss": 3.5624,
+      "step": 22000
+    },
+    {
+      "epoch": 2.371967654986523,
+      "eval_accuracy": 0.36438560431812655,
+      "eval_loss": 3.550971269607544,
+      "eval_runtime": 181.5333,
+      "eval_samples_per_second": 99.216,
+      "eval_steps_per_second": 6.203,
+      "step": 22000
+    },
+    {
+      "epoch": 2.3773584905660377,
+      "grad_norm": 0.6144738793373108,
+      "learning_rate": 0.00045789746357258497,
+      "loss": 3.538,
+      "step": 22050
+    },
+    {
+      "epoch": 2.382749326145553,
+      "grad_norm": 0.5621081590652466,
+      "learning_rate": 0.0004575736643281165,
+      "loss": 3.5622,
+      "step": 22100
+    },
+    {
+      "epoch": 2.3881401617250675,
+      "grad_norm": 0.5389125943183899,
+      "learning_rate": 0.00045724986508364807,
+      "loss": 3.5571,
+      "step": 22150
+    },
+    {
+      "epoch": 2.393530997304582,
+      "grad_norm": 0.6281595230102539,
+      "learning_rate": 0.00045692606583917967,
+      "loss": 3.5816,
+      "step": 22200
+    },
+    {
+      "epoch": 2.398921832884097,
+      "grad_norm": 0.573566734790802,
+      "learning_rate": 0.0004566022665947112,
+      "loss": 3.551,
+      "step": 22250
+    },
+    {
+      "epoch": 2.404312668463612,
+      "grad_norm": 0.6374450922012329,
+      "learning_rate": 0.0004562784673502428,
+      "loss": 3.5344,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4097035040431267,
+      "grad_norm": 0.5555562376976013,
+      "learning_rate": 0.0004559546681057744,
+      "loss": 3.5563,
+      "step": 22350
+    },
+    {
+      "epoch": 2.4150943396226414,
+      "grad_norm": 0.5741384029388428,
+      "learning_rate": 0.000455630868861306,
+      "loss": 3.5455,
+      "step": 22400
+    },
+    {
+      "epoch": 2.420485175202156,
+      "grad_norm": 0.6011936664581299,
+      "learning_rate": 0.00045530706961683753,
+      "loss": 3.5751,
+      "step": 22450
+    },
+    {
+      "epoch": 2.4258760107816713,
+      "grad_norm": 0.6304056644439697,
+      "learning_rate": 0.00045498327037236914,
+      "loss": 3.5478,
+      "step": 22500
+    },
+    {
+      "epoch": 2.431266846361186,
+      "grad_norm": 0.5553408861160278,
+      "learning_rate": 0.0004546594711279007,
+      "loss": 3.5519,
+      "step": 22550
+    },
+    {
+      "epoch": 2.4366576819407006,
+      "grad_norm": 0.514491081237793,
+      "learning_rate": 0.0004543356718834322,
+      "loss": 3.5529,
+      "step": 22600
+    },
+    {
+      "epoch": 2.442048517520216,
+      "grad_norm": 0.5898559093475342,
+      "learning_rate": 0.0004540118726389638,
+      "loss": 3.5497,
+      "step": 22650
+    },
+    {
+      "epoch": 2.4474393530997305,
+      "grad_norm": 0.5183122158050537,
+      "learning_rate": 0.00045368807339449534,
+      "loss": 3.539,
+      "step": 22700
+    },
+    {
+      "epoch": 2.452830188679245,
+      "grad_norm": 0.588056206703186,
+      "learning_rate": 0.00045336427415002694,
+      "loss": 3.5642,
+      "step": 22750
+    },
+    {
+      "epoch": 2.4582210242587603,
+      "grad_norm": 0.5579840540885925,
+      "learning_rate": 0.0004530404749055585,
+      "loss": 3.5741,
+      "step": 22800
+    },
+    {
+      "epoch": 2.463611859838275,
+      "grad_norm": 0.6022590398788452,
+      "learning_rate": 0.0004527166756610901,
+      "loss": 3.5405,
+      "step": 22850
+    },
+    {
+      "epoch": 2.4690026954177897,
+      "grad_norm": 0.5468297600746155,
+      "learning_rate": 0.00045239287641662165,
+      "loss": 3.5504,
+      "step": 22900
+    },
+    {
+      "epoch": 2.4743935309973044,
+      "grad_norm": 0.576793372631073,
+      "learning_rate": 0.0004520690771721532,
+      "loss": 3.5537,
+      "step": 22950
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "grad_norm": 0.5995750427246094,
+      "learning_rate": 0.0004517452779276848,
+      "loss": 3.5459,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4797843665768196,
+      "eval_accuracy": 0.3654815859975112,
+      "eval_loss": 3.5396082401275635,
+      "eval_runtime": 180.9737,
+      "eval_samples_per_second": 99.523,
+      "eval_steps_per_second": 6.222,
+      "step": 23000
+    },
+    {
+      "epoch": 2.4851752021563343,
+      "grad_norm": 0.6596251130104065,
+      "learning_rate": 0.00045142147868321636,
+      "loss": 3.5446,
+      "step": 23050
+    },
+    {
+      "epoch": 2.490566037735849,
+      "grad_norm": 0.5252313017845154,
+      "learning_rate": 0.00045109767943874796,
+      "loss": 3.5398,
+      "step": 23100
+    },
+    {
+      "epoch": 2.4959568733153636,
+      "grad_norm": 0.6100819706916809,
+      "learning_rate": 0.0004507738801942795,
+      "loss": 3.5374,
+      "step": 23150
+    },
+    {
+      "epoch": 2.501347708894879,
+      "grad_norm": 0.5994240045547485,
+      "learning_rate": 0.0004504500809498111,
+      "loss": 3.5518,
+      "step": 23200
+    },
+    {
+      "epoch": 2.5067385444743935,
+      "grad_norm": 0.5911756753921509,
+      "learning_rate": 0.0004501262817053426,
+      "loss": 3.5651,
+      "step": 23250
+    },
+    {
+      "epoch": 2.512129380053908,
+      "grad_norm": 0.5529770851135254,
+      "learning_rate": 0.00044980248246087427,
+      "loss": 3.5623,
+      "step": 23300
+    },
+    {
+      "epoch": 2.5175202156334233,
+      "grad_norm": 0.5851297378540039,
+      "learning_rate": 0.00044947868321640577,
+      "loss": 3.5372,
+      "step": 23350
+    },
+    {
+      "epoch": 2.522911051212938,
+      "grad_norm": 0.5575948357582092,
+      "learning_rate": 0.0004491548839719373,
+      "loss": 3.5416,
+      "step": 23400
+    },
+    {
+      "epoch": 2.5283018867924527,
+      "grad_norm": 0.5790318250656128,
+      "learning_rate": 0.0004488310847274689,
+      "loss": 3.5464,
+      "step": 23450
+    },
+    {
+      "epoch": 2.533692722371968,
+      "grad_norm": 0.5353867411613464,
+      "learning_rate": 0.0004485072854830005,
+      "loss": 3.5237,
+      "step": 23500
+    },
+    {
+      "epoch": 2.5390835579514826,
+      "grad_norm": 0.6083977818489075,
+      "learning_rate": 0.0004481834862385321,
+      "loss": 3.5601,
+      "step": 23550
+    },
+    {
+      "epoch": 2.5444743935309972,
+      "grad_norm": 0.5814011096954346,
+      "learning_rate": 0.00044785968699406363,
+      "loss": 3.5533,
+      "step": 23600
+    },
+    {
+      "epoch": 2.5498652291105124,
+      "grad_norm": 0.5686614513397217,
+      "learning_rate": 0.00044753588774959523,
+      "loss": 3.5471,
+      "step": 23650
+    },
+    {
+      "epoch": 2.555256064690027,
+      "grad_norm": 0.6276265978813171,
+      "learning_rate": 0.0004472120885051268,
+      "loss": 3.5638,
+      "step": 23700
+    },
+    {
+      "epoch": 2.560646900269542,
+      "grad_norm": 0.5724161863327026,
+      "learning_rate": 0.0004468882892606584,
+      "loss": 3.5684,
+      "step": 23750
+    },
+    {
+      "epoch": 2.5660377358490565,
+      "grad_norm": 0.6355558633804321,
+      "learning_rate": 0.00044656449001618994,
+      "loss": 3.5441,
+      "step": 23800
+    },
+    {
+      "epoch": 2.571428571428571,
+      "grad_norm": 0.6126329898834229,
+      "learning_rate": 0.0004462406907717215,
+      "loss": 3.5433,
+      "step": 23850
+    },
+    {
+      "epoch": 2.5768194070080863,
+      "grad_norm": 0.6300326585769653,
+      "learning_rate": 0.0004459168915272531,
+      "loss": 3.5409,
+      "step": 23900
+    },
+    {
+      "epoch": 2.582210242587601,
+      "grad_norm": 0.5199249386787415,
+      "learning_rate": 0.0004455930922827846,
+      "loss": 3.5678,
+      "step": 23950
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "grad_norm": 0.5622717142105103,
+      "learning_rate": 0.0004452692930383162,
+      "loss": 3.5482,
+      "step": 24000
+    },
+    {
+      "epoch": 2.5876010781671157,
+      "eval_accuracy": 0.36659712519670246,
+      "eval_loss": 3.5317554473876953,
+      "eval_runtime": 181.2936,
+      "eval_samples_per_second": 99.347,
+      "eval_steps_per_second": 6.211,
+      "step": 24000
+    },
+    {
+      "epoch": 2.592991913746631,
+      "grad_norm": 0.6593669056892395,
+      "learning_rate": 0.00044494549379384775,
+      "loss": 3.5524,
+      "step": 24050
+    },
+    {
+      "epoch": 2.5983827493261455,
+      "grad_norm": 0.6096122860908508,
+      "learning_rate": 0.00044462169454937935,
+      "loss": 3.5285,
+      "step": 24100
+    },
+    {
+      "epoch": 2.6037735849056602,
+      "grad_norm": 0.5589276552200317,
+      "learning_rate": 0.0004442978953049109,
+      "loss": 3.5507,
+      "step": 24150
+    },
+    {
+      "epoch": 2.6091644204851754,
+      "grad_norm": 0.593991219997406,
+      "learning_rate": 0.0004439740960604425,
+      "loss": 3.5445,
+      "step": 24200
+    },
+    {
+      "epoch": 2.61455525606469,
+      "grad_norm": 0.5771310329437256,
+      "learning_rate": 0.00044365029681597406,
+      "loss": 3.5493,
+      "step": 24250
+    },
+    {
+      "epoch": 2.6199460916442048,
+      "grad_norm": 0.6104580760002136,
+      "learning_rate": 0.0004433264975715056,
+      "loss": 3.5519,
+      "step": 24300
+    },
+    {
+      "epoch": 2.62533692722372,
+      "grad_norm": 0.5800246000289917,
+      "learning_rate": 0.0004430026983270372,
+      "loss": 3.5551,
+      "step": 24350
+    },
+    {
+      "epoch": 2.6307277628032346,
+      "grad_norm": 0.5594913363456726,
+      "learning_rate": 0.00044267889908256876,
+      "loss": 3.5318,
+      "step": 24400
+    },
+    {
+      "epoch": 2.6361185983827493,
+      "grad_norm": 0.570894181728363,
+      "learning_rate": 0.00044235509983810037,
+      "loss": 3.5575,
+      "step": 24450
+    },
+    {
+      "epoch": 2.641509433962264,
+      "grad_norm": 0.5746343731880188,
+      "learning_rate": 0.0004420313005936319,
+      "loss": 3.5216,
+      "step": 24500
+    },
+    {
+      "epoch": 2.6469002695417787,
+      "grad_norm": 0.5852980613708496,
+      "learning_rate": 0.0004417075013491635,
+      "loss": 3.527,
+      "step": 24550
+    },
+    {
+      "epoch": 2.652291105121294,
+      "grad_norm": 0.6170436143875122,
+      "learning_rate": 0.0004413837021046951,
+      "loss": 3.5347,
+      "step": 24600
+    },
+    {
+      "epoch": 2.6576819407008085,
+      "grad_norm": 0.5268665552139282,
+      "learning_rate": 0.00044106637884511597,
+      "loss": 3.5423,
+      "step": 24650
+    },
+    {
+      "epoch": 2.6630727762803232,
+      "grad_norm": 0.5274102687835693,
+      "learning_rate": 0.0004407425796006475,
+      "loss": 3.542,
+      "step": 24700
+    },
+    {
+      "epoch": 2.6684636118598384,
+      "grad_norm": 0.562574028968811,
+      "learning_rate": 0.00044041878035617913,
+      "loss": 3.5585,
+      "step": 24750
+    },
+    {
+      "epoch": 2.673854447439353,
+      "grad_norm": 0.6228138208389282,
+      "learning_rate": 0.0004400949811117107,
+      "loss": 3.5354,
+      "step": 24800
+    },
+    {
+      "epoch": 2.6792452830188678,
+      "grad_norm": 0.5509768724441528,
+      "learning_rate": 0.0004397711818672423,
+      "loss": 3.534,
+      "step": 24850
+    },
+    {
+      "epoch": 2.684636118598383,
+      "grad_norm": 0.5812339186668396,
+      "learning_rate": 0.00043944738262277383,
+      "loss": 3.5431,
+      "step": 24900
+    },
+    {
+      "epoch": 2.6900269541778976,
+      "grad_norm": 0.5285933017730713,
+      "learning_rate": 0.00043912358337830544,
+      "loss": 3.5385,
+      "step": 24950
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "grad_norm": 0.5929449200630188,
+      "learning_rate": 0.000438799784133837,
+      "loss": 3.5542,
+      "step": 25000
+    },
+    {
+      "epoch": 2.6954177897574123,
+      "eval_accuracy": 0.3673601944278236,
+      "eval_loss": 3.5220768451690674,
+      "eval_runtime": 181.2911,
+      "eval_samples_per_second": 99.349,
+      "eval_steps_per_second": 6.211,
+      "step": 25000
+    },
+    {
+      "epoch": 2.7008086253369274,
+      "grad_norm": 0.6103118658065796,
+      "learning_rate": 0.00043847598488936854,
+      "loss": 3.5537,
+      "step": 25050
+    },
+    {
+      "epoch": 2.706199460916442,
+      "grad_norm": 0.6209373474121094,
+      "learning_rate": 0.00043815218564490014,
+      "loss": 3.5647,
+      "step": 25100
+    },
+    {
+      "epoch": 2.711590296495957,
+      "grad_norm": 0.6222750544548035,
+      "learning_rate": 0.0004378283864004317,
+      "loss": 3.5444,
+      "step": 25150
+    },
+    {
+      "epoch": 2.7169811320754715,
+      "grad_norm": 0.5424869060516357,
+      "learning_rate": 0.0004375045871559633,
+      "loss": 3.5295,
+      "step": 25200
+    },
+    {
+      "epoch": 2.7223719676549867,
+      "grad_norm": 0.5619545578956604,
+      "learning_rate": 0.00043718078791149485,
+      "loss": 3.5323,
+      "step": 25250
+    },
+    {
+      "epoch": 2.7277628032345014,
+      "grad_norm": 0.5697439312934875,
+      "learning_rate": 0.00043685698866702645,
+      "loss": 3.5397,
+      "step": 25300
+    },
+    {
+      "epoch": 2.733153638814016,
+      "grad_norm": 0.5689240097999573,
+      "learning_rate": 0.00043653318942255795,
+      "loss": 3.5332,
+      "step": 25350
+    },
+    {
+      "epoch": 2.7385444743935308,
+      "grad_norm": 0.5533002018928528,
+      "learning_rate": 0.00043620939017808956,
+      "loss": 3.5223,
+      "step": 25400
+    },
+    {
+      "epoch": 2.743935309973046,
+      "grad_norm": 0.5482644438743591,
+      "learning_rate": 0.0004358855909336211,
+      "loss": 3.5409,
+      "step": 25450
+    },
+    {
+      "epoch": 2.7493261455525606,
+      "grad_norm": 0.5769422650337219,
+      "learning_rate": 0.00043556179168915266,
+      "loss": 3.5584,
+      "step": 25500
+    },
+    {
+      "epoch": 2.7547169811320753,
+      "grad_norm": 0.6086459159851074,
+      "learning_rate": 0.00043523799244468426,
+      "loss": 3.5519,
+      "step": 25550
+    },
+    {
+      "epoch": 2.7601078167115904,
+      "grad_norm": 0.5854797959327698,
+      "learning_rate": 0.0004349141932002158,
+      "loss": 3.5419,
+      "step": 25600
+    },
+    {
+      "epoch": 2.765498652291105,
+      "grad_norm": 0.5399619340896606,
+      "learning_rate": 0.0004345903939557474,
+      "loss": 3.5286,
+      "step": 25650
+    },
+    {
+      "epoch": 2.77088948787062,
+      "grad_norm": 0.5609824061393738,
+      "learning_rate": 0.00043426659471127897,
+      "loss": 3.5512,
+      "step": 25700
+    },
+    {
+      "epoch": 2.776280323450135,
+      "grad_norm": 0.5754132866859436,
+      "learning_rate": 0.00043394279546681057,
+      "loss": 3.5341,
+      "step": 25750
+    },
+    {
+      "epoch": 2.7816711590296497,
+      "grad_norm": 0.5857130289077759,
+      "learning_rate": 0.0004336189962223421,
+      "loss": 3.5377,
+      "step": 25800
+    },
+    {
+      "epoch": 2.7870619946091644,
+      "grad_norm": 0.5504029393196106,
+      "learning_rate": 0.0004332951969778737,
+      "loss": 3.5221,
+      "step": 25850
+    },
+    {
+      "epoch": 2.7924528301886795,
+      "grad_norm": 0.56480473279953,
+      "learning_rate": 0.0004329713977334053,
+      "loss": 3.5429,
+      "step": 25900
+    },
+    {
+      "epoch": 2.797843665768194,
+      "grad_norm": 0.5525421500205994,
+      "learning_rate": 0.0004326475984889368,
+      "loss": 3.5136,
+      "step": 25950
+    },
+    {
+      "epoch": 2.803234501347709,
+      "grad_norm": 0.6140046715736389,
+      "learning_rate": 0.0004323237992444684,
+      "loss": 3.5342,
+      "step": 26000
+    },
+    {
+      "epoch": 2.803234501347709,
+      "eval_accuracy": 0.3681909544080531,
+      "eval_loss": 3.5146098136901855,
+      "eval_runtime": 181.2808,
+      "eval_samples_per_second": 99.354,
+      "eval_steps_per_second": 6.211,
+      "step": 26000
+    },
+    {
+      "epoch": 2.8086253369272236,
+      "grad_norm": 0.5526379346847534,
+      "learning_rate": 0.00043199999999999993,
+      "loss": 3.5372,
+      "step": 26050
+    },
+    {
+      "epoch": 2.8140161725067383,
+      "grad_norm": 0.5372548699378967,
+      "learning_rate": 0.00043167620075553153,
+      "loss": 3.5251,
+      "step": 26100
+    },
+    {
+      "epoch": 2.8194070080862534,
+      "grad_norm": 0.5878198742866516,
+      "learning_rate": 0.0004313524015110631,
+      "loss": 3.536,
+      "step": 26150
+    },
+    {
+      "epoch": 2.824797843665768,
+      "grad_norm": 0.6414244174957275,
+      "learning_rate": 0.0004310286022665947,
+      "loss": 3.5469,
+      "step": 26200
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.5342559218406677,
+      "learning_rate": 0.00043070480302212624,
+      "loss": 3.533,
+      "step": 26250
+    },
+    {
+      "epoch": 2.835579514824798,
+      "grad_norm": 0.5910995602607727,
+      "learning_rate": 0.0004303810037776578,
+      "loss": 3.532,
+      "step": 26300
+    },
+    {
+      "epoch": 2.8409703504043127,
+      "grad_norm": 0.4914022982120514,
+      "learning_rate": 0.0004300572045331894,
+      "loss": 3.5451,
+      "step": 26350
+    },
+    {
+      "epoch": 2.8463611859838274,
+      "grad_norm": 0.5698440074920654,
+      "learning_rate": 0.00042973340528872095,
+      "loss": 3.528,
+      "step": 26400
+    },
+    {
+      "epoch": 2.8517520215633425,
+      "grad_norm": 0.5266543030738831,
+      "learning_rate": 0.00042940960604425255,
+      "loss": 3.5381,
+      "step": 26450
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 0.5502971410751343,
+      "learning_rate": 0.0004290858067997841,
+      "loss": 3.5261,
+      "step": 26500
+    },
+    {
+      "epoch": 2.862533692722372,
+      "grad_norm": 0.5623043775558472,
+      "learning_rate": 0.0004287620075553157,
+      "loss": 3.5085,
+      "step": 26550
+    },
+    {
+      "epoch": 2.867924528301887,
+      "grad_norm": 0.5682037472724915,
+      "learning_rate": 0.00042843820831084726,
+      "loss": 3.5304,
+      "step": 26600
+    },
+    {
+      "epoch": 2.8733153638814017,
+      "grad_norm": 0.5245981216430664,
+      "learning_rate": 0.00042811440906637886,
+      "loss": 3.5259,
+      "step": 26650
+    },
+    {
+      "epoch": 2.8787061994609164,
+      "grad_norm": 0.558342456817627,
+      "learning_rate": 0.00042779060982191036,
+      "loss": 3.5226,
+      "step": 26700
+    },
+    {
+      "epoch": 2.884097035040431,
+      "grad_norm": 0.5356988310813904,
+      "learning_rate": 0.0004274668105774419,
+      "loss": 3.5257,
+      "step": 26750
+    },
+    {
+      "epoch": 2.889487870619946,
+      "grad_norm": 0.5896562933921814,
+      "learning_rate": 0.0004271430113329735,
+      "loss": 3.5067,
+      "step": 26800
+    },
+    {
+      "epoch": 2.894878706199461,
+      "grad_norm": 0.5992358922958374,
+      "learning_rate": 0.00042682568807339447,
+      "loss": 3.5249,
+      "step": 26850
+    },
+    {
+      "epoch": 2.9002695417789757,
+      "grad_norm": 0.5337636470794678,
+      "learning_rate": 0.000426501888828926,
+      "loss": 3.5128,
+      "step": 26900
+    },
+    {
+      "epoch": 2.9056603773584904,
+      "grad_norm": 0.6004590392112732,
+      "learning_rate": 0.0004261780895844576,
+      "loss": 3.5133,
+      "step": 26950
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "grad_norm": 0.5698912739753723,
+      "learning_rate": 0.00042585429033998917,
+      "loss": 3.5381,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9110512129380055,
+      "eval_accuracy": 0.3692894351038575,
+      "eval_loss": 3.504966974258423,
+      "eval_runtime": 181.0594,
+      "eval_samples_per_second": 99.476,
+      "eval_steps_per_second": 6.219,
+      "step": 27000
+    },
+    {
+      "epoch": 2.91644204851752,
+      "grad_norm": 0.6112388372421265,
+      "learning_rate": 0.0004255304910955207,
+      "loss": 3.5501,
+      "step": 27050
+    },
+    {
+      "epoch": 2.921832884097035,
+      "grad_norm": 0.5333527326583862,
+      "learning_rate": 0.0004252066918510523,
+      "loss": 3.5325,
+      "step": 27100
+    },
+    {
+      "epoch": 2.92722371967655,
+      "grad_norm": 0.5607829689979553,
+      "learning_rate": 0.0004248828926065839,
+      "loss": 3.5464,
+      "step": 27150
+    },
+    {
+      "epoch": 2.9326145552560647,
+      "grad_norm": 0.5657780766487122,
+      "learning_rate": 0.0004245590933621155,
+      "loss": 3.5333,
+      "step": 27200
+    },
+    {
+      "epoch": 2.9380053908355794,
+      "grad_norm": 0.5746095180511475,
+      "learning_rate": 0.00042423529411764703,
+      "loss": 3.5474,
+      "step": 27250
+    },
+    {
+      "epoch": 2.9433962264150946,
+      "grad_norm": 0.604708194732666,
+      "learning_rate": 0.00042391149487317864,
+      "loss": 3.5284,
+      "step": 27300
+    },
+    {
+      "epoch": 2.9487870619946093,
+      "grad_norm": 0.6213203072547913,
+      "learning_rate": 0.00042358769562871013,
+      "loss": 3.5173,
+      "step": 27350
+    },
+    {
+      "epoch": 2.954177897574124,
+      "grad_norm": 0.5948765277862549,
+      "learning_rate": 0.00042326389638424174,
+      "loss": 3.5149,
+      "step": 27400
+    },
+    {
+      "epoch": 2.9595687331536387,
+      "grad_norm": 0.5501255989074707,
+      "learning_rate": 0.0004229400971397733,
+      "loss": 3.5339,
+      "step": 27450
+    },
+    {
+      "epoch": 2.964959568733154,
+      "grad_norm": 0.5301875472068787,
+      "learning_rate": 0.00042261629789530484,
+      "loss": 3.5331,
+      "step": 27500
+    },
+    {
+      "epoch": 2.9703504043126685,
+      "grad_norm": 0.5574668049812317,
+      "learning_rate": 0.00042229249865083644,
+      "loss": 3.522,
+      "step": 27550
+    },
+    {
+      "epoch": 2.975741239892183,
+      "grad_norm": 0.5755630135536194,
+      "learning_rate": 0.000421968699406368,
+      "loss": 3.5198,
+      "step": 27600
+    },
+    {
+      "epoch": 2.981132075471698,
+      "grad_norm": 0.54078608751297,
+      "learning_rate": 0.0004216449001618996,
+      "loss": 3.5438,
+      "step": 27650
+    },
+    {
+      "epoch": 2.986522911051213,
+      "grad_norm": 0.5769183039665222,
+      "learning_rate": 0.00042132110091743115,
+      "loss": 3.5085,
+      "step": 27700
+    },
+    {
+      "epoch": 2.9919137466307277,
+      "grad_norm": 0.5652152895927429,
+      "learning_rate": 0.00042099730167296275,
+      "loss": 3.5019,
+      "step": 27750
+    },
+    {
+      "epoch": 2.9973045822102424,
+      "grad_norm": 0.5847728848457336,
+      "learning_rate": 0.0004206735024284943,
+      "loss": 3.5239,
+      "step": 27800
+    },
+    {
+      "epoch": 3.0026954177897576,
+      "grad_norm": 0.5741530656814575,
+      "learning_rate": 0.0004203497031840259,
+      "loss": 3.4812,
+      "step": 27850
+    },
+    {
+      "epoch": 3.0080862533692723,
+      "grad_norm": 0.5797563195228577,
+      "learning_rate": 0.00042002590393955746,
+      "loss": 3.4285,
+      "step": 27900
+    },
+    {
+      "epoch": 3.013477088948787,
+      "grad_norm": 0.634753406047821,
+      "learning_rate": 0.00041970210469508896,
+      "loss": 3.4059,
+      "step": 27950
+    },
+    {
+      "epoch": 3.018867924528302,
+      "grad_norm": 0.5791683197021484,
+      "learning_rate": 0.00041937830545062056,
+      "loss": 3.4278,
+      "step": 28000
+    },
+    {
+      "epoch": 3.018867924528302,
+      "eval_accuracy": 0.37001849598109265,
+      "eval_loss": 3.502011299133301,
+      "eval_runtime": 181.3383,
+      "eval_samples_per_second": 99.323,
+      "eval_steps_per_second": 6.209,
+      "step": 28000
+    },
+    {
+      "epoch": 3.024258760107817,
+      "grad_norm": 0.603844165802002,
+      "learning_rate": 0.0004190545062061521,
+      "loss": 3.4184,
+      "step": 28050
+    },
+    {
+      "epoch": 3.0296495956873315,
+      "grad_norm": 0.5584900975227356,
+      "learning_rate": 0.0004187307069616837,
+      "loss": 3.435,
+      "step": 28100
+    },
+    {
+      "epoch": 3.035040431266846,
+      "grad_norm": 0.5561067461967468,
+      "learning_rate": 0.00041840690771721527,
+      "loss": 3.4382,
+      "step": 28150
+    },
+    {
+      "epoch": 3.0404312668463613,
+      "grad_norm": 0.6057212948799133,
+      "learning_rate": 0.00041808310847274687,
+      "loss": 3.435,
+      "step": 28200
+    },
+    {
+      "epoch": 3.045822102425876,
+      "grad_norm": 0.5753740668296814,
+      "learning_rate": 0.0004177593092282784,
+      "loss": 3.4403,
+      "step": 28250
+    },
+    {
+      "epoch": 3.0512129380053907,
+      "grad_norm": 0.5986172556877136,
+      "learning_rate": 0.00041743550998381,
+      "loss": 3.4394,
+      "step": 28300
+    },
+    {
+      "epoch": 3.056603773584906,
+      "grad_norm": 0.5863624811172485,
+      "learning_rate": 0.0004171117107393416,
+      "loss": 3.4487,
+      "step": 28350
+    },
+    {
+      "epoch": 3.0619946091644206,
+      "grad_norm": 0.6170995831489563,
+      "learning_rate": 0.00041678791149487313,
+      "loss": 3.4238,
+      "step": 28400
+    },
+    {
+      "epoch": 3.0673854447439353,
+      "grad_norm": 0.5697750449180603,
+      "learning_rate": 0.00041646411225040473,
+      "loss": 3.4469,
+      "step": 28450
+    },
+    {
+      "epoch": 3.07277628032345,
+      "grad_norm": 0.5869051814079285,
+      "learning_rate": 0.0004161403130059363,
+      "loss": 3.4314,
+      "step": 28500
+    },
+    {
+      "epoch": 3.078167115902965,
+      "grad_norm": 0.6245196461677551,
+      "learning_rate": 0.0004158165137614679,
+      "loss": 3.4267,
+      "step": 28550
+    },
+    {
+      "epoch": 3.08355795148248,
+      "grad_norm": 0.5763593912124634,
+      "learning_rate": 0.00041549271451699944,
+      "loss": 3.4496,
+      "step": 28600
+    },
+    {
+      "epoch": 3.0889487870619945,
+      "grad_norm": 0.6235705018043518,
+      "learning_rate": 0.00041516891527253104,
+      "loss": 3.4518,
+      "step": 28650
+    },
+    {
+      "epoch": 3.0943396226415096,
+      "grad_norm": 0.5487541556358337,
+      "learning_rate": 0.00041484511602806254,
+      "loss": 3.4529,
+      "step": 28700
+    },
+    {
+      "epoch": 3.0997304582210243,
+      "grad_norm": 0.5920215845108032,
+      "learning_rate": 0.0004145213167835941,
+      "loss": 3.4479,
+      "step": 28750
+    },
+    {
+      "epoch": 3.105121293800539,
+      "grad_norm": 0.6436508893966675,
+      "learning_rate": 0.0004141975175391257,
+      "loss": 3.4435,
+      "step": 28800
+    },
+    {
+      "epoch": 3.1105121293800537,
+      "grad_norm": 0.6300854682922363,
+      "learning_rate": 0.00041387371829465725,
+      "loss": 3.4504,
+      "step": 28850
+    },
+    {
+      "epoch": 3.115902964959569,
+      "grad_norm": 0.6098328828811646,
+      "learning_rate": 0.00041354991905018885,
+      "loss": 3.4489,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1212938005390836,
+      "grad_norm": 0.5779024958610535,
+      "learning_rate": 0.0004132261198057204,
+      "loss": 3.4428,
+      "step": 28950
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "grad_norm": 0.6418079137802124,
+      "learning_rate": 0.00041290879654614135,
+      "loss": 3.4513,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1266846361185983,
+      "eval_accuracy": 0.37081785527674377,
+      "eval_loss": 3.4941792488098145,
+      "eval_runtime": 181.2333,
+      "eval_samples_per_second": 99.38,
+      "eval_steps_per_second": 6.213,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1320754716981134,
+      "grad_norm": 0.5885114073753357,
+      "learning_rate": 0.00041258499730167296,
+      "loss": 3.431,
+      "step": 29050
+    },
+    {
+      "epoch": 3.137466307277628,
+      "grad_norm": 0.5644581913948059,
+      "learning_rate": 0.0004122611980572045,
+      "loss": 3.4565,
+      "step": 29100
+    },
+    {
+      "epoch": 3.142857142857143,
+      "grad_norm": 0.5936179161071777,
+      "learning_rate": 0.00041193739881273606,
+      "loss": 3.451,
+      "step": 29150
+    },
+    {
+      "epoch": 3.1482479784366575,
+      "grad_norm": 0.5671179294586182,
+      "learning_rate": 0.00041161359956826766,
+      "loss": 3.4676,
+      "step": 29200
+    },
+    {
+      "epoch": 3.1536388140161726,
+      "grad_norm": 0.5778055191040039,
+      "learning_rate": 0.0004112898003237992,
+      "loss": 3.4678,
+      "step": 29250
+    },
+    {
+      "epoch": 3.1590296495956873,
+      "grad_norm": 0.6444634795188904,
+      "learning_rate": 0.0004109660010793308,
+      "loss": 3.4602,
+      "step": 29300
+    },
+    {
+      "epoch": 3.164420485175202,
+      "grad_norm": 0.5693305730819702,
+      "learning_rate": 0.0004106422018348623,
+      "loss": 3.4477,
+      "step": 29350
+    },
+    {
+      "epoch": 3.169811320754717,
+      "grad_norm": 0.6001082062721252,
+      "learning_rate": 0.0004103184025903939,
+      "loss": 3.4583,
+      "step": 29400
+    },
+    {
+      "epoch": 3.175202156334232,
+      "grad_norm": 0.598551332950592,
+      "learning_rate": 0.00040999460334592547,
+      "loss": 3.4711,
+      "step": 29450
+    },
+    {
+      "epoch": 3.1805929919137466,
+      "grad_norm": 0.6061559319496155,
+      "learning_rate": 0.000409670804101457,
+      "loss": 3.4438,
+      "step": 29500
+    },
+    {
+      "epoch": 3.1859838274932613,
+      "grad_norm": 0.6079073548316956,
+      "learning_rate": 0.00040934700485698863,
+      "loss": 3.4549,
+      "step": 29550
+    },
+    {
+      "epoch": 3.1913746630727764,
+      "grad_norm": 0.5677346587181091,
+      "learning_rate": 0.0004090232056125202,
+      "loss": 3.4684,
+      "step": 29600
+    },
+    {
+      "epoch": 3.196765498652291,
+      "grad_norm": 0.5982911586761475,
+      "learning_rate": 0.0004086994063680518,
+      "loss": 3.4518,
+      "step": 29650
+    },
+    {
+      "epoch": 3.202156334231806,
+      "grad_norm": 0.5999668836593628,
+      "learning_rate": 0.00040837560712358333,
+      "loss": 3.4704,
+      "step": 29700
+    },
+    {
+      "epoch": 3.207547169811321,
+      "grad_norm": 0.6423681974411011,
+      "learning_rate": 0.00040805180787911494,
+      "loss": 3.4663,
+      "step": 29750
+    },
+    {
+      "epoch": 3.2129380053908356,
+      "grad_norm": 0.6018577218055725,
+      "learning_rate": 0.0004077280086346465,
+      "loss": 3.4421,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2183288409703503,
+      "grad_norm": 0.6331573128700256,
+      "learning_rate": 0.0004074042093901781,
+      "loss": 3.4647,
+      "step": 29850
+    },
+    {
+      "epoch": 3.223719676549865,
+      "grad_norm": 0.5911356210708618,
+      "learning_rate": 0.00040708041014570964,
+      "loss": 3.4587,
+      "step": 29900
+    },
+    {
+      "epoch": 3.22911051212938,
+      "grad_norm": 0.6494351625442505,
+      "learning_rate": 0.00040675661090124114,
+      "loss": 3.4525,
+      "step": 29950
+    },
+    {
+      "epoch": 3.234501347708895,
+      "grad_norm": 0.6582809090614319,
+      "learning_rate": 0.0004064328116567728,
+      "loss": 3.4627,
+      "step": 30000
+    },
+    {
+      "epoch": 3.234501347708895,
+      "eval_accuracy": 0.37125040242313323,
+      "eval_loss": 3.4898300170898438,
+      "eval_runtime": 181.4533,
+      "eval_samples_per_second": 99.26,
+      "eval_steps_per_second": 6.205,
+      "step": 30000
+    },
+    {
+      "epoch": 3.2398921832884096,
+      "grad_norm": 0.567914605140686,
+      "learning_rate": 0.0004061090124123043,
+      "loss": 3.439,
+      "step": 30050
+    },
+    {
+      "epoch": 3.2452830188679247,
+      "grad_norm": 0.8371458053588867,
+      "learning_rate": 0.0004057852131678359,
+      "loss": 3.457,
+      "step": 30100
+    },
+    {
+      "epoch": 3.2506738544474394,
+      "grad_norm": 0.5806177258491516,
+      "learning_rate": 0.00040546141392336745,
+      "loss": 3.4592,
+      "step": 30150
+    },
+    {
+      "epoch": 3.256064690026954,
+      "grad_norm": 0.5861149430274963,
+      "learning_rate": 0.00040513761467889906,
+      "loss": 3.4733,
+      "step": 30200
+    },
+    {
+      "epoch": 3.2614555256064692,
+      "grad_norm": 0.5776110291481018,
+      "learning_rate": 0.0004048138154344306,
+      "loss": 3.4591,
+      "step": 30250
+    },
+    {
+      "epoch": 3.266846361185984,
+      "grad_norm": 0.6323752403259277,
+      "learning_rate": 0.0004044900161899622,
+      "loss": 3.4368,
+      "step": 30300
+    },
+    {
+      "epoch": 3.2722371967654986,
+      "grad_norm": 0.5726909637451172,
+      "learning_rate": 0.00040416621694549376,
+      "loss": 3.4548,
+      "step": 30350
+    },
+    {
+      "epoch": 3.2776280323450133,
+      "grad_norm": 0.6342945098876953,
+      "learning_rate": 0.0004038424177010253,
+      "loss": 3.4659,
+      "step": 30400
+    },
+    {
+      "epoch": 3.2830188679245285,
+      "grad_norm": 0.8313270211219788,
+      "learning_rate": 0.0004035186184565569,
+      "loss": 3.4545,
+      "step": 30450
+    },
+    {
+      "epoch": 3.288409703504043,
+      "grad_norm": 0.6741660833358765,
+      "learning_rate": 0.00040319481921208847,
+      "loss": 3.4664,
+      "step": 30500
+    },
+    {
+      "epoch": 3.293800539083558,
+      "grad_norm": 0.6140432357788086,
+      "learning_rate": 0.00040287101996762007,
+      "loss": 3.4624,
+      "step": 30550
+    },
+    {
+      "epoch": 3.2991913746630726,
+      "grad_norm": 0.5720652937889099,
+      "learning_rate": 0.0004025472207231516,
+      "loss": 3.4554,
+      "step": 30600
+    },
+    {
+      "epoch": 3.3045822102425877,
+      "grad_norm": 0.597766637802124,
+      "learning_rate": 0.00040222342147868323,
+      "loss": 3.4516,
+      "step": 30650
+    },
+    {
+      "epoch": 3.3099730458221024,
+      "grad_norm": 0.644696831703186,
+      "learning_rate": 0.0004018996222342147,
+      "loss": 3.4671,
+      "step": 30700
+    },
+    {
+      "epoch": 3.315363881401617,
+      "grad_norm": 0.5738349556922913,
+      "learning_rate": 0.00040157582298974633,
+      "loss": 3.4429,
+      "step": 30750
+    },
+    {
+      "epoch": 3.3207547169811322,
+      "grad_norm": 0.6259222030639648,
+      "learning_rate": 0.0004012520237452779,
+      "loss": 3.4399,
+      "step": 30800
+    },
+    {
+      "epoch": 3.326145552560647,
+      "grad_norm": 0.6019585728645325,
+      "learning_rate": 0.00040092822450080943,
+      "loss": 3.4413,
+      "step": 30850
+    },
+    {
+      "epoch": 3.3315363881401616,
+      "grad_norm": 0.5780307054519653,
+      "learning_rate": 0.00040060442525634103,
+      "loss": 3.4696,
+      "step": 30900
+    },
+    {
+      "epoch": 3.3369272237196768,
+      "grad_norm": 0.5876522660255432,
+      "learning_rate": 0.0004002806260118726,
+      "loss": 3.4682,
+      "step": 30950
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "grad_norm": 0.58372563123703,
+      "learning_rate": 0.0003999568267674042,
+      "loss": 3.4463,
+      "step": 31000
+    },
+    {
+      "epoch": 3.3423180592991915,
+      "eval_accuracy": 0.37201792642265474,
+      "eval_loss": 3.4831817150115967,
+      "eval_runtime": 181.0431,
+      "eval_samples_per_second": 99.485,
+      "eval_steps_per_second": 6.22,
+      "step": 31000
+    },
+    {
+      "epoch": 3.347708894878706,
+      "grad_norm": 0.5882640480995178,
+      "learning_rate": 0.00039963950350782514,
+      "loss": 3.4554,
+      "step": 31050
+    },
+    {
+      "epoch": 3.353099730458221,
+      "grad_norm": 0.6506269574165344,
+      "learning_rate": 0.0003993157042633567,
+      "loss": 3.4735,
+      "step": 31100
+    },
+    {
+      "epoch": 3.358490566037736,
+      "grad_norm": 0.5855028033256531,
+      "learning_rate": 0.00039899190501888824,
+      "loss": 3.4553,
+      "step": 31150
+    },
+    {
+      "epoch": 3.3638814016172507,
+      "grad_norm": 0.5879990458488464,
+      "learning_rate": 0.00039866810577441985,
+      "loss": 3.4785,
+      "step": 31200
+    },
+    {
+      "epoch": 3.3692722371967654,
+      "grad_norm": 0.5899655222892761,
+      "learning_rate": 0.0003983443065299514,
+      "loss": 3.4739,
+      "step": 31250
+    },
+    {
+      "epoch": 3.37466307277628,
+      "grad_norm": 0.5837485790252686,
+      "learning_rate": 0.000398020507285483,
+      "loss": 3.4662,
+      "step": 31300
+    },
+    {
+      "epoch": 3.3800539083557952,
+      "grad_norm": 0.584699809551239,
+      "learning_rate": 0.0003976967080410145,
+      "loss": 3.4639,
+      "step": 31350
+    },
+    {
+      "epoch": 3.38544474393531,
+      "grad_norm": 0.5864905118942261,
+      "learning_rate": 0.0003973729087965461,
+      "loss": 3.4694,
+      "step": 31400
+    },
+    {
+      "epoch": 3.3908355795148246,
+      "grad_norm": 0.6279877424240112,
+      "learning_rate": 0.00039704910955207765,
+      "loss": 3.4608,
+      "step": 31450
+    },
+    {
+      "epoch": 3.3962264150943398,
+      "grad_norm": 0.7739274501800537,
+      "learning_rate": 0.00039672531030760926,
+      "loss": 3.4622,
+      "step": 31500
+    },
+    {
+      "epoch": 3.4016172506738545,
+      "grad_norm": 0.6367618441581726,
+      "learning_rate": 0.0003964015110631408,
+      "loss": 3.4693,
+      "step": 31550
+    },
+    {
+      "epoch": 3.407008086253369,
+      "grad_norm": 0.622832179069519,
+      "learning_rate": 0.00039607771181867236,
+      "loss": 3.4576,
+      "step": 31600
+    },
+    {
+      "epoch": 3.4123989218328843,
+      "grad_norm": 0.5996997356414795,
+      "learning_rate": 0.00039575391257420397,
+      "loss": 3.4453,
+      "step": 31650
+    },
+    {
+      "epoch": 3.417789757412399,
+      "grad_norm": 0.5975577235221863,
+      "learning_rate": 0.0003954301133297355,
+      "loss": 3.4643,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4231805929919137,
+      "grad_norm": 0.6235230565071106,
+      "learning_rate": 0.0003951063140852671,
+      "loss": 3.466,
+      "step": 31750
+    },
+    {
+      "epoch": 3.4285714285714284,
+      "grad_norm": 0.5875454545021057,
+      "learning_rate": 0.00039478251484079867,
+      "loss": 3.4483,
+      "step": 31800
+    },
+    {
+      "epoch": 3.4339622641509435,
+      "grad_norm": 0.6354802846908569,
+      "learning_rate": 0.0003944587155963303,
+      "loss": 3.4622,
+      "step": 31850
+    },
+    {
+      "epoch": 3.439353099730458,
+      "grad_norm": 0.6226475834846497,
+      "learning_rate": 0.0003941349163518618,
+      "loss": 3.4724,
+      "step": 31900
+    },
+    {
+      "epoch": 3.444743935309973,
+      "grad_norm": 0.5510740280151367,
+      "learning_rate": 0.00039381111710739343,
+      "loss": 3.4656,
+      "step": 31950
+    },
+    {
+      "epoch": 3.450134770889488,
+      "grad_norm": 0.5717306733131409,
+      "learning_rate": 0.000393487317862925,
+      "loss": 3.461,
+      "step": 32000
+    },
+    {
+      "epoch": 3.450134770889488,
+      "eval_accuracy": 0.3726614774771799,
+      "eval_loss": 3.4775688648223877,
+      "eval_runtime": 181.8998,
+      "eval_samples_per_second": 99.016,
+      "eval_steps_per_second": 6.19,
+      "step": 32000
+    },
+    {
+      "epoch": 3.4555256064690028,
+      "grad_norm": 0.591214120388031,
+      "learning_rate": 0.0003931635186184565,
+      "loss": 3.4622,
+      "step": 32050
+    },
+    {
+      "epoch": 3.4609164420485174,
+      "grad_norm": 0.6022748351097107,
+      "learning_rate": 0.0003928397193739881,
+      "loss": 3.4683,
+      "step": 32100
+    },
+    {
+      "epoch": 3.466307277628032,
+      "grad_norm": 0.5917901992797852,
+      "learning_rate": 0.00039251592012951963,
+      "loss": 3.4647,
+      "step": 32150
+    },
+    {
+      "epoch": 3.4716981132075473,
+      "grad_norm": 0.5644868016242981,
+      "learning_rate": 0.00039219212088505124,
+      "loss": 3.4447,
+      "step": 32200
+    },
+    {
+      "epoch": 3.477088948787062,
+      "grad_norm": 0.5835516452789307,
+      "learning_rate": 0.0003918683216405828,
+      "loss": 3.46,
+      "step": 32250
+    },
+    {
+      "epoch": 3.4824797843665767,
+      "grad_norm": 0.6037916541099548,
+      "learning_rate": 0.0003915445223961144,
+      "loss": 3.4795,
+      "step": 32300
+    },
+    {
+      "epoch": 3.487870619946092,
+      "grad_norm": 0.6386679410934448,
+      "learning_rate": 0.00039122072315164594,
+      "loss": 3.4523,
+      "step": 32350
+    },
+    {
+      "epoch": 3.4932614555256065,
+      "grad_norm": 0.5861144661903381,
+      "learning_rate": 0.0003908969239071775,
+      "loss": 3.4626,
+      "step": 32400
+    },
+    {
+      "epoch": 3.498652291105121,
+      "grad_norm": 0.6120802164077759,
+      "learning_rate": 0.0003905731246627091,
+      "loss": 3.4513,
+      "step": 32450
+    },
+    {
+      "epoch": 3.5040431266846364,
+      "grad_norm": 0.5773576498031616,
+      "learning_rate": 0.00039024932541824065,
+      "loss": 3.4613,
+      "step": 32500
+    },
+    {
+      "epoch": 3.509433962264151,
+      "grad_norm": 0.6126224994659424,
+      "learning_rate": 0.00038992552617377225,
+      "loss": 3.4569,
+      "step": 32550
+    },
+    {
+      "epoch": 3.5148247978436657,
+      "grad_norm": 0.567841649055481,
+      "learning_rate": 0.0003896017269293038,
+      "loss": 3.4605,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5202156334231804,
+      "grad_norm": 0.6432450413703918,
+      "learning_rate": 0.0003892779276848354,
+      "loss": 3.4536,
+      "step": 32650
+    },
+    {
+      "epoch": 3.525606469002695,
+      "grad_norm": 0.5933105945587158,
+      "learning_rate": 0.0003889541284403669,
+      "loss": 3.4661,
+      "step": 32700
+    },
+    {
+      "epoch": 3.5309973045822103,
+      "grad_norm": 0.6399059295654297,
+      "learning_rate": 0.0003886303291958985,
+      "loss": 3.4539,
+      "step": 32750
+    },
+    {
+      "epoch": 3.536388140161725,
+      "grad_norm": 0.5856630802154541,
+      "learning_rate": 0.00038830652995143006,
+      "loss": 3.458,
+      "step": 32800
+    },
+    {
+      "epoch": 3.5417789757412397,
+      "grad_norm": 0.6280727386474609,
+      "learning_rate": 0.0003879827307069616,
+      "loss": 3.4386,
+      "step": 32850
+    },
+    {
+      "epoch": 3.547169811320755,
+      "grad_norm": 0.5890994668006897,
+      "learning_rate": 0.0003876589314624932,
+      "loss": 3.4536,
+      "step": 32900
+    },
+    {
+      "epoch": 3.5525606469002695,
+      "grad_norm": 0.5725958943367004,
+      "learning_rate": 0.00038733513221802477,
+      "loss": 3.4458,
+      "step": 32950
+    },
+    {
+      "epoch": 3.557951482479784,
+      "grad_norm": 0.5895776748657227,
+      "learning_rate": 0.00038701133297355637,
+      "loss": 3.4454,
+      "step": 33000
+    },
+    {
+      "epoch": 3.557951482479784,
+      "eval_accuracy": 0.3733202399359991,
+      "eval_loss": 3.471012830734253,
+      "eval_runtime": 181.5491,
+      "eval_samples_per_second": 99.207,
+      "eval_steps_per_second": 6.202,
+      "step": 33000
+    },
+    {
+      "epoch": 3.5633423180592994,
+      "grad_norm": 0.6420202851295471,
+      "learning_rate": 0.0003866940097139773,
+      "loss": 3.4538,
+      "step": 33050
+    },
+    {
+      "epoch": 3.568733153638814,
+      "grad_norm": 0.6431132555007935,
+      "learning_rate": 0.0003863702104695089,
+      "loss": 3.465,
+      "step": 33100
+    },
+    {
+      "epoch": 3.5741239892183287,
+      "grad_norm": 0.5854994058609009,
+      "learning_rate": 0.0003860464112250404,
+      "loss": 3.4379,
+      "step": 33150
+    },
+    {
+      "epoch": 3.579514824797844,
+      "grad_norm": 0.5908634066581726,
+      "learning_rate": 0.00038572261198057203,
+      "loss": 3.4458,
+      "step": 33200
+    },
+    {
+      "epoch": 3.5849056603773586,
+      "grad_norm": 0.6732088327407837,
+      "learning_rate": 0.0003853988127361036,
+      "loss": 3.4777,
+      "step": 33250
+    },
+    {
+      "epoch": 3.5902964959568733,
+      "grad_norm": 0.6526401042938232,
+      "learning_rate": 0.0003850750134916352,
+      "loss": 3.466,
+      "step": 33300
+    },
+    {
+      "epoch": 3.595687331536388,
+      "grad_norm": 0.5792251229286194,
+      "learning_rate": 0.0003847512142471667,
+      "loss": 3.4519,
+      "step": 33350
+    },
+    {
+      "epoch": 3.601078167115903,
+      "grad_norm": 0.6356695890426636,
+      "learning_rate": 0.0003844274150026983,
+      "loss": 3.4533,
+      "step": 33400
+    },
+    {
+      "epoch": 3.606469002695418,
+      "grad_norm": 0.6451331973075867,
+      "learning_rate": 0.00038410361575822984,
+      "loss": 3.4628,
+      "step": 33450
+    },
+    {
+      "epoch": 3.6118598382749325,
+      "grad_norm": 0.6042508482933044,
+      "learning_rate": 0.00038377981651376144,
+      "loss": 3.4444,
+      "step": 33500
+    },
+    {
+      "epoch": 3.617250673854447,
+      "grad_norm": 0.6613562703132629,
+      "learning_rate": 0.000383456017269293,
+      "loss": 3.4521,
+      "step": 33550
+    },
+    {
+      "epoch": 3.6226415094339623,
+      "grad_norm": 0.630096971988678,
+      "learning_rate": 0.00038313221802482454,
+      "loss": 3.4584,
+      "step": 33600
+    },
+    {
+      "epoch": 3.628032345013477,
+      "grad_norm": 0.6092630624771118,
+      "learning_rate": 0.00038280841878035615,
+      "loss": 3.4683,
+      "step": 33650
+    },
+    {
+      "epoch": 3.6334231805929917,
+      "grad_norm": 0.5494804978370667,
+      "learning_rate": 0.0003824910955207771,
+      "loss": 3.4633,
+      "step": 33700
+    },
+    {
+      "epoch": 3.638814016172507,
+      "grad_norm": 0.6219018697738647,
+      "learning_rate": 0.00038216729627630865,
+      "loss": 3.4468,
+      "step": 33750
+    },
+    {
+      "epoch": 3.6442048517520216,
+      "grad_norm": 0.5919956564903259,
+      "learning_rate": 0.00038184349703184026,
+      "loss": 3.4635,
+      "step": 33800
+    },
+    {
+      "epoch": 3.6495956873315363,
+      "grad_norm": 0.6003981828689575,
+      "learning_rate": 0.0003815196977873718,
+      "loss": 3.4612,
+      "step": 33850
+    },
+    {
+      "epoch": 3.6549865229110514,
+      "grad_norm": 0.6288914680480957,
+      "learning_rate": 0.0003811958985429034,
+      "loss": 3.451,
+      "step": 33900
+    },
+    {
+      "epoch": 3.660377358490566,
+      "grad_norm": 0.6367020010948181,
+      "learning_rate": 0.00038087209929843496,
+      "loss": 3.4472,
+      "step": 33950
+    },
+    {
+      "epoch": 3.665768194070081,
+      "grad_norm": 0.6496438980102539,
+      "learning_rate": 0.00038054830005396646,
+      "loss": 3.4627,
+      "step": 34000
+    },
+    {
+      "epoch": 3.665768194070081,
+      "eval_accuracy": 0.3740616872424451,
+      "eval_loss": 3.463249444961548,
+      "eval_runtime": 181.8587,
+      "eval_samples_per_second": 99.038,
+      "eval_steps_per_second": 6.192,
+      "step": 34000
+    },
+    {
+      "epoch": 3.671159029649596,
+      "grad_norm": 0.653170645236969,
+      "learning_rate": 0.00038022450080949806,
+      "loss": 3.4457,
+      "step": 34050
+    },
+    {
+      "epoch": 3.6765498652291106,
+      "grad_norm": 0.6262568831443787,
+      "learning_rate": 0.0003799007015650296,
+      "loss": 3.4609,
+      "step": 34100
+    },
+    {
+      "epoch": 3.6819407008086253,
+      "grad_norm": 0.6341161131858826,
+      "learning_rate": 0.0003795769023205612,
+      "loss": 3.4609,
+      "step": 34150
+    },
+    {
+      "epoch": 3.68733153638814,
+      "grad_norm": 0.6720100045204163,
+      "learning_rate": 0.00037925310307609277,
+      "loss": 3.4422,
+      "step": 34200
+    },
+    {
+      "epoch": 3.6927223719676547,
+      "grad_norm": 0.6978814601898193,
+      "learning_rate": 0.0003789293038316244,
+      "loss": 3.4553,
+      "step": 34250
+    },
+    {
+      "epoch": 3.69811320754717,
+      "grad_norm": 0.6090667843818665,
+      "learning_rate": 0.0003786055045871559,
+      "loss": 3.4458,
+      "step": 34300
+    },
+    {
+      "epoch": 3.7035040431266846,
+      "grad_norm": 0.5765736103057861,
+      "learning_rate": 0.0003782817053426875,
+      "loss": 3.4688,
+      "step": 34350
+    },
+    {
+      "epoch": 3.7088948787061993,
+      "grad_norm": 0.6406888365745544,
+      "learning_rate": 0.0003779579060982191,
+      "loss": 3.4534,
+      "step": 34400
+    },
+    {
+      "epoch": 3.7142857142857144,
+      "grad_norm": 0.6476412415504456,
+      "learning_rate": 0.00037763410685375063,
+      "loss": 3.4412,
+      "step": 34450
+    },
+    {
+      "epoch": 3.719676549865229,
+      "grad_norm": 0.81647789478302,
+      "learning_rate": 0.00037731030760928223,
+      "loss": 3.4558,
+      "step": 34500
+    },
+    {
+      "epoch": 3.725067385444744,
+      "grad_norm": 0.6053717732429504,
+      "learning_rate": 0.0003769865083648138,
+      "loss": 3.4615,
+      "step": 34550
+    },
+    {
+      "epoch": 3.730458221024259,
+      "grad_norm": 0.7041976451873779,
+      "learning_rate": 0.0003766627091203454,
+      "loss": 3.4616,
+      "step": 34600
+    },
+    {
+      "epoch": 3.7358490566037736,
+      "grad_norm": 0.6277915239334106,
+      "learning_rate": 0.00037633890987587694,
+      "loss": 3.453,
+      "step": 34650
+    },
+    {
+      "epoch": 3.7412398921832883,
+      "grad_norm": 0.5918686985969543,
+      "learning_rate": 0.00037601511063140855,
+      "loss": 3.4564,
+      "step": 34700
+    },
+    {
+      "epoch": 3.7466307277628035,
+      "grad_norm": 0.6289958357810974,
+      "learning_rate": 0.00037569131138694004,
+      "loss": 3.4394,
+      "step": 34750
+    },
+    {
+      "epoch": 3.752021563342318,
+      "grad_norm": 0.6078783869743347,
+      "learning_rate": 0.0003753675121424716,
+      "loss": 3.4725,
+      "step": 34800
+    },
+    {
+      "epoch": 3.757412398921833,
+      "grad_norm": 0.6587103009223938,
+      "learning_rate": 0.0003750437128980032,
+      "loss": 3.461,
+      "step": 34850
+    },
+    {
+      "epoch": 3.7628032345013476,
+      "grad_norm": 0.6301190853118896,
+      "learning_rate": 0.00037471991365353475,
+      "loss": 3.4595,
+      "step": 34900
+    },
+    {
+      "epoch": 3.7681940700808623,
+      "grad_norm": 0.6365683078765869,
+      "learning_rate": 0.00037439611440906635,
+      "loss": 3.4571,
+      "step": 34950
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.6449572443962097,
+      "learning_rate": 0.0003740723151645979,
+      "loss": 3.444,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "eval_accuracy": 0.37428866312508957,
+      "eval_loss": 3.458812952041626,
+      "eval_runtime": 181.8189,
+      "eval_samples_per_second": 99.06,
+      "eval_steps_per_second": 6.193,
+      "step": 35000
+    },
+    {
+      "epoch": 3.778975741239892,
+      "grad_norm": 0.6829414963722229,
+      "learning_rate": 0.0003737485159201295,
+      "loss": 3.4582,
+      "step": 35050
+    },
+    {
+      "epoch": 3.784366576819407,
+      "grad_norm": 0.6117042303085327,
+      "learning_rate": 0.00037342471667566106,
+      "loss": 3.4517,
+      "step": 35100
+    },
+    {
+      "epoch": 3.789757412398922,
+      "grad_norm": 0.6252544522285461,
+      "learning_rate": 0.00037310091743119266,
+      "loss": 3.4467,
+      "step": 35150
+    },
+    {
+      "epoch": 3.7951482479784366,
+      "grad_norm": 0.6546263694763184,
+      "learning_rate": 0.0003727771181867242,
+      "loss": 3.4705,
+      "step": 35200
+    },
+    {
+      "epoch": 3.8005390835579513,
+      "grad_norm": 0.6286924481391907,
+      "learning_rate": 0.00037245331894225576,
+      "loss": 3.4452,
+      "step": 35250
+    },
+    {
+      "epoch": 3.8059299191374665,
+      "grad_norm": 0.6140318512916565,
+      "learning_rate": 0.00037212951969778737,
+      "loss": 3.4407,
+      "step": 35300
+    },
+    {
+      "epoch": 3.811320754716981,
+      "grad_norm": 0.6014268398284912,
+      "learning_rate": 0.00037180572045331887,
+      "loss": 3.4401,
+      "step": 35350
+    },
+    {
+      "epoch": 3.816711590296496,
+      "grad_norm": 0.6223017573356628,
+      "learning_rate": 0.00037148192120885047,
+      "loss": 3.4541,
+      "step": 35400
+    },
+    {
+      "epoch": 3.822102425876011,
+      "grad_norm": 0.6230337023735046,
+      "learning_rate": 0.000371158121964382,
+      "loss": 3.4638,
+      "step": 35450
+    },
+    {
+      "epoch": 3.8274932614555257,
+      "grad_norm": 0.6094276905059814,
+      "learning_rate": 0.0003708343227199136,
+      "loss": 3.444,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8328840970350404,
+      "grad_norm": 0.618259072303772,
+      "learning_rate": 0.0003705105234754452,
+      "loss": 3.4502,
+      "step": 35550
+    },
+    {
+      "epoch": 3.838274932614555,
+      "grad_norm": 0.6385563015937805,
+      "learning_rate": 0.0003701867242309768,
+      "loss": 3.4472,
+      "step": 35600
+    },
+    {
+      "epoch": 3.8436657681940702,
+      "grad_norm": 0.6820311546325684,
+      "learning_rate": 0.00036986292498650833,
+      "loss": 3.4482,
+      "step": 35650
+    },
+    {
+      "epoch": 3.849056603773585,
+      "grad_norm": 0.6169026494026184,
+      "learning_rate": 0.0003695391257420399,
+      "loss": 3.4487,
+      "step": 35700
+    },
+    {
+      "epoch": 3.8544474393530996,
+      "grad_norm": 0.6038658022880554,
+      "learning_rate": 0.0003692153264975715,
+      "loss": 3.4566,
+      "step": 35750
+    },
+    {
+      "epoch": 3.8598382749326143,
+      "grad_norm": 0.6171316504478455,
+      "learning_rate": 0.00036889152725310304,
+      "loss": 3.4453,
+      "step": 35800
+    },
+    {
+      "epoch": 3.8652291105121295,
+      "grad_norm": 0.6223262548446655,
+      "learning_rate": 0.00036856772800863464,
+      "loss": 3.4441,
+      "step": 35850
+    },
+    {
+      "epoch": 3.870619946091644,
+      "grad_norm": 0.6220224499702454,
+      "learning_rate": 0.0003682439287641662,
+      "loss": 3.4563,
+      "step": 35900
+    },
+    {
+      "epoch": 3.876010781671159,
+      "grad_norm": 0.5950342416763306,
+      "learning_rate": 0.0003679201295196978,
+      "loss": 3.4536,
+      "step": 35950
+    },
+    {
+      "epoch": 3.881401617250674,
+      "grad_norm": 0.6365296244621277,
+      "learning_rate": 0.00036759633027522935,
+      "loss": 3.4513,
+      "step": 36000
+    },
+    {
+      "epoch": 3.881401617250674,
+      "eval_accuracy": 0.37520667137423414,
+      "eval_loss": 3.452235221862793,
+      "eval_runtime": 181.5679,
+      "eval_samples_per_second": 99.197,
+      "eval_steps_per_second": 6.202,
+      "step": 36000
+    },
+    {
+      "epoch": 3.8867924528301887,
+      "grad_norm": 0.6023239493370056,
+      "learning_rate": 0.00036727253103076084,
+      "loss": 3.4493,
+      "step": 36050
+    },
+    {
+      "epoch": 3.8921832884097034,
+      "grad_norm": 0.5765787959098816,
+      "learning_rate": 0.00036694873178629245,
+      "loss": 3.458,
+      "step": 36100
+    },
+    {
+      "epoch": 3.8975741239892185,
+      "grad_norm": 0.6466318368911743,
+      "learning_rate": 0.000366624932541824,
+      "loss": 3.4327,
+      "step": 36150
+    },
+    {
+      "epoch": 3.9029649595687332,
+      "grad_norm": 0.6745396852493286,
+      "learning_rate": 0.0003663011332973556,
+      "loss": 3.4447,
+      "step": 36200
+    },
+    {
+      "epoch": 3.908355795148248,
+      "grad_norm": 0.6407946348190308,
+      "learning_rate": 0.00036597733405288715,
+      "loss": 3.4635,
+      "step": 36250
+    },
+    {
+      "epoch": 3.913746630727763,
+      "grad_norm": 0.6124605536460876,
+      "learning_rate": 0.00036565353480841876,
+      "loss": 3.4481,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9191374663072778,
+      "grad_norm": 0.6159313321113586,
+      "learning_rate": 0.0003653297355639503,
+      "loss": 3.4557,
+      "step": 36350
+    },
+    {
+      "epoch": 3.9245283018867925,
+      "grad_norm": 0.6387877464294434,
+      "learning_rate": 0.0003650059363194819,
+      "loss": 3.4407,
+      "step": 36400
+    },
+    {
+      "epoch": 3.929919137466307,
+      "grad_norm": 0.6212127804756165,
+      "learning_rate": 0.00036468213707501347,
+      "loss": 3.4654,
+      "step": 36450
+    },
+    {
+      "epoch": 3.935309973045822,
+      "grad_norm": 0.6384363174438477,
+      "learning_rate": 0.000364358337830545,
+      "loss": 3.4575,
+      "step": 36500
+    },
+    {
+      "epoch": 3.940700808625337,
+      "grad_norm": 0.5889016389846802,
+      "learning_rate": 0.0003640345385860766,
+      "loss": 3.4574,
+      "step": 36550
+    },
+    {
+      "epoch": 3.9460916442048517,
+      "grad_norm": 0.6244920492172241,
+      "learning_rate": 0.00036371073934160817,
+      "loss": 3.4584,
+      "step": 36600
+    },
+    {
+      "epoch": 3.9514824797843664,
+      "grad_norm": 0.6703576445579529,
+      "learning_rate": 0.0003633869400971398,
+      "loss": 3.4541,
+      "step": 36650
+    },
+    {
+      "epoch": 3.9568733153638815,
+      "grad_norm": 0.720070481300354,
+      "learning_rate": 0.00036306961683756073,
+      "loss": 3.4476,
+      "step": 36700
+    },
+    {
+      "epoch": 3.9622641509433962,
+      "grad_norm": 0.6241286993026733,
+      "learning_rate": 0.0003627458175930922,
+      "loss": 3.4797,
+      "step": 36750
+    },
+    {
+      "epoch": 3.967654986522911,
+      "grad_norm": 0.637942373752594,
+      "learning_rate": 0.0003624220183486238,
+      "loss": 3.4629,
+      "step": 36800
+    },
+    {
+      "epoch": 3.973045822102426,
+      "grad_norm": 0.6324536800384521,
+      "learning_rate": 0.0003620982191041554,
+      "loss": 3.4389,
+      "step": 36850
+    },
+    {
+      "epoch": 3.9784366576819408,
+      "grad_norm": 0.6562104821205139,
+      "learning_rate": 0.00036177441985968693,
+      "loss": 3.4525,
+      "step": 36900
+    },
+    {
+      "epoch": 3.9838274932614555,
+      "grad_norm": 0.634568452835083,
+      "learning_rate": 0.00036145062061521854,
+      "loss": 3.4295,
+      "step": 36950
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "grad_norm": 0.6241843104362488,
+      "learning_rate": 0.0003611268213707501,
+      "loss": 3.4411,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9892183288409706,
+      "eval_accuracy": 0.3758031757283356,
+      "eval_loss": 3.4459524154663086,
+      "eval_runtime": 181.806,
+      "eval_samples_per_second": 99.067,
+      "eval_steps_per_second": 6.193,
+      "step": 37000
+    },
+    {
+      "epoch": 3.9946091644204853,
+      "grad_norm": 0.5959682464599609,
+      "learning_rate": 0.0003608030221262817,
+      "loss": 3.4464,
+      "step": 37050
+    },
+    {
+      "epoch": 4.0,
+      "grad_norm": 1.1611329317092896,
+      "learning_rate": 0.00036047922288181324,
+      "loss": 3.4311,
+      "step": 37100
+    },
+    {
+      "epoch": 4.005390835579515,
+      "grad_norm": 0.6404488682746887,
+      "learning_rate": 0.00036015542363734485,
+      "loss": 3.3566,
+      "step": 37150
+    },
+    {
+      "epoch": 4.010781671159029,
+      "grad_norm": 0.6125085949897766,
+      "learning_rate": 0.0003598316243928764,
+      "loss": 3.3577,
+      "step": 37200
+    },
+    {
+      "epoch": 4.0161725067385445,
+      "grad_norm": 0.6174116730690002,
+      "learning_rate": 0.00035950782514840795,
+      "loss": 3.3508,
+      "step": 37250
+    },
+    {
+      "epoch": 4.02156334231806,
+      "grad_norm": 0.6580213308334351,
+      "learning_rate": 0.00035918402590393955,
+      "loss": 3.3726,
+      "step": 37300
+    },
+    {
+      "epoch": 4.026954177897574,
+      "grad_norm": 0.6208986043930054,
+      "learning_rate": 0.00035886022665947105,
+      "loss": 3.3577,
+      "step": 37350
+    },
+    {
+      "epoch": 4.032345013477089,
+      "grad_norm": 0.6044255495071411,
+      "learning_rate": 0.0003585364274150027,
+      "loss": 3.353,
+      "step": 37400
+    },
+    {
+      "epoch": 4.037735849056604,
+      "grad_norm": 0.660776674747467,
+      "learning_rate": 0.0003582126281705342,
+      "loss": 3.3675,
+      "step": 37450
+    },
+    {
+      "epoch": 4.0431266846361185,
+      "grad_norm": 0.6297988891601562,
+      "learning_rate": 0.0003578888289260658,
+      "loss": 3.3677,
+      "step": 37500
+    },
+    {
+      "epoch": 4.048517520215634,
+      "grad_norm": 0.6227061748504639,
+      "learning_rate": 0.00035756502968159736,
+      "loss": 3.3648,
+      "step": 37550
+    },
+    {
+      "epoch": 4.053908355795148,
+      "grad_norm": 0.6688127517700195,
+      "learning_rate": 0.00035724123043712896,
+      "loss": 3.3517,
+      "step": 37600
+    },
+    {
+      "epoch": 4.059299191374663,
+      "grad_norm": 0.6485785841941833,
+      "learning_rate": 0.0003569174311926605,
+      "loss": 3.3742,
+      "step": 37650
+    },
+    {
+      "epoch": 4.064690026954178,
+      "grad_norm": 0.6131174564361572,
+      "learning_rate": 0.00035659363194819206,
+      "loss": 3.3793,
+      "step": 37700
+    },
+    {
+      "epoch": 4.070080862533692,
+      "grad_norm": 0.665742814540863,
+      "learning_rate": 0.00035626983270372367,
+      "loss": 3.3788,
+      "step": 37750
+    },
+    {
+      "epoch": 4.0754716981132075,
+      "grad_norm": 0.6330723166465759,
+      "learning_rate": 0.0003559460334592552,
+      "loss": 3.3495,
+      "step": 37800
+    },
+    {
+      "epoch": 4.080862533692723,
+      "grad_norm": 0.6350688934326172,
+      "learning_rate": 0.0003556222342147868,
+      "loss": 3.3652,
+      "step": 37850
+    },
+    {
+      "epoch": 4.086253369272237,
+      "grad_norm": 0.6854297518730164,
+      "learning_rate": 0.0003552984349703184,
+      "loss": 3.3649,
+      "step": 37900
+    },
+    {
+      "epoch": 4.091644204851752,
+      "grad_norm": 0.618898868560791,
+      "learning_rate": 0.00035497463572585,
+      "loss": 3.3665,
+      "step": 37950
+    },
+    {
+      "epoch": 4.097035040431267,
+      "grad_norm": 0.6617404818534851,
+      "learning_rate": 0.00035465083648138153,
+      "loss": 3.3781,
+      "step": 38000
+    },
+    {
+      "epoch": 4.097035040431267,
+      "eval_accuracy": 0.37639392147938294,
+      "eval_loss": 3.4480299949645996,
+      "eval_runtime": 181.7693,
+      "eval_samples_per_second": 99.087,
+      "eval_steps_per_second": 6.195,
+      "step": 38000
+    },
+    {
+      "epoch": 4.1024258760107815,
+      "grad_norm": 0.6297301054000854,
+      "learning_rate": 0.00035432703723691314,
+      "loss": 3.373,
+      "step": 38050
+    },
+    {
+      "epoch": 4.107816711590297,
+      "grad_norm": 0.6363517045974731,
+      "learning_rate": 0.00035400323799244463,
+      "loss": 3.3811,
+      "step": 38100
+    },
+    {
+      "epoch": 4.113207547169812,
+      "grad_norm": 0.6690065860748291,
+      "learning_rate": 0.0003536794387479762,
+      "loss": 3.3791,
+      "step": 38150
+    },
+    {
+      "epoch": 4.118598382749326,
+      "grad_norm": 0.5962274074554443,
+      "learning_rate": 0.0003533556395035078,
+      "loss": 3.3629,
+      "step": 38200
+    },
+    {
+      "epoch": 4.123989218328841,
+      "grad_norm": 0.6696664690971375,
+      "learning_rate": 0.00035303184025903934,
+      "loss": 3.3706,
+      "step": 38250
+    },
+    {
+      "epoch": 4.129380053908355,
+      "grad_norm": 0.6310851573944092,
+      "learning_rate": 0.00035270804101457094,
+      "loss": 3.3778,
+      "step": 38300
+    },
+    {
+      "epoch": 4.1347708894878705,
+      "grad_norm": 0.6481786370277405,
+      "learning_rate": 0.0003523842417701025,
+      "loss": 3.3564,
+      "step": 38350
+    },
+    {
+      "epoch": 4.140161725067386,
+      "grad_norm": 0.5998034477233887,
+      "learning_rate": 0.0003520604425256341,
+      "loss": 3.3726,
+      "step": 38400
+    },
+    {
+      "epoch": 4.1455525606469,
+      "grad_norm": 0.642882227897644,
+      "learning_rate": 0.00035173664328116565,
+      "loss": 3.3581,
+      "step": 38450
+    },
+    {
+      "epoch": 4.150943396226415,
+      "grad_norm": 0.5714357495307922,
+      "learning_rate": 0.00035141284403669725,
+      "loss": 3.3907,
+      "step": 38500
+    },
+    {
+      "epoch": 4.15633423180593,
+      "grad_norm": 0.6234452128410339,
+      "learning_rate": 0.0003510890447922288,
+      "loss": 3.3647,
+      "step": 38550
+    },
+    {
+      "epoch": 4.1617250673854445,
+      "grad_norm": 0.6253894567489624,
+      "learning_rate": 0.00035076524554776035,
+      "loss": 3.3738,
+      "step": 38600
+    },
+    {
+      "epoch": 4.16711590296496,
+      "grad_norm": 0.6415000557899475,
+      "learning_rate": 0.00035044144630329196,
+      "loss": 3.3739,
+      "step": 38650
+    },
+    {
+      "epoch": 4.172506738544475,
+      "grad_norm": 0.6307880282402039,
+      "learning_rate": 0.00035011764705882346,
+      "loss": 3.3956,
+      "step": 38700
+    },
+    {
+      "epoch": 4.177897574123989,
+      "grad_norm": 0.6561170816421509,
+      "learning_rate": 0.0003497938478143551,
+      "loss": 3.3671,
+      "step": 38750
+    },
+    {
+      "epoch": 4.183288409703504,
+      "grad_norm": 0.6375702619552612,
+      "learning_rate": 0.0003494700485698866,
+      "loss": 3.3902,
+      "step": 38800
+    },
+    {
+      "epoch": 4.188679245283019,
+      "grad_norm": 0.6860988140106201,
+      "learning_rate": 0.0003491462493254182,
+      "loss": 3.3791,
+      "step": 38850
+    },
+    {
+      "epoch": 4.1940700808625335,
+      "grad_norm": 0.6401588320732117,
+      "learning_rate": 0.00034882245008094977,
+      "loss": 3.3893,
+      "step": 38900
+    },
+    {
+      "epoch": 4.199460916442049,
+      "grad_norm": 0.6353194713592529,
+      "learning_rate": 0.0003484986508364813,
+      "loss": 3.3833,
+      "step": 38950
+    },
+    {
+      "epoch": 4.204851752021563,
+      "grad_norm": 0.6985985040664673,
+      "learning_rate": 0.0003481748515920129,
+      "loss": 3.3767,
+      "step": 39000
+    },
+    {
+      "epoch": 4.204851752021563,
+      "eval_accuracy": 0.3767065158376252,
+      "eval_loss": 3.442295789718628,
+      "eval_runtime": 181.7129,
+      "eval_samples_per_second": 99.118,
+      "eval_steps_per_second": 6.197,
+      "step": 39000
+    },
+    {
+      "epoch": 4.210242587601078,
+      "grad_norm": 0.6564880609512329,
+      "learning_rate": 0.00034785105234754447,
+      "loss": 3.3834,
+      "step": 39050
+    },
+    {
+      "epoch": 4.215633423180593,
+      "grad_norm": 0.6590373516082764,
+      "learning_rate": 0.0003475272531030761,
+      "loss": 3.3852,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2210242587601075,
+      "grad_norm": 0.63273686170578,
+      "learning_rate": 0.00034720345385860763,
+      "loss": 3.3931,
+      "step": 39150
+    },
+    {
+      "epoch": 4.226415094339623,
+      "grad_norm": 0.6495418548583984,
+      "learning_rate": 0.00034687965461413923,
+      "loss": 3.389,
+      "step": 39200
+    },
+    {
+      "epoch": 4.231805929919138,
+      "grad_norm": 0.6506909728050232,
+      "learning_rate": 0.0003465558553696708,
+      "loss": 3.3861,
+      "step": 39250
+    },
+    {
+      "epoch": 4.237196765498652,
+      "grad_norm": 0.6353509426116943,
+      "learning_rate": 0.0003462320561252024,
+      "loss": 3.3829,
+      "step": 39300
+    },
+    {
+      "epoch": 4.242587601078167,
+      "grad_norm": 0.6566866040229797,
+      "learning_rate": 0.00034590825688073394,
+      "loss": 3.3971,
+      "step": 39350
+    },
+    {
+      "epoch": 4.247978436657682,
+      "grad_norm": 0.6250466108322144,
+      "learning_rate": 0.00034558445763626543,
+      "loss": 3.3818,
+      "step": 39400
+    },
+    {
+      "epoch": 4.2533692722371965,
+      "grad_norm": 0.6622294783592224,
+      "learning_rate": 0.00034526065839179704,
+      "loss": 3.3849,
+      "step": 39450
+    },
+    {
+      "epoch": 4.258760107816712,
+      "grad_norm": 0.6352642774581909,
+      "learning_rate": 0.0003449368591473286,
+      "loss": 3.3847,
+      "step": 39500
+    },
+    {
+      "epoch": 4.264150943396227,
+      "grad_norm": 0.6231552958488464,
+      "learning_rate": 0.0003446130599028602,
+      "loss": 3.3864,
+      "step": 39550
+    },
+    {
+      "epoch": 4.269541778975741,
+      "grad_norm": 0.6041277647018433,
+      "learning_rate": 0.00034428926065839174,
+      "loss": 3.3811,
+      "step": 39600
+    },
+    {
+      "epoch": 4.274932614555256,
+      "grad_norm": 0.6078187227249146,
+      "learning_rate": 0.00034396546141392335,
+      "loss": 3.3788,
+      "step": 39650
+    },
+    {
+      "epoch": 4.280323450134771,
+      "grad_norm": 0.6268119812011719,
+      "learning_rate": 0.0003436416621694549,
+      "loss": 3.3947,
+      "step": 39700
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 0.6630831956863403,
+      "learning_rate": 0.0003433178629249865,
+      "loss": 3.3834,
+      "step": 39750
+    },
+    {
+      "epoch": 4.291105121293801,
+      "grad_norm": 0.6812276244163513,
+      "learning_rate": 0.00034299406368051806,
+      "loss": 3.3911,
+      "step": 39800
+    },
+    {
+      "epoch": 4.296495956873315,
+      "grad_norm": 0.6586547493934631,
+      "learning_rate": 0.0003426702644360496,
+      "loss": 3.3586,
+      "step": 39850
+    },
+    {
+      "epoch": 4.30188679245283,
+      "grad_norm": 0.6651487350463867,
+      "learning_rate": 0.0003423464651915812,
+      "loss": 3.3987,
+      "step": 39900
+    },
+    {
+      "epoch": 4.307277628032345,
+      "grad_norm": 0.6707507371902466,
+      "learning_rate": 0.00034202266594711276,
+      "loss": 3.3748,
+      "step": 39950
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "grad_norm": 0.6894720792770386,
+      "learning_rate": 0.00034169886670264437,
+      "loss": 3.3947,
+      "step": 40000
+    },
+    {
+      "epoch": 4.3126684636118595,
+      "eval_accuracy": 0.3773171450671426,
+      "eval_loss": 3.438546895980835,
+      "eval_runtime": 181.5176,
+      "eval_samples_per_second": 99.225,
+      "eval_steps_per_second": 6.203,
+      "step": 40000
+    },
+    {
+      "epoch": 4.318059299191375,
+      "grad_norm": 0.5998165607452393,
+      "learning_rate": 0.0003413815434430653,
+      "loss": 3.4067,
+      "step": 40050
+    },
+    {
+      "epoch": 4.32345013477089,
+      "grad_norm": 0.6333771347999573,
+      "learning_rate": 0.0003410577441985968,
+      "loss": 3.3767,
+      "step": 40100
+    },
+    {
+      "epoch": 4.328840970350404,
+      "grad_norm": 0.6432942152023315,
+      "learning_rate": 0.00034073394495412837,
+      "loss": 3.3926,
+      "step": 40150
+    },
+    {
+      "epoch": 4.334231805929919,
+      "grad_norm": 0.6244867444038391,
+      "learning_rate": 0.00034041014570965997,
+      "loss": 3.4092,
+      "step": 40200
+    },
+    {
+      "epoch": 4.339622641509434,
+      "grad_norm": 0.624860405921936,
+      "learning_rate": 0.0003400863464651915,
+      "loss": 3.3898,
+      "step": 40250
+    },
+    {
+      "epoch": 4.345013477088949,
+      "grad_norm": 0.6423871517181396,
+      "learning_rate": 0.0003397625472207231,
+      "loss": 3.4,
+      "step": 40300
+    },
+    {
+      "epoch": 4.350404312668464,
+      "grad_norm": 0.6460752487182617,
+      "learning_rate": 0.0003394387479762547,
+      "loss": 3.4063,
+      "step": 40350
+    },
+    {
+      "epoch": 4.355795148247978,
+      "grad_norm": 0.6665635108947754,
+      "learning_rate": 0.0003391149487317863,
+      "loss": 3.3966,
+      "step": 40400
+    },
+    {
+      "epoch": 4.361185983827493,
+      "grad_norm": 0.6376999020576477,
+      "learning_rate": 0.00033879114948731783,
+      "loss": 3.4036,
+      "step": 40450
+    },
+    {
+      "epoch": 4.366576819407008,
+      "grad_norm": 0.6386997103691101,
+      "learning_rate": 0.00033846735024284944,
+      "loss": 3.3693,
+      "step": 40500
+    },
+    {
+      "epoch": 4.3719676549865225,
+      "grad_norm": 0.6432493329048157,
+      "learning_rate": 0.000338143550998381,
+      "loss": 3.3786,
+      "step": 40550
+    },
+    {
+      "epoch": 4.377358490566038,
+      "grad_norm": 0.6118904948234558,
+      "learning_rate": 0.00033781975175391254,
+      "loss": 3.3857,
+      "step": 40600
+    },
+    {
+      "epoch": 4.382749326145553,
+      "grad_norm": 0.6847822666168213,
+      "learning_rate": 0.00033749595250944414,
+      "loss": 3.3869,
+      "step": 40650
+    },
+    {
+      "epoch": 4.388140161725067,
+      "grad_norm": 0.6140970587730408,
+      "learning_rate": 0.00033717215326497564,
+      "loss": 3.3911,
+      "step": 40700
+    },
+    {
+      "epoch": 4.393530997304582,
+      "grad_norm": 0.6901987195014954,
+      "learning_rate": 0.0003368483540205073,
+      "loss": 3.3914,
+      "step": 40750
+    },
+    {
+      "epoch": 4.398921832884097,
+      "grad_norm": 0.5972651243209839,
+      "learning_rate": 0.0003365245547760388,
+      "loss": 3.3906,
+      "step": 40800
+    },
+    {
+      "epoch": 4.404312668463612,
+      "grad_norm": 0.7040302157402039,
+      "learning_rate": 0.0003362007555315704,
+      "loss": 3.3822,
+      "step": 40850
+    },
+    {
+      "epoch": 4.409703504043127,
+      "grad_norm": 0.6535053849220276,
+      "learning_rate": 0.00033587695628710195,
+      "loss": 3.3886,
+      "step": 40900
+    },
+    {
+      "epoch": 4.415094339622642,
+      "grad_norm": 0.6086871027946472,
+      "learning_rate": 0.00033555315704263355,
+      "loss": 3.3965,
+      "step": 40950
+    },
+    {
+      "epoch": 4.420485175202156,
+      "grad_norm": 0.6123448610305786,
+      "learning_rate": 0.0003352293577981651,
+      "loss": 3.389,
+      "step": 41000
+    },
+    {
+      "epoch": 4.420485175202156,
+      "eval_accuracy": 0.37761822221927654,
+      "eval_loss": 3.433894395828247,
+      "eval_runtime": 181.957,
+      "eval_samples_per_second": 98.985,
+      "eval_steps_per_second": 6.188,
+      "step": 41000
+    },
+    {
+      "epoch": 4.425876010781671,
+      "grad_norm": 0.6953914761543274,
+      "learning_rate": 0.00033490555855369665,
+      "loss": 3.3761,
+      "step": 41050
+    },
+    {
+      "epoch": 4.431266846361186,
+      "grad_norm": 0.6522581577301025,
+      "learning_rate": 0.00033458175930922826,
+      "loss": 3.3945,
+      "step": 41100
+    },
+    {
+      "epoch": 4.436657681940701,
+      "grad_norm": 0.6128993630409241,
+      "learning_rate": 0.0003342579600647598,
+      "loss": 3.3911,
+      "step": 41150
+    },
+    {
+      "epoch": 4.442048517520216,
+      "grad_norm": 0.6275089979171753,
+      "learning_rate": 0.0003339341608202914,
+      "loss": 3.3761,
+      "step": 41200
+    },
+    {
+      "epoch": 4.44743935309973,
+      "grad_norm": 0.6717378497123718,
+      "learning_rate": 0.00033361036157582297,
+      "loss": 3.3872,
+      "step": 41250
+    },
+    {
+      "epoch": 4.452830188679245,
+      "grad_norm": 0.6755267381668091,
+      "learning_rate": 0.00033328656233135457,
+      "loss": 3.3795,
+      "step": 41300
+    },
+    {
+      "epoch": 4.45822102425876,
+      "grad_norm": 0.6669474840164185,
+      "learning_rate": 0.0003329627630868861,
+      "loss": 3.4041,
+      "step": 41350
+    },
+    {
+      "epoch": 4.463611859838275,
+      "grad_norm": 0.6538794040679932,
+      "learning_rate": 0.0003326389638424177,
+      "loss": 3.3974,
+      "step": 41400
+    },
+    {
+      "epoch": 4.46900269541779,
+      "grad_norm": 0.6036146283149719,
+      "learning_rate": 0.0003323151645979492,
+      "loss": 3.387,
+      "step": 41450
+    },
+    {
+      "epoch": 4.474393530997305,
+      "grad_norm": 0.6463048458099365,
+      "learning_rate": 0.00033199136535348077,
+      "loss": 3.3739,
+      "step": 41500
+    },
+    {
+      "epoch": 4.479784366576819,
+      "grad_norm": 0.6582531332969666,
+      "learning_rate": 0.0003316675661090124,
+      "loss": 3.3918,
+      "step": 41550
+    },
+    {
+      "epoch": 4.485175202156334,
+      "grad_norm": 0.6958242654800415,
+      "learning_rate": 0.00033134376686454393,
+      "loss": 3.3972,
+      "step": 41600
+    },
+    {
+      "epoch": 4.490566037735849,
+      "grad_norm": 0.6295124292373657,
+      "learning_rate": 0.00033101996762007553,
+      "loss": 3.3811,
+      "step": 41650
+    },
+    {
+      "epoch": 4.495956873315364,
+      "grad_norm": 0.627613365650177,
+      "learning_rate": 0.0003306961683756071,
+      "loss": 3.4056,
+      "step": 41700
+    },
+    {
+      "epoch": 4.501347708894879,
+      "grad_norm": 0.6739616990089417,
+      "learning_rate": 0.0003303723691311387,
+      "loss": 3.3856,
+      "step": 41750
+    },
+    {
+      "epoch": 4.506738544474393,
+      "grad_norm": 0.6398614645004272,
+      "learning_rate": 0.00033004856988667024,
+      "loss": 3.3847,
+      "step": 41800
+    },
+    {
+      "epoch": 4.512129380053908,
+      "grad_norm": 0.6853240728378296,
+      "learning_rate": 0.0003297247706422018,
+      "loss": 3.3945,
+      "step": 41850
+    },
+    {
+      "epoch": 4.517520215633423,
+      "grad_norm": 0.6887792348861694,
+      "learning_rate": 0.0003294009713977334,
+      "loss": 3.3924,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5229110512129385,
+      "grad_norm": 0.6502972841262817,
+      "learning_rate": 0.00032907717215326494,
+      "loss": 3.3939,
+      "step": 41950
+    },
+    {
+      "epoch": 4.528301886792453,
+      "grad_norm": 0.6029436588287354,
+      "learning_rate": 0.00032875337290879655,
+      "loss": 3.3849,
+      "step": 42000
+    },
+    {
+      "epoch": 4.528301886792453,
+      "eval_accuracy": 0.37859935779624126,
+      "eval_loss": 3.428929090499878,
+      "eval_runtime": 181.7172,
+      "eval_samples_per_second": 99.116,
+      "eval_steps_per_second": 6.196,
+      "step": 42000
+    },
+    {
+      "epoch": 4.533692722371968,
+      "grad_norm": 0.62586510181427,
+      "learning_rate": 0.00032842957366432805,
+      "loss": 3.382,
+      "step": 42050
+    },
+    {
+      "epoch": 4.539083557951482,
+      "grad_norm": 0.6761049628257751,
+      "learning_rate": 0.0003281057744198597,
+      "loss": 3.4097,
+      "step": 42100
+    },
+    {
+      "epoch": 4.544474393530997,
+      "grad_norm": 0.6782863140106201,
+      "learning_rate": 0.0003277819751753912,
+      "loss": 3.3976,
+      "step": 42150
+    },
+    {
+      "epoch": 4.549865229110512,
+      "grad_norm": 0.6204219460487366,
+      "learning_rate": 0.0003274581759309228,
+      "loss": 3.3791,
+      "step": 42200
+    },
+    {
+      "epoch": 4.555256064690027,
+      "grad_norm": 0.8040519952774048,
+      "learning_rate": 0.00032713437668645436,
+      "loss": 3.3998,
+      "step": 42250
+    },
+    {
+      "epoch": 4.560646900269542,
+      "grad_norm": 0.699002742767334,
+      "learning_rate": 0.0003268105774419859,
+      "loss": 3.3999,
+      "step": 42300
+    },
+    {
+      "epoch": 4.566037735849057,
+      "grad_norm": 0.6923718452453613,
+      "learning_rate": 0.0003264867781975175,
+      "loss": 3.382,
+      "step": 42350
+    },
+    {
+      "epoch": 4.571428571428571,
+      "grad_norm": 0.6704239845275879,
+      "learning_rate": 0.00032616297895304906,
+      "loss": 3.4073,
+      "step": 42400
+    },
+    {
+      "epoch": 4.576819407008086,
+      "grad_norm": 0.6082397103309631,
+      "learning_rate": 0.00032583917970858067,
+      "loss": 3.3739,
+      "step": 42450
+    },
+    {
+      "epoch": 4.5822102425876015,
+      "grad_norm": 0.6278915405273438,
+      "learning_rate": 0.0003255153804641122,
+      "loss": 3.3761,
+      "step": 42500
+    },
+    {
+      "epoch": 4.587601078167116,
+      "grad_norm": 0.628350555896759,
+      "learning_rate": 0.0003251915812196438,
+      "loss": 3.3917,
+      "step": 42550
+    },
+    {
+      "epoch": 4.592991913746631,
+      "grad_norm": 0.6095200181007385,
+      "learning_rate": 0.00032486778197517537,
+      "loss": 3.387,
+      "step": 42600
+    },
+    {
+      "epoch": 4.598382749326145,
+      "grad_norm": 0.6962038278579712,
+      "learning_rate": 0.000324543982730707,
+      "loss": 3.4012,
+      "step": 42650
+    },
+    {
+      "epoch": 4.60377358490566,
+      "grad_norm": 0.6633439660072327,
+      "learning_rate": 0.00032422018348623853,
+      "loss": 3.3961,
+      "step": 42700
+    },
+    {
+      "epoch": 4.609164420485175,
+      "grad_norm": 0.6363673210144043,
+      "learning_rate": 0.0003239028602266595,
+      "loss": 3.382,
+      "step": 42750
+    },
+    {
+      "epoch": 4.6145552560646905,
+      "grad_norm": 0.6578346490859985,
+      "learning_rate": 0.000323579060982191,
+      "loss": 3.4043,
+      "step": 42800
+    },
+    {
+      "epoch": 4.619946091644205,
+      "grad_norm": 0.6377918124198914,
+      "learning_rate": 0.0003232552617377226,
+      "loss": 3.3988,
+      "step": 42850
+    },
+    {
+      "epoch": 4.62533692722372,
+      "grad_norm": 0.5883036851882935,
+      "learning_rate": 0.00032293146249325413,
+      "loss": 3.3863,
+      "step": 42900
+    },
+    {
+      "epoch": 4.630727762803234,
+      "grad_norm": 0.6473262906074524,
+      "learning_rate": 0.00032260766324878574,
+      "loss": 3.3877,
+      "step": 42950
+    },
+    {
+      "epoch": 4.636118598382749,
+      "grad_norm": 0.6402585506439209,
+      "learning_rate": 0.0003222838640043173,
+      "loss": 3.4041,
+      "step": 43000
+    },
+    {
+      "epoch": 4.636118598382749,
+      "eval_accuracy": 0.37907547475064435,
+      "eval_loss": 3.424429178237915,
+      "eval_runtime": 182.0489,
+      "eval_samples_per_second": 98.935,
+      "eval_steps_per_second": 6.185,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6415094339622645,
+      "grad_norm": 0.6390340924263,
+      "learning_rate": 0.00032196006475984884,
+      "loss": 3.3832,
+      "step": 43050
+    },
+    {
+      "epoch": 4.646900269541779,
+      "grad_norm": 0.6697818636894226,
+      "learning_rate": 0.00032163626551538044,
+      "loss": 3.3847,
+      "step": 43100
+    },
+    {
+      "epoch": 4.652291105121294,
+      "grad_norm": 0.6102224588394165,
+      "learning_rate": 0.000321312466270912,
+      "loss": 3.3822,
+      "step": 43150
+    },
+    {
+      "epoch": 4.657681940700809,
+      "grad_norm": 0.6941066384315491,
+      "learning_rate": 0.0003209886670264436,
+      "loss": 3.3905,
+      "step": 43200
+    },
+    {
+      "epoch": 4.663072776280323,
+      "grad_norm": 0.6344305276870728,
+      "learning_rate": 0.00032066486778197515,
+      "loss": 3.3888,
+      "step": 43250
+    },
+    {
+      "epoch": 4.668463611859838,
+      "grad_norm": 0.6397703886032104,
+      "learning_rate": 0.00032034106853750675,
+      "loss": 3.3968,
+      "step": 43300
+    },
+    {
+      "epoch": 4.6738544474393535,
+      "grad_norm": 0.615652322769165,
+      "learning_rate": 0.0003200172692930383,
+      "loss": 3.3742,
+      "step": 43350
+    },
+    {
+      "epoch": 4.679245283018868,
+      "grad_norm": 0.6831540465354919,
+      "learning_rate": 0.0003196934700485699,
+      "loss": 3.383,
+      "step": 43400
+    },
+    {
+      "epoch": 4.684636118598383,
+      "grad_norm": 0.670258104801178,
+      "learning_rate": 0.0003193696708041014,
+      "loss": 3.391,
+      "step": 43450
+    },
+    {
+      "epoch": 4.690026954177897,
+      "grad_norm": 0.6558921337127686,
+      "learning_rate": 0.00031904587155963296,
+      "loss": 3.3886,
+      "step": 43500
+    },
+    {
+      "epoch": 4.695417789757412,
+      "grad_norm": 0.6475565433502197,
+      "learning_rate": 0.00031872207231516456,
+      "loss": 3.3769,
+      "step": 43550
+    },
+    {
+      "epoch": 4.7008086253369274,
+      "grad_norm": 0.7060928344726562,
+      "learning_rate": 0.0003183982730706961,
+      "loss": 3.3797,
+      "step": 43600
+    },
+    {
+      "epoch": 4.706199460916442,
+      "grad_norm": 0.6488223671913147,
+      "learning_rate": 0.0003180744738262277,
+      "loss": 3.3987,
+      "step": 43650
+    },
+    {
+      "epoch": 4.711590296495957,
+      "grad_norm": 0.651401937007904,
+      "learning_rate": 0.00031775067458175927,
+      "loss": 3.389,
+      "step": 43700
+    },
+    {
+      "epoch": 4.716981132075472,
+      "grad_norm": 0.6610608100891113,
+      "learning_rate": 0.00031742687533729087,
+      "loss": 3.3873,
+      "step": 43750
+    },
+    {
+      "epoch": 4.722371967654986,
+      "grad_norm": 0.6515519618988037,
+      "learning_rate": 0.0003171030760928224,
+      "loss": 3.3807,
+      "step": 43800
+    },
+    {
+      "epoch": 4.727762803234501,
+      "grad_norm": 0.6313045620918274,
+      "learning_rate": 0.000316779276848354,
+      "loss": 3.3998,
+      "step": 43850
+    },
+    {
+      "epoch": 4.7331536388140165,
+      "grad_norm": 0.6147524118423462,
+      "learning_rate": 0.0003164554776038856,
+      "loss": 3.4078,
+      "step": 43900
+    },
+    {
+      "epoch": 4.738544474393531,
+      "grad_norm": 0.6559792160987854,
+      "learning_rate": 0.00031613167835941713,
+      "loss": 3.4056,
+      "step": 43950
+    },
+    {
+      "epoch": 4.743935309973046,
+      "grad_norm": 0.6663808822631836,
+      "learning_rate": 0.00031580787911494873,
+      "loss": 3.3693,
+      "step": 44000
+    },
+    {
+      "epoch": 4.743935309973046,
+      "eval_accuracy": 0.3795856000589333,
+      "eval_loss": 3.4179606437683105,
+      "eval_runtime": 181.5494,
+      "eval_samples_per_second": 99.207,
+      "eval_steps_per_second": 6.202,
+      "step": 44000
+    },
+    {
+      "epoch": 4.74932614555256,
+      "grad_norm": 0.6471132636070251,
+      "learning_rate": 0.00031548407987048023,
+      "loss": 3.4089,
+      "step": 44050
+    },
+    {
+      "epoch": 4.754716981132075,
+      "grad_norm": 0.6646540760993958,
+      "learning_rate": 0.0003151602806260119,
+      "loss": 3.3837,
+      "step": 44100
+    },
+    {
+      "epoch": 4.7601078167115904,
+      "grad_norm": 0.6546226143836975,
+      "learning_rate": 0.0003148364813815434,
+      "loss": 3.3773,
+      "step": 44150
+    },
+    {
+      "epoch": 4.765498652291106,
+      "grad_norm": 0.5949311852455139,
+      "learning_rate": 0.000314512682137075,
+      "loss": 3.3991,
+      "step": 44200
+    },
+    {
+      "epoch": 4.77088948787062,
+      "grad_norm": 0.6574331521987915,
+      "learning_rate": 0.00031418888289260654,
+      "loss": 3.3919,
+      "step": 44250
+    },
+    {
+      "epoch": 4.776280323450135,
+      "grad_norm": 0.6199044585227966,
+      "learning_rate": 0.0003138650836481381,
+      "loss": 3.3865,
+      "step": 44300
+    },
+    {
+      "epoch": 4.781671159029649,
+      "grad_norm": 0.6571835875511169,
+      "learning_rate": 0.0003135412844036697,
+      "loss": 3.3986,
+      "step": 44350
+    },
+    {
+      "epoch": 4.787061994609164,
+      "grad_norm": 0.6753513813018799,
+      "learning_rate": 0.00031321748515920124,
+      "loss": 3.3776,
+      "step": 44400
+    },
+    {
+      "epoch": 4.7924528301886795,
+      "grad_norm": 0.6330493092536926,
+      "learning_rate": 0.00031289368591473285,
+      "loss": 3.3942,
+      "step": 44450
+    },
+    {
+      "epoch": 4.797843665768194,
+      "grad_norm": 0.6691889762878418,
+      "learning_rate": 0.0003125698866702644,
+      "loss": 3.3777,
+      "step": 44500
+    },
+    {
+      "epoch": 4.803234501347709,
+      "grad_norm": 0.6553536653518677,
+      "learning_rate": 0.000312246087425796,
+      "loss": 3.3769,
+      "step": 44550
+    },
+    {
+      "epoch": 4.808625336927224,
+      "grad_norm": 0.648166298866272,
+      "learning_rate": 0.00031192228818132756,
+      "loss": 3.3905,
+      "step": 44600
+    },
+    {
+      "epoch": 4.814016172506738,
+      "grad_norm": 0.6351345777511597,
+      "learning_rate": 0.00031159848893685916,
+      "loss": 3.3821,
+      "step": 44650
+    },
+    {
+      "epoch": 4.819407008086253,
+      "grad_norm": 0.591360330581665,
+      "learning_rate": 0.0003112746896923907,
+      "loss": 3.3744,
+      "step": 44700
+    },
+    {
+      "epoch": 4.824797843665769,
+      "grad_norm": 0.7283137440681458,
+      "learning_rate": 0.0003109508904479222,
+      "loss": 3.3773,
+      "step": 44750
+    },
+    {
+      "epoch": 4.830188679245283,
+      "grad_norm": 0.6123622059822083,
+      "learning_rate": 0.0003106270912034538,
+      "loss": 3.391,
+      "step": 44800
+    },
+    {
+      "epoch": 4.835579514824798,
+      "grad_norm": 0.6461575031280518,
+      "learning_rate": 0.00031030329195898536,
+      "loss": 3.3865,
+      "step": 44850
+    },
+    {
+      "epoch": 4.840970350404312,
+      "grad_norm": 0.6624020934104919,
+      "learning_rate": 0.00030997949271451697,
+      "loss": 3.3887,
+      "step": 44900
+    },
+    {
+      "epoch": 4.846361185983827,
+      "grad_norm": 0.6682701110839844,
+      "learning_rate": 0.0003096556934700485,
+      "loss": 3.3842,
+      "step": 44950
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "grad_norm": 0.6075751781463623,
+      "learning_rate": 0.0003093318942255801,
+      "loss": 3.3953,
+      "step": 45000
+    },
+    {
+      "epoch": 4.8517520215633425,
+      "eval_accuracy": 0.379875703269398,
+      "eval_loss": 3.41325306892395,
+      "eval_runtime": 182.0922,
+      "eval_samples_per_second": 98.911,
+      "eval_steps_per_second": 6.184,
+      "step": 45000
+    },
+    {
+      "epoch": 4.857142857142857,
+      "grad_norm": 0.618019163608551,
+      "learning_rate": 0.0003090080949811117,
+      "loss": 3.3711,
+      "step": 45050
+    },
+    {
+      "epoch": 4.862533692722372,
+      "grad_norm": 0.6713760495185852,
+      "learning_rate": 0.0003086842957366433,
+      "loss": 3.3965,
+      "step": 45100
+    },
+    {
+      "epoch": 4.867924528301887,
+      "grad_norm": 0.7070072889328003,
+      "learning_rate": 0.00030836049649217483,
+      "loss": 3.3691,
+      "step": 45150
+    },
+    {
+      "epoch": 4.873315363881401,
+      "grad_norm": 0.6715174913406372,
+      "learning_rate": 0.0003080431732325958,
+      "loss": 3.3933,
+      "step": 45200
+    },
+    {
+      "epoch": 4.878706199460916,
+      "grad_norm": 0.6704628467559814,
+      "learning_rate": 0.00030771937398812733,
+      "loss": 3.3891,
+      "step": 45250
+    },
+    {
+      "epoch": 4.884097035040432,
+      "grad_norm": 0.6364473700523376,
+      "learning_rate": 0.00030739557474365894,
+      "loss": 3.394,
+      "step": 45300
+    },
+    {
+      "epoch": 4.889487870619946,
+      "grad_norm": 0.6634293794631958,
+      "learning_rate": 0.0003070717754991905,
+      "loss": 3.3871,
+      "step": 45350
+    },
+    {
+      "epoch": 4.894878706199461,
+      "grad_norm": 0.6384555101394653,
+      "learning_rate": 0.0003067479762547221,
+      "loss": 3.3828,
+      "step": 45400
+    },
+    {
+      "epoch": 4.900269541778976,
+      "grad_norm": 0.6213516592979431,
+      "learning_rate": 0.0003064241770102536,
+      "loss": 3.3905,
+      "step": 45450
+    },
+    {
+      "epoch": 4.90566037735849,
+      "grad_norm": 0.6964032649993896,
+      "learning_rate": 0.00030610037776578514,
+      "loss": 3.3905,
+      "step": 45500
+    },
+    {
+      "epoch": 4.9110512129380055,
+      "grad_norm": 0.6544898152351379,
+      "learning_rate": 0.00030577657852131674,
+      "loss": 3.3783,
+      "step": 45550
+    },
+    {
+      "epoch": 4.916442048517521,
+      "grad_norm": 0.6489967107772827,
+      "learning_rate": 0.0003054527792768483,
+      "loss": 3.4039,
+      "step": 45600
+    },
+    {
+      "epoch": 4.921832884097035,
+      "grad_norm": 0.6411947011947632,
+      "learning_rate": 0.0003051289800323799,
+      "loss": 3.376,
+      "step": 45650
+    },
+    {
+      "epoch": 4.92722371967655,
+      "grad_norm": 0.624286949634552,
+      "learning_rate": 0.00030480518078791145,
+      "loss": 3.3895,
+      "step": 45700
+    },
+    {
+      "epoch": 4.932614555256064,
+      "grad_norm": 0.6586293578147888,
+      "learning_rate": 0.00030448138154344305,
+      "loss": 3.3934,
+      "step": 45750
+    },
+    {
+      "epoch": 4.938005390835579,
+      "grad_norm": 0.683337390422821,
+      "learning_rate": 0.0003041575822989746,
+      "loss": 3.3861,
+      "step": 45800
+    },
+    {
+      "epoch": 4.943396226415095,
+      "grad_norm": 0.6735714673995972,
+      "learning_rate": 0.0003038337830545062,
+      "loss": 3.3883,
+      "step": 45850
+    },
+    {
+      "epoch": 4.948787061994609,
+      "grad_norm": 0.687961995601654,
+      "learning_rate": 0.00030350998381003776,
+      "loss": 3.3902,
+      "step": 45900
+    },
+    {
+      "epoch": 4.954177897574124,
+      "grad_norm": 0.6364692449569702,
+      "learning_rate": 0.0003031861845655693,
+      "loss": 3.3726,
+      "step": 45950
+    },
+    {
+      "epoch": 4.959568733153639,
+      "grad_norm": 0.6474451422691345,
+      "learning_rate": 0.0003028623853211009,
+      "loss": 3.3841,
+      "step": 46000
+    },
+    {
+      "epoch": 4.959568733153639,
+      "eval_accuracy": 0.3804872017220179,
+      "eval_loss": 3.409752368927002,
+      "eval_runtime": 181.5189,
+      "eval_samples_per_second": 99.224,
+      "eval_steps_per_second": 6.203,
+      "step": 46000
+    },
+    {
+      "epoch": 4.964959568733153,
+      "grad_norm": 0.630465030670166,
+      "learning_rate": 0.00030253858607663247,
+      "loss": 3.3909,
+      "step": 46050
+    },
+    {
+      "epoch": 4.9703504043126685,
+      "grad_norm": 0.655409038066864,
+      "learning_rate": 0.00030221478683216407,
+      "loss": 3.3874,
+      "step": 46100
+    },
+    {
+      "epoch": 4.975741239892184,
+      "grad_norm": 0.6540499925613403,
+      "learning_rate": 0.00030189098758769557,
+      "loss": 3.3845,
+      "step": 46150
+    },
+    {
+      "epoch": 4.981132075471698,
+      "grad_norm": 0.6544473171234131,
+      "learning_rate": 0.00030156718834322717,
+      "loss": 3.3843,
+      "step": 46200
+    },
+    {
+      "epoch": 4.986522911051213,
+      "grad_norm": 0.6690456867218018,
+      "learning_rate": 0.0003012433890987587,
+      "loss": 3.4047,
+      "step": 46250
+    },
+    {
+      "epoch": 4.991913746630727,
+      "grad_norm": 0.6413595080375671,
+      "learning_rate": 0.0003009195898542903,
+      "loss": 3.3936,
+      "step": 46300
+    },
+    {
+      "epoch": 4.997304582210242,
+      "grad_norm": 0.6787312030792236,
+      "learning_rate": 0.0003005957906098219,
+      "loss": 3.382,
+      "step": 46350
+    },
+    {
+      "epoch": 5.002695417789758,
+      "grad_norm": 0.6582103371620178,
+      "learning_rate": 0.00030027199136535343,
+      "loss": 3.3388,
+      "step": 46400
+    },
+    {
+      "epoch": 5.008086253369272,
+      "grad_norm": 0.6663526296615601,
+      "learning_rate": 0.00029994819212088503,
+      "loss": 3.288,
+      "step": 46450
+    },
+    {
+      "epoch": 5.013477088948787,
+      "grad_norm": 0.6610379219055176,
+      "learning_rate": 0.0002996243928764166,
+      "loss": 3.2923,
+      "step": 46500
+    },
+    {
+      "epoch": 5.018867924528302,
+      "grad_norm": 0.652783989906311,
+      "learning_rate": 0.0002993005936319482,
+      "loss": 3.3012,
+      "step": 46550
+    },
+    {
+      "epoch": 5.024258760107816,
+      "grad_norm": 0.6830565333366394,
+      "learning_rate": 0.00029897679438747974,
+      "loss": 3.299,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0296495956873315,
+      "grad_norm": 0.7056826949119568,
+      "learning_rate": 0.0002986529951430113,
+      "loss": 3.3031,
+      "step": 46650
+    },
+    {
+      "epoch": 5.035040431266847,
+      "grad_norm": 0.6342096924781799,
+      "learning_rate": 0.0002983291958985429,
+      "loss": 3.2964,
+      "step": 46700
+    },
+    {
+      "epoch": 5.040431266846361,
+      "grad_norm": 0.6923931241035461,
+      "learning_rate": 0.00029800539665407444,
+      "loss": 3.3102,
+      "step": 46750
+    },
+    {
+      "epoch": 5.045822102425876,
+      "grad_norm": 0.677852988243103,
+      "learning_rate": 0.000297681597409606,
+      "loss": 3.2961,
+      "step": 46800
+    },
+    {
+      "epoch": 5.051212938005391,
+      "grad_norm": 0.6728231310844421,
+      "learning_rate": 0.0002973577981651376,
+      "loss": 3.314,
+      "step": 46850
+    },
+    {
+      "epoch": 5.056603773584905,
+      "grad_norm": 0.6960410475730896,
+      "learning_rate": 0.00029703399892066915,
+      "loss": 3.3012,
+      "step": 46900
+    },
+    {
+      "epoch": 5.061994609164421,
+      "grad_norm": 0.6692420244216919,
+      "learning_rate": 0.00029671019967620076,
+      "loss": 3.2907,
+      "step": 46950
+    },
+    {
+      "epoch": 5.067385444743936,
+      "grad_norm": 0.672247052192688,
+      "learning_rate": 0.0002963864004317323,
+      "loss": 3.3078,
+      "step": 47000
+    },
+    {
+      "epoch": 5.067385444743936,
+      "eval_accuracy": 0.3803283512000331,
+      "eval_loss": 3.413437843322754,
+      "eval_runtime": 181.7151,
+      "eval_samples_per_second": 99.117,
+      "eval_steps_per_second": 6.197,
+      "step": 47000
+    },
+    {
+      "epoch": 5.07277628032345,
+      "grad_norm": 0.6593713164329529,
+      "learning_rate": 0.00029606260118726386,
+      "loss": 3.3107,
+      "step": 47050
+    },
+    {
+      "epoch": 5.078167115902965,
+      "grad_norm": 0.6897931694984436,
+      "learning_rate": 0.0002957388019427954,
+      "loss": 3.3144,
+      "step": 47100
+    },
+    {
+      "epoch": 5.083557951482479,
+      "grad_norm": 0.682111918926239,
+      "learning_rate": 0.000295415002698327,
+      "loss": 3.3084,
+      "step": 47150
+    },
+    {
+      "epoch": 5.0889487870619945,
+      "grad_norm": 0.6645491719245911,
+      "learning_rate": 0.00029509120345385856,
+      "loss": 3.3045,
+      "step": 47200
+    },
+    {
+      "epoch": 5.09433962264151,
+      "grad_norm": 0.7227234840393066,
+      "learning_rate": 0.00029476740420939017,
+      "loss": 3.3093,
+      "step": 47250
+    },
+    {
+      "epoch": 5.099730458221024,
+      "grad_norm": 0.6667834520339966,
+      "learning_rate": 0.0002944436049649217,
+      "loss": 3.3061,
+      "step": 47300
+    },
+    {
+      "epoch": 5.105121293800539,
+      "grad_norm": 0.6737233996391296,
+      "learning_rate": 0.0002941198057204533,
+      "loss": 3.2953,
+      "step": 47350
+    },
+    {
+      "epoch": 5.110512129380054,
+      "grad_norm": 0.6560413837432861,
+      "learning_rate": 0.00029379600647598487,
+      "loss": 3.3114,
+      "step": 47400
+    },
+    {
+      "epoch": 5.115902964959568,
+      "grad_norm": 0.669833242893219,
+      "learning_rate": 0.0002934722072315164,
+      "loss": 3.3264,
+      "step": 47450
+    },
+    {
+      "epoch": 5.121293800539084,
+      "grad_norm": 0.6529123187065125,
+      "learning_rate": 0.000293148407987048,
+      "loss": 3.3197,
+      "step": 47500
+    },
+    {
+      "epoch": 5.126684636118599,
+      "grad_norm": 0.6688790917396545,
+      "learning_rate": 0.0002928246087425796,
+      "loss": 3.3277,
+      "step": 47550
+    },
+    {
+      "epoch": 5.132075471698113,
+      "grad_norm": 0.7134573459625244,
+      "learning_rate": 0.00029250080949811113,
+      "loss": 3.3201,
+      "step": 47600
+    },
+    {
+      "epoch": 5.137466307277628,
+      "grad_norm": 0.6406720280647278,
+      "learning_rate": 0.00029217701025364273,
+      "loss": 3.3089,
+      "step": 47650
+    },
+    {
+      "epoch": 5.142857142857143,
+      "grad_norm": 0.6913286447525024,
+      "learning_rate": 0.0002918532110091743,
+      "loss": 3.3229,
+      "step": 47700
+    },
+    {
+      "epoch": 5.1482479784366575,
+      "grad_norm": 0.628642201423645,
+      "learning_rate": 0.0002915294117647059,
+      "loss": 3.3174,
+      "step": 47750
+    },
+    {
+      "epoch": 5.153638814016173,
+      "grad_norm": 0.6638661026954651,
+      "learning_rate": 0.00029120561252023744,
+      "loss": 3.3054,
+      "step": 47800
+    },
+    {
+      "epoch": 5.159029649595688,
+      "grad_norm": 0.7017157673835754,
+      "learning_rate": 0.000290881813275769,
+      "loss": 3.3304,
+      "step": 47850
+    },
+    {
+      "epoch": 5.164420485175202,
+      "grad_norm": 0.6650938987731934,
+      "learning_rate": 0.00029055801403130054,
+      "loss": 3.3221,
+      "step": 47900
+    },
+    {
+      "epoch": 5.169811320754717,
+      "grad_norm": 0.6689581871032715,
+      "learning_rate": 0.00029023421478683215,
+      "loss": 3.3087,
+      "step": 47950
+    },
+    {
+      "epoch": 5.175202156334231,
+      "grad_norm": 0.6488136649131775,
+      "learning_rate": 0.0002899104155423637,
+      "loss": 3.315,
+      "step": 48000
+    },
+    {
+      "epoch": 5.175202156334231,
+      "eval_accuracy": 0.3809863530886376,
+      "eval_loss": 3.410261631011963,
+      "eval_runtime": 181.8994,
+      "eval_samples_per_second": 99.016,
+      "eval_steps_per_second": 6.19,
+      "step": 48000
+    },
+    {
+      "epoch": 5.180592991913747,
+      "grad_norm": 0.6930067539215088,
+      "learning_rate": 0.0002895866162978953,
+      "loss": 3.3195,
+      "step": 48050
+    },
+    {
+      "epoch": 5.185983827493262,
+      "grad_norm": 0.6869164109230042,
+      "learning_rate": 0.00028926281705342685,
+      "loss": 3.3301,
+      "step": 48100
+    },
+    {
+      "epoch": 5.191374663072776,
+      "grad_norm": 0.6907550096511841,
+      "learning_rate": 0.0002889390178089584,
+      "loss": 3.3239,
+      "step": 48150
+    },
+    {
+      "epoch": 5.196765498652291,
+      "grad_norm": 0.7010151147842407,
+      "learning_rate": 0.00028861521856449,
+      "loss": 3.3341,
+      "step": 48200
+    },
+    {
+      "epoch": 5.202156334231806,
+      "grad_norm": 0.6742527484893799,
+      "learning_rate": 0.00028829141932002156,
+      "loss": 3.3272,
+      "step": 48250
+    },
+    {
+      "epoch": 5.2075471698113205,
+      "grad_norm": 0.6344496011734009,
+      "learning_rate": 0.0002879676200755531,
+      "loss": 3.3019,
+      "step": 48300
+    },
+    {
+      "epoch": 5.212938005390836,
+      "grad_norm": 0.6495826840400696,
+      "learning_rate": 0.0002876438208310847,
+      "loss": 3.3375,
+      "step": 48350
+    },
+    {
+      "epoch": 5.218328840970351,
+      "grad_norm": 0.6612682342529297,
+      "learning_rate": 0.00028732002158661626,
+      "loss": 3.3277,
+      "step": 48400
+    },
+    {
+      "epoch": 5.223719676549865,
+      "grad_norm": 0.7070847749710083,
+      "learning_rate": 0.00028699622234214787,
+      "loss": 3.3326,
+      "step": 48450
+    },
+    {
+      "epoch": 5.22911051212938,
+      "grad_norm": 0.6680747270584106,
+      "learning_rate": 0.0002866724230976794,
+      "loss": 3.3387,
+      "step": 48500
+    },
+    {
+      "epoch": 5.234501347708895,
+      "grad_norm": 0.7149842977523804,
+      "learning_rate": 0.00028634862385321097,
+      "loss": 3.3139,
+      "step": 48550
+    },
+    {
+      "epoch": 5.2398921832884096,
+      "grad_norm": 0.6795401573181152,
+      "learning_rate": 0.0002860248246087426,
+      "loss": 3.3161,
+      "step": 48600
+    },
+    {
+      "epoch": 5.245283018867925,
+      "grad_norm": 0.6816480755805969,
+      "learning_rate": 0.0002857010253642741,
+      "loss": 3.339,
+      "step": 48650
+    },
+    {
+      "epoch": 5.250673854447439,
+      "grad_norm": 0.7223415970802307,
+      "learning_rate": 0.0002853772261198057,
+      "loss": 3.3277,
+      "step": 48700
+    },
+    {
+      "epoch": 5.256064690026954,
+      "grad_norm": 0.715975284576416,
+      "learning_rate": 0.0002850534268753373,
+      "loss": 3.3273,
+      "step": 48750
+    },
+    {
+      "epoch": 5.261455525606469,
+      "grad_norm": 0.6848128437995911,
+      "learning_rate": 0.00028472962763086883,
+      "loss": 3.3383,
+      "step": 48800
+    },
+    {
+      "epoch": 5.2668463611859835,
+      "grad_norm": 0.6746169924736023,
+      "learning_rate": 0.0002844058283864004,
+      "loss": 3.3299,
+      "step": 48850
+    },
+    {
+      "epoch": 5.272237196765499,
+      "grad_norm": 0.6700478792190552,
+      "learning_rate": 0.000284082029141932,
+      "loss": 3.3196,
+      "step": 48900
+    },
+    {
+      "epoch": 5.277628032345014,
+      "grad_norm": 0.7243698835372925,
+      "learning_rate": 0.00028375822989746354,
+      "loss": 3.3327,
+      "step": 48950
+    },
+    {
+      "epoch": 5.283018867924528,
+      "grad_norm": 0.6648624539375305,
+      "learning_rate": 0.00028343443065299514,
+      "loss": 3.3108,
+      "step": 49000
+    },
+    {
+      "epoch": 5.283018867924528,
+      "eval_accuracy": 0.38090779705074773,
+      "eval_loss": 3.408730983734131,
+      "eval_runtime": 181.6804,
+      "eval_samples_per_second": 99.136,
+      "eval_steps_per_second": 6.198,
+      "step": 49000
+    },
+    {
+      "epoch": 5.288409703504043,
+      "grad_norm": 0.7246460914611816,
+      "learning_rate": 0.0002831106314085267,
+      "loss": 3.3337,
+      "step": 49050
+    },
+    {
+      "epoch": 5.293800539083558,
+      "grad_norm": 0.6399522423744202,
+      "learning_rate": 0.0002827868321640583,
+      "loss": 3.3354,
+      "step": 49100
+    },
+    {
+      "epoch": 5.2991913746630726,
+      "grad_norm": 0.6992799043655396,
+      "learning_rate": 0.0002824630329195898,
+      "loss": 3.3293,
+      "step": 49150
+    },
+    {
+      "epoch": 5.304582210242588,
+      "grad_norm": 0.6970546841621399,
+      "learning_rate": 0.0002821392336751214,
+      "loss": 3.3261,
+      "step": 49200
+    },
+    {
+      "epoch": 5.309973045822103,
+      "grad_norm": 0.6449671387672424,
+      "learning_rate": 0.00028181543443065295,
+      "loss": 3.3052,
+      "step": 49250
+    },
+    {
+      "epoch": 5.315363881401617,
+      "grad_norm": 0.6572911143302917,
+      "learning_rate": 0.00028149163518618455,
+      "loss": 3.3106,
+      "step": 49300
+    },
+    {
+      "epoch": 5.320754716981132,
+      "grad_norm": 0.6900778412818909,
+      "learning_rate": 0.0002811743119266055,
+      "loss": 3.341,
+      "step": 49350
+    },
+    {
+      "epoch": 5.3261455525606465,
+      "grad_norm": 0.6336671710014343,
+      "learning_rate": 0.00028085051268213706,
+      "loss": 3.3116,
+      "step": 49400
+    },
+    {
+      "epoch": 5.331536388140162,
+      "grad_norm": 0.698238730430603,
+      "learning_rate": 0.0002805267134376686,
+      "loss": 3.3287,
+      "step": 49450
+    },
+    {
+      "epoch": 5.336927223719677,
+      "grad_norm": 0.6817668080329895,
+      "learning_rate": 0.00028020291419320016,
+      "loss": 3.3373,
+      "step": 49500
+    },
+    {
+      "epoch": 5.342318059299191,
+      "grad_norm": 0.6314173340797424,
+      "learning_rate": 0.00027987911494873176,
+      "loss": 3.3249,
+      "step": 49550
+    },
+    {
+      "epoch": 5.347708894878706,
+      "grad_norm": 0.6567521691322327,
+      "learning_rate": 0.0002795553157042633,
+      "loss": 3.3204,
+      "step": 49600
+    },
+    {
+      "epoch": 5.353099730458221,
+      "grad_norm": 0.6802349090576172,
+      "learning_rate": 0.0002792315164597949,
+      "loss": 3.3259,
+      "step": 49650
+    },
+    {
+      "epoch": 5.3584905660377355,
+      "grad_norm": 0.6689594388008118,
+      "learning_rate": 0.00027890771721532647,
+      "loss": 3.3023,
+      "step": 49700
+    },
+    {
+      "epoch": 5.363881401617251,
+      "grad_norm": 0.6894006729125977,
+      "learning_rate": 0.00027858391797085807,
+      "loss": 3.3348,
+      "step": 49750
+    },
+    {
+      "epoch": 5.369272237196766,
+      "grad_norm": 0.673187255859375,
+      "learning_rate": 0.0002782601187263896,
+      "loss": 3.3259,
+      "step": 49800
+    },
+    {
+      "epoch": 5.37466307277628,
+      "grad_norm": 0.6820120215415955,
+      "learning_rate": 0.0002779363194819212,
+      "loss": 3.3265,
+      "step": 49850
+    },
+    {
+      "epoch": 5.380053908355795,
+      "grad_norm": 0.6760383248329163,
+      "learning_rate": 0.0002776125202374527,
+      "loss": 3.3287,
+      "step": 49900
+    },
+    {
+      "epoch": 5.38544474393531,
+      "grad_norm": 0.6421095132827759,
+      "learning_rate": 0.00027728872099298433,
+      "loss": 3.3256,
+      "step": 49950
+    },
+    {
+      "epoch": 5.390835579514825,
+      "grad_norm": 0.630399763584137,
+      "learning_rate": 0.0002769649217485159,
+      "loss": 3.3356,
+      "step": 50000
+    },
+    {
+      "epoch": 5.390835579514825,
+      "eval_accuracy": 0.3818138534822327,
+      "eval_loss": 3.403038263320923,
+      "eval_runtime": 181.7655,
+      "eval_samples_per_second": 99.089,
+      "eval_steps_per_second": 6.195,
+      "step": 50000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 92750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 10000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.18034589696e+17,
+  "train_batch_size": 32,
+  "trial_name": null,
+  "trial_params": null
+}